id stringlengths 9 13 | content unknown | decision stringclasses 13
values | reviews listlengths 3 12 | metareview listlengths 1 3 | sentence_texts listlengths 15 712 | opinion_groups listlengths 1 114 | conflicts_validation listlengths 1 155 | rebuttal_validation listlengths 1 155 | opinions listlengths 1 155 | PDF_path stringlengths 36 43 | PDF_version stringclasses 4
values | MD_path stringlengths 34 41 | conference stringclasses 13
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
SyzrLjA5FQ | {
"title": "Selective Self-Training for semi-supervised Learning",
"abstract": "Semi-supervised learning (SSL) is a study that efficiently exploits a large amount of unlabeled data to improve performance in conditions of limited labeled data. Most of the conventional SSL methods assume that the classes of unlabeled... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "Reviewers have concerns about poor writing of the paper, lack of technical novelty, and the methodology taken by the paper not being very principled. "
}
}
]
] | [
"This paper describes the method for performing self-training where the unlabeled datapoints are iteratively added to the training set only if their predictions by the classifier are confident enough. The contributions of this paper are to add datapoints based on the prediction of the confidence level by a separate... | [
[
19
],
[
31
],
[
1,
37
],
[
39
],
[
16
],
[
17
],
[
21
],
[
30
],
[
32,
33
],
[
2
],
[
23
],
[
36
],
[
6
],
[
14
],
[
4
],
[
5
],
[
8
],
[
10
],
[
11
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1,
2
]
},
{
"role": "Author",
... | benchmark/PDF/ICLR2019_SyzrLjA5FQ.pdf | openreview | benchmark/MD/ICLR2019_SyzrLjA5FQ.md | ICLR 2019 |
rJedV3R5tm | {
"title": "RelGAN: Relational Generative Adversarial Networks for Text Generation",
"abstract": "Generative adversarial networks (GANs) have achieved great success at generating realistic images. However, the text generation still remains a challenging task for modern GAN architectures. In this work, we propose Re... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "\npros:\n- well-written and clear\n- good evaluation with convincing ablations\n- moderately novel\n\ncons:\n- Reviewers 1 and 3 feel the paper is somewhat incremental over previous work, combining previously proposed ideas.\n\n(Reviewer 2 originall... | [
"I have read the authors' response and other reviewers' comments. Thanks the authors for taking great effort in answering my questions. Generally, I feel satisfied with the repsonse, and prefer an acceptance recommendation.",
"Contributions:\nThe main contribution of this paper is the proposed RelGAN.",
"First,... | [
[
5,
27
],
[
23
],
[
28
],
[
1,
2,
16,
19
],
[
0
],
[
3,
4
],
[
10
],
[
18
],
[
25
],
[
34
],
[
6
],
[
7,
14
],
[
11
],
[
12
],
[
13
],
[
20
],
[
21
],
[
... | [
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2,
3,
4
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
5
]
... | benchmark/PDF/ICLR2019_rJedV3R5tm.pdf | openreview | benchmark/MD/ICLR2019_rJedV3R5tm.md | ICLR 2019 |
BJxbYoC9FQ | {
"title": "Classifier-agnostic saliency map extraction",
"abstract": "Extracting saliency maps, which indicate parts of the image important to classification, requires many tricks to achieve satisfactory performance when using classifier-dependent methods. Instead, we propose classifier-agnostic saliency map extra... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "{418}; {Classifier-agnostic saliency map extraction}; {Avg: 4.33}; {}\n\n1. Describe the strengths of the paper. As pointed out by the reviewers and based on your expert opinion.\n\nThe paper is well-written and the method is simple, effective, and... | [
"This paper focuses on the extraction of high-quality model-agnostic saliency maps. The authors argue that when an extracted saliency map is directly dependent on a model, then it might not be useful for a different classifier and thus not general enough. To overcome this problem, they consider all the possible cla... | [
[
16
],
[
17
],
[
17
],
[
9
],
[
10
],
[
11
],
[
30
],
[
33
],
[
38
],
[
46
],
[
2,
7,
19,
23,
32,
39
],
[
13
],
[
14
],
[
42
],
[
3,
25
],
[
4
],
[
5
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
2
]
}
],
"category": [
"QUAL-MET"... | benchmark/PDF/ICLR2019_BJxbYoC9FQ.pdf | openreview | benchmark/MD/ICLR2019_BJxbYoC9FQ.md | ICLR 2019 |
ryeaZhRqFm | {
"title": "Link Prediction in Hypergraphs using Graph Convolutional Networks",
"abstract": "Link prediction in simple graphs is a fundamental problem in which new links between nodes are predicted based on the observed structure of the graph. However, in many real-world applications, there is a need to model relat... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9
]
},
"scores... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper describes a method for the link prediction problem in both directed and undirected hypergraphs. While the problem discussed in the paper is clearly importnant and interesting, all reviewers agree that the novelty of the proposed approach... | [
"This paper proposed to use graph convolutional neural networks for link prediction. The authors proposed to use the dual graph to simultaneously learn node and edge embeddings. The label of the edges (positive or negative) are used as supervised signal for training the GCNs. Experiments on a few small data set pro... | [
[
33
],
[
12
],
[
34
],
[
18
],
[
22
],
[
1,
3,
7,
19,
36
],
[
15
],
[
5
],
[
8
],
[
11
],
[
13
],
[
27
],
[
2
],
[
4
],
[
10
],
[
16
],
[
20
],
[
24
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1,
2,
3,
4
]
},
{
... | benchmark/PDF/ICLR2019_ryeaZhRqFm.pdf | openreview | benchmark/MD/ICLR2019_ryeaZhRqFm.md | ICLR 2019 |
S1zlmnA5K7 | {
"title": "Where Off-Policy Deep Reinforcement Learning Fails",
"abstract": "This work examines batch reinforcement learning--the task of maximally exploiting a given batch of off-policy data, without further data collection. We demonstrate that due to errors introduced by extrapolation, standard off-policy deep r... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8
]
},
"scores": {
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper proposes batch-constrained approach to batch RL, where the policy is optimized under the constrain that at a state only actions appearing in the training data are allowed. An extension to continuous cases is given.\n\nWhile the paper has ... | [
"This paper studies extrapolation error in off-policy batch reinforcement learning (RL), where the extrapolation error refers to the overestimation of the value for the state-action pairs that are not in the training data.",
"The authors propose batch-constrained RL, where the policy is optimize under the constra... | [
[
8,
24
],
[
10
],
[
12
],
[
21
],
[
19
],
[
7
],
[
11
],
[
14
],
[
22
],
[
2
],
[
17,
26
],
[
18
],
[
6
],
[
13
],
[
1
],
[
3
],
[
9
],
[
16
],
[
20
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
3,
4,
5
]
},
... | benchmark/PDF/ICLR2019_S1zlmnA5K7.pdf | openreview | benchmark/MD/ICLR2019_S1zlmnA5K7.md | ICLR 2019 |
rJfUCoR5KX | {
"title": "An Empirical study of Binary Neural Networks' Optimisation",
"abstract": "Binary neural networks using the Straight-Through-Estimator (STE) have been shown to achieve state-of-the-art results, but their training process is not well-founded. This is due to the discrepancy between the evaluated function i... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5
]
},
"scores": {
"Solid": null,
"Presentation": null... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper summarizes existing work on binary neural network optimization and performs an empirical study across a few datasets and neural network architectures. I agree with the reviewers that this is a valuable study and it can establish a benchmar... | [
"The authors made several claims and provide suggestions on training binary networks, however, they are not proved or theoretically analyzed. The empirical verification of the proposed hypothesis was viewed as weak as the only two datasets used are small datasets MNIST and CIFAR-10, and the used network architectu... | [
[
4
],
[
15
],
[
20
],
[
0
],
[
21
],
[
23
],
[
10,
13
],
[
18
],
[
1
],
[
2
],
[
3
],
[
5
],
[
11
],
[
14
],
[
16
],
[
22
],
[
6
],
[
7
],
[
8
],
[
9
],
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
},
{
"role": "Author",
"data": [
7,
8,
9,
10,
56
]
}
],
"category": [
"ORIG-ANL",
"QUAL-EXP"
]
},
... | benchmark/PDF/ICLR2019_rJfUCoR5KX.pdf | openreview | benchmark/MD/ICLR2019_rJfUCoR5KX.md | ICLR 2019 |
BkzeUiRcY7 | {
"title": "M^3RL: Mind-aware Multi-agent Management Reinforcement Learning",
"abstract": "Most of the prior work on multi-agent reinforcement learning (MARL) achieves optimal collaboration by directly learning a policy for each agent to maximize a common reward. In this paper, we aim to address this from a differe... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper addresses a variant of multi-agent reinforcement learning that aligns well with real-world applications - it considers the case where agents may have individual, diverging preferences. The proposed approach trains a \"manager\" agent which... | [
"This paper studies the problem of coordinating many strategic agents with private valuation to perform a series of common goals. The algorithm designer is a manager who can assign goals to various agents but cannot see their valuation or control them explicitly. The manager has a utility function for various goals... | [
[
5
],
[
1
],
[
4
],
[
6
],
[
7,
22
],
[
8,
23
],
[
9
],
[
21
],
[
24
],
[
10
],
[
0,
15
],
[
11
],
[
14
],
[
16
],
[
17
],
[
18
],
[
19,
25
],
[
2
],
[
... | [
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2,
3
]
}
],
"category": [
"QUAL-CMP"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
4,
5,
6
... | benchmark/PDF/ICLR2019_BkzeUiRcY7.pdf | openreview | benchmark/MD/ICLR2019_BkzeUiRcY7.md | ICLR 2019 |
S1erHoR5t7 | {
"title": " The relativistic discriminator: a key element missing from standard GAN",
"abstract": "In standard generative adversarial network (SGAN), the discriminator estimates the probability that the input data is real. The generator is trained to increase the probability that fake data is real. We argue that i... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5
]
},
"scores": {
"Solid": null,
"Presentation": null... | [
[
{
"role": "AC",
"data": {
"metareview": "All authors agree that the relativistic discriminator is an interesting idea, and a useful proposal to improve the stability and sample quality of GANs. In earlier drafts there were some clarity issues and missing details, but those have been fixed ... | [
"In this work, the authors considers a variation of GAN by consider simultaneously decrease the probability that real data is real for the generator. To include such a property, the authors propose a relativistic discriminator which estimate the probability that the given real data is more realistic than the fake d... | [
[
1
],
[
3
],
[
6,
11,
13
],
[
7
],
[
9
],
[
5
],
[
8
],
[
14
],
[
2
],
[
12
],
[
4
],
[
0
],
[
10
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"incorrect",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"CLAR-WRT"
]
},
... | benchmark/PDF/ICLR2019_S1erHoR5t7.pdf | openreview | benchmark/MD/ICLR2019_S1erHoR5t7.md | ICLR 2019 |
r1gNni0qtm | {
"title": "Generalized Tensor Models for Recurrent Neural Networks",
"abstract": "Recurrent Neural Networks (RNNs) are very successful at solving challenging problems with sequential data. However, this observed efficiency is not yet entirely explained by theory. It is known that a certain class of multiplicative ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "AR1 finds that extension of the previously presented ICLR'18 paper are interesting and sufficient due to the provided analysis of universality and depth efficiency. AR2 is concerned with the lack of any concrete toy example between the proposed arch... | [
"The authors extend the theoretical results of a paper previously presented in the last edition of ICLR (2018), where it was demonstrated that Recurrent Neural Network can be interpreted as a tensor network decomposition based on the Tensor-Train (TT, Oseledets et al, 2011).",
"While previous results covered the ... | [
[
34
],
[
6
],
[
8
],
[
22
],
[
32
],
[
2,
29
],
[
3
],
[
4
],
[
5
],
[
7
],
[
11
],
[
12
],
[
20
],
[
21,
23
],
[
0
],
[
1,
27
],
[
30
],
[
9
],
[
25
],
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"ORIG-ANL"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"ORIG-ANL"
]
... | benchmark/PDF/ICLR2019_r1gNni0qtm.pdf | openreview | benchmark/MD/ICLR2019_r1gNni0qtm.md | ICLR 2019 |
B1x-LjAcKX | {
"title": "Local Critic Training of Deep Neural Networks",
"abstract": "This paper proposes a novel approach to train deep neural networks by unlocking the layer-wise dependency of backpropagation training. The approach employs additional modules called local critic networks besides the main network model to be tr... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper proposes a new training approach for deep neural interfaces. The idea is to bootstrap from critics of other layers instead of using the final loss as target. The method is evaluated of CIFAR-10 and CIFAR-100 and found to improve performan... | [
"This paper describes a method of training neural networks without update locking. The idea is a small modification on top of Czarnecki et al. Critic training, where instead of using final loss as a critic target, one bootstraps from critics on other layers.",
"In particular, if only one module is present, these ... | [
[
6
],
[
1
],
[
18
],
[
5
],
[
9
],
[
12
],
[
13,
15
],
[
17
],
[
20
],
[
10
],
[
4
],
[
8
],
[
19
],
[
14
],
[
7
],
[
16
],
[
0
],
[
2
],
[
3
],
[
11
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1,
2,
3
]
},
{
"role": "Au... | benchmark/PDF/ICLR2019_B1x-LjAcKX.pdf | openreview | benchmark/MD/ICLR2019_B1x-LjAcKX.md | ICLR 2019 |
Byx93sC9tm | {
"title": "Deep Ensemble Bayesian Active Learning : Adressing the Mode Collapse issue in Monte Carlo dropout via Ensembles",
"abstract": "In image classification tasks, the ability of deep convolutional neural networks (CNNs) to deal with complex image data has proved to be unrivalled. Deep CNNs, however, require ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5
]
},
"scores": {
"Solid": null,
"Presentation": null... | [
[
{
"role": "AC",
"data": {
"metareview": "The reviewers in general found the paper approachable, well written and clear. They noted that the empirical observation of mode collapse in active learning was an interesting insight. However, all the reviewers had concerns with novelty, particul... | [
"This paper introduces a technique using ensembles of models with MC-dropout to perform uncertainty sampling for active learning.",
"In active learning, there is generally a trade-off between data efficiency and computational cost. This paper proposes a combination of existing techniques, not just ensembling neur... | [
[
29
],
[
4,
25
],
[
1,
28
],
[
24
],
[
3,
5,
8,
20
],
[
23
],
[
16
],
[
9
],
[
10,
17,
18
],
[
12
],
[
2
],
[
11
],
[
13
],
[
14
],
[
19,
22
],
[
26,
30
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
},
{
"role": "Author",
"data": [
... | benchmark/PDF/ICLR2019_Byx93sC9tm.pdf | openreview | benchmark/MD/ICLR2019_Byx93sC9tm.md | ICLR 2019 |
BJej72AqF7 | {
"title": "A Max-Affine Spline Perspective of Recurrent Neural Networks",
"abstract": "We develop a framework for understanding and improving recurrent neural networks (RNNs) using max-affine spline operators (MASOs). We prove that RNNs using piecewise affine and convex nonlinearities can be written as a simple pi... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "While the reformulation of RNNs is not practical as it is missing sigmoids and tanhs that are common in LSTMs it does provide an interesting analysis of traditional RNNs and a technique that's novel for many in the ICLR community.\n"
}
}
... | [
"This paper builds upon recent work by Balestriero and Baraniuk (ICML 2018) that concern max-affine spline opertaor (MASO) interpretation of a substantial class of deep networks. In the new paper a special focus is put on Recurrent Neural Networks (RNNs), and it is highlighted based on theoretical considerations le... | [
[
22
],
[
24
],
[
2
],
[
3
],
[
4
],
[
5
],
[
6
],
[
7
],
[
9
],
[
8
],
[
10
],
[
11
],
[
12
],
[
14
],
[
21
],
[
23
],
[
25
],
[
15
],
[
1
],
[
13,
19
],... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"ORIG-MTH"
]
},
... | benchmark/PDF/ICLR2019_BJej72AqF7.pdf | openreview | benchmark/MD/ICLR2019_BJej72AqF7.md | ICLR 2019 |
BJgklhAcK7 | {
"title": "Meta-Learning with Latent Embedding Optimization",
"abstract": "Gradient-based meta-learning techniques are both widely applicable and proficient at solving challenging few-shot learning and fast adaptation problems. However, they have practical difficulties when operating on high-dimensional parameter ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This work builds on MAML by (1) switching from a single underlying set of parameters to a distribution in a latent lower-dimensional space, and (2) conditioning the initial parameter of each subproblem on the input data.\nAll reviewers agree that th... | [
"This work presents an extension of the MAML framework for \"learning to learn.\" This extension changes the space in which \"inner-loop\" gradient steps are taken to adapt the model to a new task, and also introduces stochasticity. The authors validate their proposed method with regression experiments in a toy set... | [
[
3
],
[
6
],
[
12
],
[
16
],
[
17
],
[
18
],
[
14
],
[
1
],
[
4
],
[
19
],
[
2,
5
],
[
8
],
[
9
],
[
7
],
[
11
],
[
15
],
[
0
],
[
10
],
[
13
],
[
20
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1,
2
]
}
],
"category": [
"ORIG-MTH"... | benchmark/PDF/ICLR2019_BJgklhAcK7.pdf | openreview | benchmark/MD/ICLR2019_BJgklhAcK7.md | ICLR 2019 |
HkxjYoCqKX | {
"title": "Relaxed Quantization for Discretized Neural Networks",
"abstract": "Neural network quantization has become an important research area due to its great impact on deployment of large models on resource constrained devices. In order to train networks that can be effectively discretized without loss of perf... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4
]
},
"scores": {
"Solid": null,
"Presentation": null,
"No... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper proposes an effective method to train neural networks with quantized reduced precision. It's fairly straight-forward idea and achieved good results and solid empirical work. reviewers have a consensus on acceptance. "
}
}
]
] | [
"Summary\nThis paper introduces a method for learning neural networks with quantized weights and activations. The main idea is to stochastically – rather than deterministically – quantize values, and to replace the resulting categorical distribution over quantized values with a continuous relaxation (the \"concrete... | [
[
4,
7
],
[
10
],
[
12
],
[
15
],
[
2
],
[
8
],
[
3
],
[
5
],
[
6
],
[
13
],
[
14
],
[
1
],
[
9
],
[
11
],
[
0
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"SIGN-BRD"
]
},
... | benchmark/PDF/ICLR2019_HkxjYoCqKX.pdf | openreview | benchmark/MD/ICLR2019_HkxjYoCqKX.md | ICLR 2019 |
SyMDXnCcF7 | {
"title": "A Mean Field Theory of Batch Normalization",
"abstract": "We develop a mean field theory for batch normalization in fully-connected feedforward neural networks. In so doing, we provide a precise characterization of signal propagation and gradient backpropagation in wide batch-normalized networks at init... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4
]
},
"scores": {
"Solid": null,
"Presentation": null,
"No... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper provides a mean-field-theory analysis of batch normalization. First there is a negative result as to the necessity of gradient explosion when using batch normalization in a fully connected network. They then provide further insights as to... | [
"This paper provides a new dynamic perspective on deep neural network. Based on Gaussian weights and biases, the paper investigates the evolution of the covariance matrix along with the layers.",
"Eventually the matrices achieve a stationary point, i.e., fixed point of the dynamic system. Local performance around... | [
[
2
],
[
4
],
[
5
],
[
9
],
[
1
],
[
8
],
[
10
],
[
0
],
[
11
],
[
6
],
[
3
],
[
7
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
2
]
},
{
"role": "Author",
... | benchmark/PDF/ICLR2019_SyMDXnCcF7.pdf | openreview | benchmark/MD/ICLR2019_SyMDXnCcF7.md | ICLR 2019 |
HJGtFoC5Fm | {
"title": "On the Margin Theory of Feedforward Neural Networks",
"abstract": "Past works have shown that, somewhat surprisingly, over-parametrization can help generalization in neural networks. Towards explaining this phenomenon, we adopt a margin-based perspective. We establish: 1) for multi-layer feedforward rel... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper has received reviews from multiple experts who raise a litany of issues. These have been addressed quite convincingly by the authors, but I believe that ultimately this work needs to go through another round of reviewing, and this cannot ... | [
"This paper studies the implicit bias of minimizers of a regularized cross entropy loss of a two-layer network with ReLU activations. By combining several results, the authors obtain a generalization upper bound which does not increase with the network size.",
"Furthermore, they show that the maximum normalized m... | [
[
6
],
[
9
],
[
21
],
[
27
],
[
46
],
[
47
],
[
2
],
[
24,
25,
33
],
[
26
],
[
29,
32
],
[
39
],
[
41
],
[
3,
42
],
[
4,
28
],
[
5
],
[
30
],
[
35
],
[
37,
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"incorrect",
"correct",
"incorrect",
"incorrect",
"correct",
"incorrect",
"correct",
"correct",
"co... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer",
"data": [
3
]
}
],
"category": [
... | benchmark/PDF/ICLR2019_HJGtFoC5Fm.pdf | openreview | benchmark/MD/ICLR2019_HJGtFoC5Fm.md | ICLR 2019 |
HJGven05Y7 | {
"title": "How to train your MAML",
"abstract": "The field of few-shot learning has recently seen substantial advancements. Most of these advancements came from casting few-shot learning as a meta-learning problem.Model Agnostic Meta Learning or MAML is currently one of the best approaches for few-shot learning vi... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper proposes several improvements for the MAML algorithm that improve its stability and performance.\nStrengths: The improvements are useful for future researchers building upon the MAML algorithm. The results demonstrate a significant improv... | [
"Paper summary - This paper provides a bag of sensible tricks for making MAML more stable, faster to learn, and better in final performance.",
"Quality - The quality of the work is strong: the results demonstrate that tweaks to MAML produce significant improvements in performance.",
"However, I have some concer... | [
[
29
],
[
2,
5
],
[
7,
21
],
[
8,
20
],
[
10
],
[
11
],
[
12
],
[
13
],
[
14
],
[
15
],
[
16
],
[
17
],
[
18
],
[
19
],
[
28
],
[
34
],
[
3
],
[
26
],
[
27,... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"QUAL-EXP"
]
},
... | benchmark/PDF/ICLR2019_HJGven05Y7.pdf | openreview | benchmark/MD/ICLR2019_HJGven05Y7.md | ICLR 2019 |
SkMuPjRcKQ | {
"title": "Feed-forward Propagation in Probabilistic Neural Networks with Categorical and Max Layers",
"abstract": "Probabilistic Neural Networks deal with various sources of stochasticity: input noise, dropout, stochastic neurons, parameter uncertainties modeled as random variables, etc.\nIn this paper we revisit... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1
]
},
"scores": {
"Solid": null,
"Presentation": null,
"Novelty": null,
"Overall": 6,
... | [
[
{
"role": "AC",
"data": {
"metareview": "Reviewers are in a consensus and recommended to accept after engaging with the authors. Please take reviewers' comments into consideration to improve your submission for the camera ready.\n"
}
}
]
] | [
"The main contribution of the paper are methods for propagating approximate uncertainty in neural networks through max and argmax layers. The proposed methods are explained well. The paper is clearly written. The methods are validated in small scale experiments and seem to work well.",
"The proposed approach is n... | [
[
5
],
[
6,
16
],
[
27
],
[
7
],
[
10
],
[
20
],
[
4
],
[
1,
2
],
[
14
],
[
24
],
[
25
],
[
26
],
[
8
],
[
9
],
[
11
],
[
12
],
[
13
],
[
17
],
[
0
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
},
{
"role": "Author",
"data": [
... | benchmark/PDF/ICLR2019_SkMuPjRcKQ.pdf | openreview | benchmark/MD/ICLR2019_SkMuPjRcKQ.md | ICLR 2019 |
BJgGhiR5KX | {
"title": "Learning Cross-Lingual Sentence Representations via a Multi-task Dual-Encoder Model",
"abstract": "A significant roadblock in multilingual neural language modeling is the lack of labeled non-English data. One potential method for overcoming this issue is learning cross-lingual text representations that ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11
... | [
[
{
"role": "AC",
"data": {
"metareview": "Pros:\n\n- A new framework for learning sentence representations\n- Solid experiments and analyses\n- En-Zh / XNLI dataset was added, addressing the comment that no distant languages were considered; also ablation tests\n\nCons:\n\n- The considered... | [
"The paper presents an intuitive architecture for learning cross-lingual sentence representations. I see weaknesses and strengths:",
"(i) The approach is not very novel. Using parallel data and similarity training (siamese, adversarial, etc.) to facilitate transfer has been done before; see [0] and references the... | [
[
9
],
[
10
],
[
45
],
[
16
],
[
17
],
[
22
],
[
24
],
[
31
],
[
32
],
[
33
],
[
13,
35
],
[
1,
11
],
[
18
],
[
19
],
[
21
],
[
39
],
[
40
],
[
2
],
[
3
],
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
},
{
"role": "Author",
"data": [
... | benchmark/PDF/ICLR2019_BJgGhiR5KX.pdf | openreview | benchmark/MD/ICLR2019_BJgGhiR5KX.md | ICLR 2019 |
HylTXn0qYX | {
"title": "Efficiently testing local optimality and escaping saddles for ReLU networks",
"abstract": "We provide a theoretical algorithm for checking local optimality and escaping saddles at nondifferentiable points of empirical risks of two-layer ReLU networks. Our algorithm receives any parameter value and retur... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8
]
},
"scores": {
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper proposes a new method for verifying whether a given point of a two layer ReLU network is a local minima or a second order stationary point and checks for descent directions. All reviewers agree that the algorithm is based on number of new... | [
"The paper proposes a method to check if a given point is a stationary point or not (if not, it provides a descent direction), and then classify stationary points as either local min or second-order stationary. The method works for a specific non-differentiable loss. In the worst case, there can be exponentially m... | [
[
13
],
[
14
],
[
15
],
[
16
],
[
2
],
[
3
],
[
32
],
[
33
],
[
34
],
[
6
],
[
18
],
[
5,
8,
23
],
[
25
],
[
26
],
[
0
],
[
1
],
[
4,
20,
28,
31,
36,
37
]... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
}
],
"category": [
"QUAL-MET"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"QUAL-MET"
]... | benchmark/PDF/ICLR2019_HylTXn0qYX.pdf | openreview | benchmark/MD/ICLR2019_HylTXn0qYX.md | ICLR 2019 |
H1gRM2A5YX | {
"title": "Analysis of Memory Organization for Dynamic Neural Networks",
"abstract": "An increasing number of neural memory networks have been developed, leading to the need for a systematic approach to analyze and compare their underlying memory capabilities. Thus, in this paper, we propose a taxonomy for four po... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5
]
},
"scores": {
"Solid": null,
"Presentation": null... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper presents a taxonomic study of neural network architectures, focussing on those which seek to map onto different part of the hierarchy of models of computation (DFAs, PDAs, etc). The paper splits between defining the taxonomy and comparing... | [
"The authors propose a review-style overview of memory systems within neural networks, from simple RNNs to stack-based memory architectures and NTM / MemNet-style architectures. They propose some reductions to imply how one model can be used (or modify) to simulate another. They then make predictions about which ty... | [
[
14
],
[
15
],
[
1
],
[
2,
6,
7
],
[
12
],
[
18
],
[
19
],
[
20
],
[
25
],
[
27
],
[
28
],
[
30
],
[
3
],
[
8
],
[
9
],
[
24
],
[
4
],
[
21
],
[
10
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
},
{
"role": "Author",
"data": [
... | benchmark/PDF/ICLR2019_H1gRM2A5YX.pdf | openreview | benchmark/MD/ICLR2019_H1gRM2A5YX.md | ICLR 2019 |
HyxCxhRcY7 | {
"title": "Deep Anomaly Detection with Outlier Exposure",
"abstract": "It is important to detect anomalous inputs when deploying machine learning systems. The use of larger and more complex inputs in deep learning magnifies the difficulty of distinguishing between anomalous and in-distribution examples. At the sam... | Accept (Poster) | [
[
{
"role": "Author",
"data": {
"value": {
"comment": [
0,
1
]
}
}
}
],
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
2,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper proposes a new fine-tuning method for improving the performance of existing anomaly detectors.\n\nThe reviewers and AC note the limitation of novelty beyond existing literature.\n\nThis is quite a borader line paper, but AC decided to reco... | [
"In Section 4.3 we observe that a cutting-edge CIFAR-10 density model unexpectedly assigns higher density to SVHN images than to CIFAR-10 images.",
"As it happens, a concurrent submission is based on this observation. Their work can be found here: https://openreview.net/forum?id=H1xwNhCcYm",
"This paper propose... | [
[
1
],
[
7
],
[
14,
16
],
[
19
],
[
6,
11
],
[
2
],
[
3
],
[
8
],
[
17
],
[
18
],
[
20
],
[
21
],
[
13
],
[
10
],
[
4
],
[
12
],
[
9
],
[
0
],
[
5
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
2
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 2",
"data": [
3,
4
]
},
{
"role": "Author",
... | benchmark/PDF/ICLR2019_HyxCxhRcY7.pdf | openreview | benchmark/MD/ICLR2019_HyxCxhRcY7.md | ICLR 2019 |
BJxhijAcY7 | {
"title": "signSGD with Majority Vote is Communication Efficient and Fault Tolerant",
"abstract": "Training neural networks on large datasets can be accelerated by distributing the workload over a network of machines. As datasets grow ever larger, networks of hundreds or thousands of machines become economically v... | Accept (Poster) | [
[
{
"role": "Author",
"data": {
"value": {
"comment": [
0,
1,
2
]
}
}
}
],
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
... | [
[
{
"role": "AC",
"data": {
"metareview": "The Reviewers noticed that the paper undergone many editions and raise concern about the content. They encourage improving experimental section further and strengthening the message of the paper. "
}
}
]
] | [
"Dear anonReviewers,\nHere's a Jupyter notebook in case you'd like to play with the algorithm: https://colab.research.google.com/drive/1PlD2jXoXr2a8e57aIDINCw1-7RIttRTt",
"It can be run in the browser, or you can just download it and run on your machine.\nBest wishes,",
"anonAuthors",
"The paper proposes a di... | [
[
12
],
[
6
],
[
11
],
[
8,
20
],
[
14
],
[
17
],
[
5
],
[
24,
25
],
[
26
],
[
27,
30
],
[
29
],
[
2
],
[
3
],
[
7
],
[
16
],
[
19
],
[
22,
23
],
[
28
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
3,
4
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer",
"data": [
5,
6
]
}
],
"category": [
... | benchmark/PDF/ICLR2019_BJxhijAcY7.pdf | openreview | benchmark/MD/ICLR2019_BJxhijAcY7.md | ICLR 2019 |
B1MbDj0ctQ | {
"title": "Switching Linear Dynamics for Variational Bayes Filtering",
"abstract": "System identification of complex and nonlinear systems is a central problem for model predictive control and model-based reinforcement learning. Despite their complexity, such systems can often be approximated well by a set of line... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The overall view of the reviewers is that the paper is not quite good enough as it stands. The reviewers also appreciates the contributions so taking the comments into account and resubmit elsewhere is encouraged. "
}
}
]
] | [
"This paper proposes a deep probabilistic model for temporal data that leverages latent variables to switch between different learned linear dynamics. The probability distributions are parameterized by deep neural networks and learning is performed end-to-end with amortized variational inference using inference net... | [
[
2
],
[
3,
4
],
[
5
],
[
6
],
[
7
],
[
11
],
[
13
],
[
14
],
[
17
],
[
18
],
[
20
],
[
21
],
[
30
],
[
8,
33
],
[
16
],
[
22
],
[
26
],
[
28
],
[
1
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"ORIG-COM"
]
},
... | benchmark/PDF/ICLR2019_B1MbDj0ctQ.pdf | openreview | benchmark/MD/ICLR2019_B1MbDj0ctQ.md | ICLR 2019 |
H1MgjoR9tQ | {
"title": "CBOW Is Not All You Need: Combining CBOW with the Compositional Matrix Space Model",
"abstract": "Continuous Bag of Words (CBOW) is a powerful text embedding method. Due to its strong capabilities to encode word content, CBOW embeddings perform well on a wide range of downstream tasks while being effici... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3
]
},
"scores": {
"Solid": null,
"Presentation": null,
"Novelty": null,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper presents CMOW—an unsupervised sentence representation learning method that treats sentences as the product of their word matrices. This method is not entirely novel, as the authors acknowledge, but it has not been successfully applied to ... | [
"The main contribution of this paper in practice seems to be a way to initialize the Continuous Matrix Space Model so that training actually converges, followed by a slightly different contrastive loss function used to train these models. The paper explores the pure matrix model and a mixed matrix / vector model, s... | [
[
6
],
[
10
],
[
11
],
[
1
],
[
4,
9
],
[
7
],
[
8
],
[
12
],
[
2
],
[
5
],
[
0
],
[
3
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1,
2
]
},
{
"role": "Author",
... | benchmark/PDF/ICLR2019_H1MgjoR9tQ.pdf | openreview | benchmark/MD/ICLR2019_H1MgjoR9tQ.md | ICLR 2019 |
H1MW72AcK7 | {
"title": "Optimal Control Via Neural Networks: A Convex Approach",
"abstract": "Control of complex systems involves both system identification and controller design. Deep neural networks have proven to be successful in many identification tasks, however, from model-based control perspective, these networks are di... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11
... | [
[
{
"role": "AC",
"data": {
"comment": "The detailed reviews and responses are commendable. Thanks to all.\n\nReviewers: can you comment on whether the revised-paper and author responses have addressed your concerns?\nIn particular, for reviewer 1, this would be important. Note that the r... | [
"The paper proposes neural networks which are convex on inputs data to control problems. These types of networks, constructed based on either MLP or RNN, are shown to have similar representation power as their non-convex versions, thus are potentially able to better capture the dynamics behind complex systems compa... | [
[
6
],
[
9
],
[
19
],
[
17
],
[
30
],
[
2
],
[
7
],
[
20,
21,
36
],
[
22
],
[
10
],
[
31
],
[
32
],
[
33
],
[
35
],
[
4
],
[
34
],
[
1
],
[
3
],
[
14
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
2
]
}
],
"category": [
"QUAL-MET"... | benchmark/PDF/ICLR2019_H1MW72AcK7.pdf | openreview | benchmark/MD/ICLR2019_H1MW72AcK7.md | ICLR 2019 |
SJe8DsR9tm | {
"title": "Dynamic Early Terminating of Multiply Accumulate Operations for Saving Computation Cost in Convolutional Neural Networks",
"abstract": "Deep learning has been attracting enormous attention from academia as well as industry due to its great success in many artificial intelligence applications. As more ap... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6
]
},
"scores": {
"Solid": null,
"Pres... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper proposes a new method for speeding up convolutional neural networks. It uses the idea of early terminating the computation of convolutional layers. It saves FLOPs, but the reviewers raised a critical concern that it doesn't save wall-cloc... | [
"In convolutional neural networks, a convolutional filter conducts a series of multiply-accumulate (MAC) operations, which is computationally heavy. To save computational cost, this manuscript proposes an algorithm to set a checkpoint in the MAC process to determine whether a filter could terminate early based on t... | [
[
18
],
[
7
],
[
8
],
[
9
],
[
10
],
[
13
],
[
16
],
[
17
],
[
22
],
[
24
],
[
6
],
[
19
],
[
2
],
[
15
],
[
3,
12
],
[
11
],
[
14
],
[
20
],
[
21
],
[
23,
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
},
{
"role": "Author",
"data": [
... | benchmark/PDF/ICLR2019_SJe8DsR9tm.pdf | openreview | benchmark/MD/ICLR2019_SJe8DsR9tm.md | ICLR 2019 |
BklCusRct7 | {
"title": "Optimal Transport Maps For Distribution Preserving Operations on Latent Spaces of Generative Models",
"abstract": "Generative models such as Variational Auto Encoders (VAEs) and Generative Adversarial Networks (GANs) are typically trained for a fixed prior distribution in the latent space, such as unifo... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6
]
},
"scores": {
"Solid": null,
"Pres... | [
[
{
"role": "AC",
"data": {
"metareview": "This is a well-written paper that shows how to use optimal transport to perform smooth interpolation, between two random vectors sampled from the prior distribution of the latent space of a deep generative model. By encouraging the marginal of the i... | [
"The paper addresses the latent space distribution mismatch in VAEs and GANs. The authors try to solve the issue by optimal transport theory and the proposed method on the latent space yields better quality in the generated samples.",
"To me, the motivation is not very strong. In DCGAN, amazingly, latent space li... | [
[
14
],
[
5,
8
],
[
13
],
[
2
],
[
3
],
[
6
],
[
9
],
[
10
],
[
11
],
[
12
],
[
1
],
[
0
],
[
4
],
[
7
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1,
2,
3
]
},
{
"role": "Au... | benchmark/PDF/ICLR2019_BklCusRct7.pdf | openreview | benchmark/MD/ICLR2019_BklCusRct7.md | ICLR 2019 |
r1xFE3Rqt7 | {
"title": "Adaptive Mixture of Low-Rank Factorizations for Compact Neural Modeling",
"abstract": "Modern deep neural networks have a large amount of weights, which make them difficult to deploy on computation constrained devices such as mobile phones. One common approach to reduce the model size and computational ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5
]
},
"scores": {
"Solid": null,
"Presentation": null... | [
[
{
"role": "AC",
"data": {
"comment": "Dear Authors and Reviewers,\n\nGiven the disagreements between the reviewers, I took a careful read of the key technical parts of your paper. I have a couple of questions:\n\n1) Is the proposed adaptive mixture of low-rank factorization network trained... | [
"In this paper, the authors propose a compression technique to reduce the number of parameters to learn in a neural network without losing expressiveness.",
"The paper nicely introduces the problem of lack in espressiveness with low-rank factorizations, a well-known technique to reduce the number of parameters in... | [
[
13
],
[
5
],
[
9
],
[
1
],
[
7
],
[
30
],
[
32
],
[
35
],
[
17
],
[
19
],
[
28
],
[
24
],
[
25
],
[
36
],
[
2
],
[
3
],
[
8
],
[
10
],
[
16
],
[
20
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"CLAR-WRT"
]
},
... | benchmark/PDF/ICLR2019_r1xFE3Rqt7.pdf | openreview | benchmark/MD/ICLR2019_r1xFE3Rqt7.md | ICLR 2019 |
Hyl_vjC5KQ | {
"title": "Hierarchical Reinforcement Learning via Advantage-Weighted Information Maximization",
"abstract": "Real-world tasks are often highly structured. Hierarchical reinforcement learning (HRL) has attracted research interest as an approach for leveraging the hierarchical structure of a given task in reinforce... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper proposes a method for hierarchical reinforcement learning that aims to maximize mutual information between options and state-action pairs. The approach and empirical analysis is interesting. The initial submission had many issues with cla... | [
"Revision: The authors addressed most of my concerns and clearly put in effort to improve the paper. The paper explains the central idea better, is more precise in terminology in general, and the additional ablation gives more insight into the relative importance of the advantage weighting. I still think that the r... | [
[
28
],
[
3
],
[
8
],
[
10
],
[
11
],
[
14
],
[
2
],
[
9
],
[
13
],
[
15
],
[
23
],
[
24
],
[
26
],
[
25
],
[
0
],
[
7
],
[
18
],
[
6
],
[
16,
20
],
[
17
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"inc... | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
},
{
"role": "Author",
"data": [
19,
80
]
}
],
"category": [
"QUAL-CMP",
"QUAL-EXP"
]
},
{
"sentences": [
{
"... | benchmark/PDF/ICLR2019_Hyl_vjC5KQ.pdf | openreview | benchmark/MD/ICLR2019_Hyl_vjC5KQ.md | ICLR 2019 |
BJe1E2R5KX | {
"title": "Algorithmic Framework for Model-based Deep Reinforcement Learning with Theoretical Guarantees",
"abstract": "Model-based reinforcement learning (RL) is considered to be a promising approach to reduce the sample complexity that hinders model-free RL. However, the theoretical understanding of such methods... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper proposes model-based reinforcement learning algorithms that have theoretical guarantees. These methods are shown to good results on Mujuco benchmark tasks. All of the reviewers have given a reasonable score to the paper, and the paper can... | [
"This paper proposed a new class of meta-algorithm for reinforcement learning and proved the monotone improvement for a local maximum of the expected reward, which could be used in deep RL setting. The framework seems to be quite general but does not include any specific example, like what non-linear dynamical mode... | [
[
3
],
[
11
],
[
0
],
[
4
],
[
5
],
[
19
],
[
21
],
[
15
],
[
16
],
[
1
],
[
6
],
[
20
],
[
13
],
[
22
],
[
2
],
[
8
],
[
9
],
[
10
],
[
12
],
[
14
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"... | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0,
1,
2
]
},
{
"role": "Author",
"data": [
13,
14,
15,
16,
17,
18,
19,
29,
30,
... | benchmark/PDF/ICLR2019_BJe1E2R5KX.pdf | openreview | benchmark/MD/ICLR2019_BJe1E2R5KX.md | ICLR 2019 |
HkzRQhR9YX | {
"title": "Tree-Structured Recurrent Switching Linear Dynamical Systems for Multi-Scale Modeling",
"abstract": "Many real-world systems studied are governed by complex, nonlinear dynamics. By modeling these dynamics, we can gain insight into how these systems work, make predictions about how they will behave, and... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper presents a recurrent tree-structured linear dynamical system to model the dynamics of a complex nonlinear dynamical system. All reviewers agree that the paper is interesting and useful, and is likely to have an impact in the community. So... | [
"The authors develop a tree structured extension to the recently proposed recurrent switching linear dynamical systems.",
"Like switching linear dynamical systems (sLDS) the proposed models capture non-linear dynamics by switching between a collection of linear regimes.",
"However, unlike SLDS, the transition b... | [
[
16,
21
],
[
1
],
[
0,
12,
20
],
[
8
],
[
13
],
[
6
],
[
14
],
[
18
],
[
22
],
[
2
],
[
3
],
[
5,
9
],
[
15
],
[
17
],
[
4
],
[
7
],
[
10
],
[
11
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
3,
4
]
}
],
... | benchmark/PDF/ICLR2019_HkzRQhR9YX.pdf | openreview | benchmark/MD/ICLR2019_HkzRQhR9YX.md | ICLR 2019 |
Skh4jRcKQ | {
"title": "Understanding Straight-Through Estimator in Training Activation Quantized Neural Nets",
"abstract": "Training activation quantized neural networks involves minimizing a piecewise constant training loss whose gradient vanishes almost everywhere, which is undesirable for the standard back-propagation or c... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7
]
},
"scores": {
"Solid": null,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper contributes to the understanding of straight-through estimation for single hidden layer neural networks, revealing advantages for ReLU and clipped ReLU over identity activations. A thorough and convincing theoretical analysis is provided ... | [
"This paper provides theoretical analysis for two kinds of straight-through estimation (STE) for activation bianrized neural networks. It is theoretically shown that the ReLU STE has better convergence properties the identity STE, by studying the properties of the orientation and norm of the course gradients for S... | [
[
2
],
[
5
],
[
1
],
[
4
],
[
10
],
[
16
],
[
18
],
[
26
],
[
38
],
[
9
],
[
21
],
[
29
],
[
37
],
[
20
],
[
25
],
[
11
],
[
12
],
[
13,
19
],
[
15,
31
],
[... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"QUAL-MET"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
},
{
"role": "Author",
"data"... | benchmark/PDF/ICLR2019_Skh4jRcKQ.pdf | openreview | benchmark/MD/ICLR2019_Skh4jRcKQ.md | ICLR 2019 |
S1gBgnR9Y7 | {
"title": "End-to-end learning of pharmacological assays from high-resolution microscopy images",
"abstract": "Predicting the outcome of pharmacological assays based on high-resolution microscopy\nimages of treated cells is a crucial task in drug discovery which tremendously\nincreases discovery rates. However, en... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This work studies the performance of several end-to-end CNN architectures for the prediction of biomedical assays in microscopy images. One of the architectures, GAPnet, is a minor modification of existing global average pooling (GAP) networks, invo... | [
"The authors explore the possibility of using an end-to-end approach for predicting pharmacological assay outcome using fluorescence microscopy images from the public Cell Painting dataset. In my view, the primary contributions are the following: an interesting and relatively new application (predicting assay outco... | [
[
15
],
[
16
],
[
17
],
[
4
],
[
18
],
[
25,
38
],
[
32
],
[
35
],
[
36
],
[
39
],
[
40
],
[
41
],
[
8
],
[
11
],
[
0,
6,
14
],
[
3,
13
],
[
22
],
[
1,
12,
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
},
{
"role": "Author",
"data": [
26,
27,
28,
29,
30
]
}
],
"category": [
"ORIG-MTH",
"QUAL-CMP"
]
}... | benchmark/PDF/ICLR2019_S1gBgnR9Y7.pdf | openreview | benchmark/MD/ICLR2019_S1gBgnR9Y7.md | ICLR 2019 |
Sklr9i09KQ | {
"title": "Neural Networks for Modeling Source Code Edits",
"abstract": "Programming languages are emerging as a challenging and interesting domain for machine learning. A core task, which has received significant attention in recent years, is building generative models of source code. However, to our knowledge, p... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper focuses on neural network models for source code edits. Compared to prior literature that focused on generative models of source codes, this paper focuses on the generative models of edit sequences of the source code. The paper explores b... | [
"Summary: The authors study building models for edits in source code. The application is obvious: a system to accurately predict what the next edit should be would be very valuable for developers. Here, edits are modeled by two types of sequences: one that tracks the state of all edits at each time step (and is thu... | [
[
8
],
[
6,
24,
34
],
[
18,
32
],
[
25
],
[
27
],
[
28
],
[
30
],
[
44
],
[
16
],
[
1
],
[
35
],
[
9
],
[
23
],
[
15
],
[
33
],
[
7
],
[
14,
21
],
[
10
],
[... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"inc... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"ORIG-COM",
"CLAR... | benchmark/PDF/ICLR2019_Sklr9i09KQ.pdf | openreview | benchmark/MD/ICLR2019_Sklr9i09KQ.md | ICLR 2019 |
rklaWn0qK7 | {
"title": "Learning Neural PDE Solvers with Convergence Guarantees",
"abstract": "Partial differential equations (PDEs) are widely used across the physical and computational sciences. Decades of research and engineering went into designing fast iterative solution methods. Existing solvers are general purpose, but ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9
]
},
"scores... | [
[
{
"role": "AC",
"data": {
"metareview": "Quality: The overall quality of the work is high. The main idea and technical choices are well-motivated, and the method is about as simple as it could be while achieving its stated objectives.\n\nClarity: The writing is clear, with the exception ... | [
"Summary:\nThe authors propose a method to learn and improve problem-tailored PDE solvers from existing ones. The linear updates of the target solver, specified by the problem's geometry and boundary conditions, are computed from the updates of a well-known solver through an optimized linear map. The obtained solv... | [
[
24
],
[
25
],
[
3
],
[
13
],
[
14
],
[
8
],
[
4
],
[
5
],
[
7
],
[
11,
12
],
[
22
],
[
23
],
[
2,
6
],
[
9
],
[
15
],
[
16
],
[
17
],
[
18
],
[
19
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct"... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
2
]
}
],
"category": [
"SIGN-SOT"... | benchmark/PDF/ICLR2019_rklaWn0qK7.pdf | openreview | benchmark/MD/ICLR2019_rklaWn0qK7.md | ICLR 2019 |
rkx1m2C5YQ | {
"title": "Recurrent Kalman Networks: Factorized Inference in High-Dimensional Deep Feature Spaces",
"abstract": "In order to integrate uncertainty estimates into deep time-series modelling, Kalman Filters (KFs) (Kalman et al., 1960) have been integrated with deep learning models. Yet, such approaches typically re... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7
]
},
"scores": {
"Solid": null,
... | [
[
{
"role": "AC",
"data": {
"metareview": "A lot of work has appeared recently on recurrent state space models. So although this paper is in general considered favorable by the reviewers it is unclear exactly how the paper places itself in that (crowded) space. So rejection with a strong enc... | [
"This paper proposes, Recurrent Kalman Network, a modified Kaman filter in which the latent dynamics is projected into a higher dimensional space; efficient inference in this high-dimensional latent space is possible due to the space being locally linear. The state representation, transition, and observation models... | [
[
15,
18
],
[
17
],
[
1,
8
],
[
16
],
[
6
],
[
2
],
[
3
],
[
10
],
[
11
],
[
12
],
[
4
],
[
7
],
[
9
],
[
14
],
[
0
],
[
5
],
[
13
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"CLAR-WRT"
]
},
... | benchmark/PDF/ICLR2019_rkx1m2C5YQ.pdf | openreview | benchmark/MD/ICLR2019_rkx1m2C5YQ.md | ICLR 2019 |
rJxHsjRqFQ | {
"title": "Hyperbolic Attention Networks",
"abstract": "Recent approaches have successfully demonstrated the benefits of learning the parameters of shallow networks in hyperbolic space. We extend this line of work by imposing hyperbolic geometry on the embeddings used to compute the ubiquitous attention mechanisms... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2
]
},
"scores": {
"Solid": null,
"Presentation": null,
"Novelty": null,
"Overa... | [
[
{
"role": "AC",
"data": {
"metareview": "Reviewers all agree that this is a strong submission.\nI also believe it is interesting that only by changing the geometry of embeddings, they can use the space more efficiently without increasing the number of parameters."
}
}
]
] | [
"The authors proposed to exploit hyperbolic geometry in computing the attention mechanisms for neural networks.",
"Specifically, they break the attention read operation into two parts: matching and aggregation. In matching step, they use the hyperbolic distance to quantify the macthing between a query and a key; ... | [
[
4
],
[
8,
12
],
[
9
],
[
6
],
[
14
],
[
1
],
[
5
],
[
7
],
[
11
],
[
16
],
[
3
],
[
15
],
[
10
],
[
13
],
[
0
],
[
2
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"incorrect",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
2
]
},
{
"role": "Author",
... | benchmark/PDF/ICLR2019_rJxHsjRqFQ.pdf | openreview | benchmark/MD/ICLR2019_rJxHsjRqFQ.md | ICLR 2019 |
Syx72jC9tm | {
"title": "Invariant and Equivariant Graph Networks",
"abstract": "Invariant and equivariant networks have been successfully used for learning images, sets, point clouds, and graphs. A basic challenge in developing such networks is finding the maximal collection of invariant and equivariant \\emph{linear} layers. ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9
]
},
"scores... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper provides a comprehensive study and generalisations of previous results on linear permutation invariant and equivariant operators / layers for the case of hypergraph data on multiple node sets. Reviewers indicate that the paper makes a part... | [
"This paper explores maximally expressive linear layers for jointly exchangeable data and in doing so presents a surprisingly expressive model. I have given it a strong accept because the paper takes a very well-studied area (convolutions on graphs) and manages to find a far more expressive model (in terms of numbe... | [
[
12
],
[
11
],
[
0
],
[
3
],
[
4
],
[
5
],
[
14
],
[
17
],
[
18
],
[
2
],
[
7
],
[
8
],
[
9,
10
],
[
13
],
[
15
],
[
1
],
[
6
],
[
16
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
"role": "Reviewer",
"data": [
1,
2
]
}
],
"category": [
"SIGN-D... | benchmark/PDF/ICLR2019_Syx72jC9tm.pdf | openreview | benchmark/MD/ICLR2019_Syx72jC9tm.md | ICLR 2019 |
Hyls7h05FQ | {
"title": "A Differentiable Self-disambiguated Sense Embedding Model via Scaled Gumbel Softmax",
"abstract": "We present a differentiable multi-prototype word representation model that disentangles senses of polysemous words and produces meaningful sense-specific embeddings without external resources. It jointly l... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "\nPros:\n\n* High quality evaluation across different benchmarks, plus human eval\n\n* The paper is well written (though one could quibble about the motivation for the method, see Cons)\n\nCons:\n\n* The approach is incremental, the main contribu... | [
"This paper proposes GASI to disambiguate different sense identities and learn sense representations given contextual information.",
"The main idea is to use scaled Gumbel softmax as the sense selection method instead of soft or hard attention, which is the novelty and contribution of this paper.",
"In addition... | [
[
13
],
[
16
],
[
28
],
[
11
],
[
15
],
[
17
],
[
18
],
[
20
],
[
21
],
[
1
],
[
5
],
[
7
],
[
27
],
[
2
],
[
3,
24
],
[
4
],
[
6
],
[
12
],
[
14
],
[
23
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"inc... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2,
3
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
4,
5,
6,
... | benchmark/PDF/ICLR2019_Hyls7h05FQ.pdf | openreview | benchmark/MD/ICLR2019_Hyls7h05FQ.md | ICLR 2019 |
B1GHJ3R9tQ | {
"title": "HyperGAN: Exploring the Manifold of Neural Networks",
"abstract": "We introduce HyperGAN, a generative network that learns to generate all the weight parameters of deep neural networks. HyperGAN first transforms low dimensional noise into a latent space, which can be sampled from to obtain diverse, per... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "All of the reviewers find this paper to contain interesting ideas. Originally, clarity was a major issue, although a few issues remain (see the comments of reviewer 3). The reviewers believe that the paper has been substantially improved from its or... | [
"TL;DR. I find the manuscript to contain interesting ideas, yet I believe there is room for improvement.",
"* Summary\nFor any given specific network architecture, the manuscript aims at learning a distribution over the weights (rather than point-wise estimates of the weights). This is achieved through using a tw... | [
[
51
],
[
12
],
[
15,
28
],
[
16
],
[
17
],
[
18,
48
],
[
19
],
[
37
],
[
45
],
[
47
],
[
49
],
[
53
],
[
42
],
[
55
],
[
0
],
[
6
],
[
7
],
[
8
],
[
9,
29
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
}
],
"category": [
"CLAR-WRT"
]
},
{
"sentences": [
{
"role": "Reviewer",
"data": [
1
]
}
],
"category": [
"N/A"
]
},
... | benchmark/PDF/ICLR2019_B1GHJ3R9tQ.pdf | openreview | benchmark/MD/ICLR2019_B1GHJ3R9tQ.md | ICLR 2019 |
H1xD9sR5Fm | {
"title": "Minimum Divergence vs. Maximum Margin: an Empirical Comparison on Seq2Seq Models",
"abstract": "Sequence to sequence (seq2seq) models have become a popular framework for neural sequence prediction. While traditional seq2seq models are trained by Maximum Likelihood Estimation (MLE), much recent work has ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The reviewers agree that the paper is worthy of publication at ICLR, hence I recommend accept.\n\nRegarding section 4.3 of the submission and the claim that this paper presents the first insight for existing work from a divergence minimization pers... | [
"In this paper the authors distinguish between two families of training objectives for seq2seq models, namely, divergence minimization objectives and max-margin objectives. They primarily focus on the divergence minimization family, and show that the MRT and RAML objectives can be related to minimizing the KL diver... | [
[
13
],
[
14
],
[
17
],
[
18
],
[
20
],
[
21
],
[
23
],
[
27
],
[
10
],
[
16
],
[
22
],
[
25,
26,
28
],
[
1,
4
],
[
2
],
[
7
],
[
5
],
[
8
],
[
30
],
[
12
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"inc... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
2,
3,
4,
5,
6
... | benchmark/PDF/ICLR2019_H1xD9sR5Fm.pdf | openreview | benchmark/MD/ICLR2019_H1xD9sR5Fm.md | ICLR 2019 |
HyGcghRct7 | {
"title": "Random mesh projectors for inverse problems",
"abstract": "We propose a new learning-based approach to solve ill-posed inverse problems in imaging. We address the case where ground truth training samples are rare and the problem is severely ill-posed---both because of the underlying physics and because ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper proposes a novel method of solving inverse problems that avoids direct inversion by first reconstructing various piecewise-constant projections of the unknown image (using a different CNN to learn each) and then combining them via optim... | [
"This paper describes a novel method for solving inverse problems in imaging.",
"The basic idea of this approach is use the following steps:",
"1. initialize with nonnegative least squares solution to inverse problem (x0)",
"2. compute m different projections of x0",
"3. estimate x from the m different proj... | [
[
12
],
[
4
],
[
11
],
[
30
],
[
6
],
[
18,
24
],
[
20
],
[
22
],
[
25
],
[
26
],
[
5
],
[
8
],
[
13,
15
],
[
9
],
[
16
],
[
27,
29
],
[
32
],
[
7
],
[
14
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"incorrect",
"incorrect",
"incorrect",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"co... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer",
"data": [
2
]
}
],
"category": [
"N/A"
]
... | benchmark/PDF/ICLR2019_HyGcghRct7.pdf | openreview | benchmark/MD/ICLR2019_HyGcghRct7.md | ICLR 2019 |
SkEqro0ctQ | {
"title": "Hierarchical interpretations for neural network predictions",
"abstract": "Deep neural networks (DNNs) have achieved impressive predictive performance due to their ability to learn complex, non-linear relationships between variables. However, the inability to effectively visualize these relationships ha... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10
]
},... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper receives a unanimous accept over reviewers, though some concerns on novelty exist. So it is suggested to be a probable accept. "
}
}
]
] | [
"This paper proposes a novel approach to explain neural network predictions by learning hierarchical representations of groups of input features and their contribution to the final prediction. The proposed method is a straightforward extension of the contextual decomposition work by (Murdoch et. al. 2018) which est... | [
[
1
],
[
5
],
[
14
],
[
20
],
[
3,
10,
15,
17,
21
],
[
12
],
[
23
],
[
6
],
[
7
],
[
13
],
[
16
],
[
19,
22,
24
],
[
4
],
[
0
],
[
2
],
[
8
],
[
9
],
[
... | [
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
3
]
}
],
"category": [
... | benchmark/PDF/ICLR2019_SkEqro0ctQ.pdf | openreview | benchmark/MD/ICLR2019_SkEqro0ctQ.md | ICLR 2019 |
rkgW0oA9FX | {
"title": "Graph HyperNetworks for Neural Architecture Search",
"abstract": "Neural architecture search (NAS) automatically finds the best task-specific neural network topology, outperforming many manual architecture designs. However, it can be prohibitively expensive as the search requires training thousands of d... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper proposes an architecture search method based on graph hypernetworks (GHN). The core idea is that given a candidate architecture, GHN predicts its weights (similar to SMASH), which allows for fast evaluation w/o training the architecture fr... | [
"This paper proposes to accelerate architecture search by replacing the expensive inner loop (wherein candidate architectures are trained to completion) with a HyperNetwork which predicts the weights of candidate architectures, as in SMASH.",
"Contrary to SMASH, this work employs a Graph neural network to allow f... | [
[
9
],
[
14
],
[
15
],
[
18
],
[
2
],
[
8
],
[
19
],
[
29
],
[
1
],
[
6
],
[
7
],
[
10
],
[
11
],
[
12
],
[
13
],
[
20
],
[
25
],
[
31
],
[
23
],
[
5
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
2,
3
]
}
],
"category": [
... | benchmark/PDF/ICLR2019_rkgW0oA9FX.pdf | openreview | benchmark/MD/ICLR2019_rkgW0oA9FX.md | ICLR 2019 |
H1gsz30cKX | {
"title": "Fixup Initialization: Residual Learning Without Normalization",
"abstract": "Normalization layers are a staple in state-of-the-art deep neural network architectures. They are widely believed to stabilize training, enable higher learning rate, accelerate convergence and improve generalization, though the... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8
]
},
"scores": {
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper explores the effect of normalization and initialization in residual networks, motivated by the need to avoid exploding and vanishing activations and gradients. Based on some theoretical analysis of stepsizes in SGD, the authors propose a s... | [
"This paper shows that with a clever initialization method ResNets can be trained without using batch-norm (and other normalization techniques). The network can still reach state-of-the-art performance.",
"The authors propose a new initialization method called \"ZeroInit\" and use it to train very deep ResNets (... | [
[
23
],
[
13
],
[
16
],
[
20
],
[
29
],
[
1
],
[
11
],
[
14
],
[
15
],
[
17
],
[
19
],
[
22
],
[
33
],
[
3
],
[
7
],
[
9,
30
],
[
35
],
[
28
],
[
8
],
[
34
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"incorrect",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
3
]
}
],
"category": [
... | benchmark/PDF/ICLR2019_H1gsz30cKX.pdf | openreview | benchmark/MD/ICLR2019_H1gsz30cKX.md | ICLR 2019 |
rkeMHjR9Ym | {
"title": "Stochastic Gradient Descent Learns State Equations with Nonlinear Activations",
"abstract": "We study discrete time dynamical systems governed by the state equation $h_{t+1}=ϕ(Ah_t+Bu_t)$. Here A,B are weight matrices, ϕ is an activation function, and $u_t$ is the input data. This relation is the backbo... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper shows convergence of stochastic gradient descent for the problem of learning weight matrices for a linear dynamical system with non-linear activation. Reviewers agree that the problem considered is both interesting and challenging. How... | [
"The paper studies discrete time dynamical systems with a non-linear state equation. They assume the non-linear function is assumed to be \\beta-increasing like leaky ReLU. Under this setting, the authors prove that for the given state equation for stable systems with random gaussian input at each time step, runni... | [
[
10
],
[
29
],
[
24,
31
],
[
25
],
[
28
],
[
1
],
[
3
],
[
11
],
[
13
],
[
9
],
[
12
],
[
16
],
[
17
],
[
20
],
[
26,
30
],
[
27
],
[
4,
21
],
[
5
],
[
6,
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1,
2
]
},
{
"role": "Author",
... | benchmark/PDF/ICLR2019_rkeMHjR9Ym.pdf | openreview | benchmark/MD/ICLR2019_rkeMHjR9Ym.md | ICLR 2019 |
B1gstsCqt7 | {
"title": "Sparse Dictionary Learning by Dynamical Neural Networks",
"abstract": "A dynamical neural network consists of a set of interconnected neurons that interact over time continuously. It can exhibit computational properties in the sense that the dynamical system’s evolution and/or limit points in the associ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3
]
},
"scores": {
"Solid": null,
"Presentation": null,
"Novelty": null,
... | [
[
{
"role": "AC",
"data": {
"metareview": "While there has been lots of previous work on training dictionaries for sparse coding, this work tackles the problem of doing son in a purely local way. While previous work suggests that the exact computation of gradient addressed in the paper is no... | [
"This paper proposes a dynamical neural network for sparse coding where all the interactions terms are learned. In previous approaches (Rozell et al.) some weights were tied to the others. Here the network consists of feedforward, lateral, and feedback weights, all of which have their own learning rule. The auth... | [
[
1
],
[
10
],
[
0
],
[
3
],
[
11
],
[
13
],
[
14
],
[
2
],
[
12
],
[
5,
7
],
[
8
],
[
4
],
[
6
],
[
9
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect"
] | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
},
{
"role": "Author",
"dat... | benchmark/PDF/ICLR2019_B1gstsCqt7.pdf | openreview | benchmark/MD/ICLR2019_B1gstsCqt7.md | ICLR 2019 |
HJeRkh05Km | {
"title": "Visual Semantic Navigation using Scene Priors",
"abstract": "How do humans navigate to target objects in novel scenes? Do we use the semantic/functional priors we have built over years to efficiently search and navigate? For example, to search for mugs, we search cabinets near the coffee machine and for... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3
]
},
"scores": {
"Solid": null,
"Presentation": null,
"Novelty": null,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The authors propose an approach for visual navigation that leverages a semantic knowledge graph to ground and inform the policy of an RL agent. The agent uses a graphnet to learn relationships and support the navigation. The empirical protocol is so... | [
"This work proposes to use semantic knowledge about the relationships and functionality of different objects, to help in navigation tasks, in both familiar and unfamiliar situations. The paper is very well written and it is clear what the authors did. The approach seems sound, and while it combines two existing app... | [
[
6
],
[
0
],
[
13
],
[
1,
3
],
[
11,
17
],
[
2
],
[
5
],
[
14
],
[
9
],
[
12
],
[
16
],
[
15
],
[
8
],
[
4
],
[
7
],
[
10
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
}
],
"category": [
"CLAR-WRT",
"QUAL-MET",
"ORIG-COM",
"QUAL-EXP"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1,
... | benchmark/PDF/ICLR2019_HJeRkh05Km.pdf | openreview | benchmark/MD/ICLR2019_HJeRkh05Km.md | ICLR 2019 |
r1xdH3CcKX | {
"title": "Stochastic Prediction of Multi-Agent Interactions from Partial Observations",
"abstract": "We present a method which learns to integrate temporal information, from a learned dynamics model, with ambiguous visual information, from a learned vision model, in the context of interacting agents. Our method i... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper proposes a unified approach for performing state estimation and future forecasting for agents interacting within a multi-agent system. The method relies on a graph-structured recurrent neural network trained on temporal and visual (pixel)... | [
"1) Summary\nThis paper presents a graph neural network based architecture that is trained to locate and model the interactions of agents in an environment directly from pixels. They propose an architecture that is a composition of recurrent neural networks where each models a single object independently and commun... | [
[
6
],
[
9
],
[
24
],
[
14
],
[
19
],
[
22
],
[
26
],
[
2,
7
],
[
5
],
[
1
],
[
12
],
[
4
],
[
8
],
[
15
],
[
25
],
[
28,
29
],
[
3
],
[
13
],
[
16
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"ORIG-MTH"
]
},
... | benchmark/PDF/ICLR2019_r1xdH3CcKX.pdf | openreview | benchmark/MD/ICLR2019_r1xdH3CcKX.md | ICLR 2019 |
Hke-JhA9Y7 | {
"title": "Learning concise representations for regression by evolving networks of trees",
"abstract": "We propose and study a method for learning interpretable representations for the task of regression. Features are represented as networks of multi-type expression trees comprised of activation functions common i... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The reviewers all feel that the paper should be accepted to the conference. The main strengths that they noted were the quality of writing, the wide applicability of the proposed method and the strength of the empirical evaluation. It's nice to se... | [
"The paper proposes a method for learning regression models through evolutionary",
"algorithms that promise to be more interpretable than other models while",
"achieving similar or higher performance. The authors evaluate their approach on",
"99 datasets from OpenML, demonstrating very promising performance."... | [
[
5
],
[
17
],
[
7,
18,
19
],
[
2
],
[
3
],
[
4
],
[
8
],
[
9
],
[
10,
13
],
[
12
],
[
16
],
[
15
],
[
20
],
[
0
],
[
1
],
[
6
],
[
11
],
[
14
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer",
"data": [
3,
4,
5,
6,
... | benchmark/PDF/ICLR2019_Hke-JhA9Y7.pdf | openreview | benchmark/MD/ICLR2019_Hke-JhA9Y7.md | ICLR 2019 |
H1eSS3CcKX | {
"title": "Stochastic Optimization of Sorting Networks via Continuous Relaxations",
"abstract": "Sorting input objects is an important step in many machine learning pipelines. However, the sorting operator is non-differentiable with respect to its inputs, which prohibits end-to-end gradient-based optimization. In ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10
]
},... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper proposes a general-purpose continuous relaxation of the output of the sorting operator. This enables end-to-end training to enable more efficient stochastic optimization over the combinatorially large space of permutations.\n\nIn the subm... | [
"In many machine learning applications, sorting is an important step such as ranking.",
"However, the sorting operator is not differentiable with respect to its inputs. The main idea of the paper is to introduce a continuous relaxation of the sorting operator in order to construct an end-to-end gradient-based opt... | [
[
3
],
[
4
],
[
6
],
[
15
],
[
2,
14
],
[
5
],
[
7
],
[
11
],
[
12
],
[
13
],
[
22
],
[
1
],
[
16
],
[
23
],
[
17
],
[
19
],
[
10
],
[
20
],
[
18
],
[
9
]... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2,
3
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
4
]
}
],
"cat... | benchmark/PDF/ICLR2019_H1eSS3CcKX.pdf | openreview | benchmark/MD/ICLR2019_H1eSS3CcKX.md | ICLR 2019 |
HyGEM3C9KQ | {
"title": "Improving Differentiable Neural Computers Through Memory Masking, De-allocation, and Link Distribution Sharpness Control",
"abstract": "The Differentiable Neural Computer (DNC) can learn algorithmic and question answering tasks. An analysis of its internal activation patterns reveals three problems: Mos... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3
]
},
"scores": {
"Solid": null,
"Presentation": null,
"Novelty": null,
... | [
[
{
"role": "AC",
"data": {
"metareview": "\npros:\n- Identification of several interesting problems with the original DNC model: masked attention, erasion of de-allocated elements, and sharpened temporal links\n- An improved architecture which addresses the issues and shows improved perform... | [
"The authors propose three improvements to the DNC model: masked attention, erasion of de-allocated elements, and sharpened temporal links --- and show that this allows the model to solve synthetic memory tasks faster and with better precision. They also show the model performs better on average on bAbI than the or... | [
[
12
],
[
10
],
[
11
],
[
13
],
[
2
],
[
9
],
[
15
],
[
1
],
[
6
],
[
7
],
[
4
],
[
5
],
[
8
],
[
17
],
[
16
],
[
0
],
[
3
],
[
14
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"ORIG-MTH"
]
},
... | benchmark/PDF/ICLR2019_HyGEM3C9KQ.pdf | openreview | benchmark/MD/ICLR2019_HyGEM3C9KQ.md | ICLR 2019 |
rJlRKjActQ | {
"title": "Manifold Mixup: Learning Better Representations by Interpolating Hidden States",
"abstract": "Deep networks often perform well on the data distribution on which they are trained, yet give incorrect (and often very confident) answers when evaluated on points from off of the training distribution. This is... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper contains useful information and shows relative improvements compared to mixup. However, some of the main claims are not substantiated enough to be fully convincing. For example, the claims that manifold mixup can prevent can manifold colli... | [
"TL;DR. a generalization of the mixup algorithm to any layer, improving generalization abilities.",
"* Summary\nThe manuscript generalizes the mixup algorithm (Zhang et al., 2017) which proposed to interpolate between inputs to yield better generalization. The present manuscript addresses a fairly more general se... | [
[
29
],
[
41
],
[
9
],
[
11
],
[
25
],
[
35
],
[
3
],
[
12
],
[
39
],
[
6
],
[
7
],
[
14
],
[
17
],
[
19
],
[
5
],
[
15,
22
],
[
16,
23
],
[
18
],
[
27
],
[... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer",
"data": [
1
]
}
],
"category": [
"N/A"
]
},
{
... | benchmark/PDF/ICLR2019_rJlRKjActQ.pdf | openreview | benchmark/MD/ICLR2019_rJlRKjActQ.md | ICLR 2019 |
BJx0sjC5FX | {
"title": "RNNs implicitly implement tensor-product representations",
"abstract": "Recurrent neural networks (RNNs) can learn continuous vector representations of symbolic structures such as sequences and sentences; these representations often exhibit linear regularities (analogies). Such regularities motivate ou... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7
]
},
"scores": {
"Solid": null,
... | [
[
{
"role": "AC",
"data": {
"metareview": "AR1 seeks the paper to be more standalone and easier to read. As this comment comes from the reviewer who is very experienced in tensor models, it is highly recommended that the authors make further efforts to make the paper easier to follow. AR2 is... | [
"This paper presents an analysis of popularly-use RNN model for structure modeling abilities by designing Tensor Product Decomposition Networks to approximate the encoder. The results show that the representations exhibit interpretable compositional structure. To provide better understanding, the paper evaluates th... | [
[
1
],
[
3
],
[
4
],
[
8
],
[
15
],
[
9
],
[
10
],
[
11
],
[
7
],
[
2
],
[
5
],
[
6
],
[
13
],
[
12,
14
],
[
0
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"CLAR-WRT",
"SIGN... | benchmark/PDF/ICLR2019_BJx0sjC5FX.pdf | openreview | benchmark/MD/ICLR2019_BJx0sjC5FX.md | ICLR 2019 |
S1lhbnRqF7 | {
"title": "Building Dynamic Knowledge Graphs from Text using Machine Reading Comprehension",
"abstract": "We propose a neural machine-reading model that constructs dynamic knowledge graphs from procedural text. It builds these graphs recurrently for each step of the described procedure, and uses them to track the ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4
]
},
"scores": {
"Solid": null,
"Presentation": null,
"No... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper investigates a new approach to machine reading for procedural text, where the task of reading comprehension is formulated as dynamic construction of a procedural knowledge graph. The proposed model constructs a recurrent knowledge graph (... | [
"The paper addresses a challenging problem of predicting the states of entities over the description of a process. The paper is very well written, and easily understandable. The authors propose a graph structure for entity states, which is updated at each step using the outputs of a machine comprehension system. Th... | [
[
3
],
[
8
],
[
10
],
[
7,
14
],
[
9
],
[
12
],
[
17
],
[
18
],
[
19
],
[
22
],
[
1
],
[
2
],
[
4,
11
],
[
20
],
[
21
],
[
13
],
[
6
],
[
0
],
[
5
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
},
{
"role": "Author",
"data": [
... | benchmark/PDF/ICLR2019_S1lhbnRqF7.pdf | openreview | benchmark/MD/ICLR2019_S1lhbnRqF7.md | ICLR 2019 |
S1zk9iRqF7 | {
"title": "PATE-GAN: Generating Synthetic Data with Differential Privacy Guarantees",
"abstract": "Machine learning has the potential to assist many communities in using the large datasets that are becoming more and more available. Unfortunately, much of that potential is not being realized because it would requir... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3
]
},
"scores": {
"Solid": null,
"Presentation": null,
"Novelty": null,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper improves upon the PATE-GAN framework for differentially-private synthetic data generation. They eliminate the need for public data samples for training the GAN, by providing a distribution which can be sampled from instead.\n\nThe authors... | [
"The paper studies the problem of generating synthetic datasets (while ensuring differential privacy) via training a GAN. One natural approach is the teacher-student framework considered in the PATE framework. In the original PATE framework, while the teachers are ensured to preserve differential privacy, the stud... | [
[
5
],
[
10
],
[
11
],
[
15
],
[
0
],
[
6
],
[
8
],
[
9
],
[
14
],
[
2
],
[
3
],
[
7
],
[
13
],
[
1
],
[
4
],
[
12
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"SIGN-BRD"
]
... | benchmark/PDF/ICLR2019_S1zk9iRqF7.pdf | openreview | benchmark/MD/ICLR2019_S1zk9iRqF7.md | ICLR 2019 |
SyVU6s05K7 | {
"title": "Deep Frank-Wolfe For Neural Network Optimization",
"abstract": "Learning a deep neural network requires solving a challenging optimization problem: it is a high-dimensional, non-convex and non-smooth minimization problem with a large number of terms. The current practice in neural network optimization i... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3
]
},
"scores": {
"Solid": null,
"Presentation": null,
"Novelty": null,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper was judged by the reviewers as providing interesting ideas, well-written and potentially having impact on future research on NN optimization. The authors are asked to make sure they addressed reviewers comments clearly in the paper."
... | [
"This paper introduced a proximal approach to optimize neural networks by linearizing the network output instead of the loss function. They demonstrate their algorithm on multi-class hinge loss, where they can show that optimal step size can be computed in close form without significant additional cost. Their exper... | [
[
9
],
[
10
],
[
0,
7
],
[
3
],
[
12
],
[
1
],
[
2
],
[
11
],
[
5
],
[
6
],
[
13
],
[
4
],
[
8
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1,
2
]
},
{
"role": "Author",
... | benchmark/PDF/ICLR2019_SyVU6s05K7.pdf | openreview | benchmark/MD/ICLR2019_SyVU6s05K7.md | ICLR 2019 |
rylIAsCqYm | {
"title": "A2BCD: Asynchronous Acceleration with Optimal Complexity",
"abstract": "\tIn this paper, we propose the Asynchronous Accelerated Nonuniform Randomized Block Coordinate Descent algorithm (A2BCD). We prove A2BCD converges linearly to a solution of the convex minimization problem at the same rate as NU_ACD... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8
]
},
"scores": {
... | [
[
{
"role": "AC",
"data": {
"metareview": "The reviewers all agreed that this paper makes a strong contribution to ICLR by providing the first asynchronous analysis of a Nesterov-accelerated coordinate descent method."
}
}
]
] | [
"In distributed optimisation, it is well known that asynchronous methods outperform synchronous methods in many cases.",
"However, the questions as to whether (and when) asynchronous methods can be shown to have any speed-up, as the number of nodes increases, has been open. The paper under review answers the ques... | [
[
7
],
[
8
],
[
2
],
[
3
],
[
1
],
[
9
],
[
4
],
[
5
],
[
6
],
[
11
],
[
12
],
[
13
],
[
0
],
[
10
],
[
14
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
2,
3,
4,
5
]
},
... | benchmark/PDF/ICLR2019_rylIAsCqYm.pdf | openreview | benchmark/MD/ICLR2019_rylIAsCqYm.md | ICLR 2019 |
B1l08oAct7 | {
"title": "Deterministic Variational Inference for Robust Bayesian Neural Networks",
"abstract": "Bayesian neural networks (BNNs) hold great promise as a flexible and principled solution to deal with uncertainty when learning from finite data. Among approaches to realize probabilistic inference in deep neural netw... | Accept (Oral) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The manuscript proposes deterministic approximations for Bayesian neural networks as an alternative to the standard Monte-Carlo approach. The results suggest that the deterministic approximation can be more accurate than previous methods. Some expli... | [
"Summary:\nThis work is tackling two difficulties in current VB applied to DNNs (\"Bayes by backprop\").",
"First, MC approximations of intractable expectations are replaced by deterministic approximations.",
"While this has been done before, the solution here is new and very interesting.",
"Second, a Gaussia... | [
[
5
],
[
4
],
[
7
],
[
6
],
[
1
],
[
10,
35
],
[
9
],
[
11,
29,
30
],
[
14
],
[
22
],
[
24
],
[
25
],
[
26
],
[
28
],
[
31
],
[
33
],
[
3,
21,
36
],
[
13,
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1,
2
]
}
],
"category": [
"ORIG-MTH"... | benchmark/PDF/ICLR2019_B1l08oAct7.pdf | openreview | benchmark/MD/ICLR2019_B1l08oAct7.md | ICLR 2019 |
ryxepo0cFX | {
"title": "AntisymmetricRNN: A Dynamical System View on Recurrent Neural Networks",
"abstract": "Recurrent neural networks have gained widespread use in modeling sequential data. Learning long-term dependencies using these models remains difficult though, due to exploding or vanishing gradients. In this paper, we ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3
]
},
"scores": {
"Solid": null,
"Presentation": null,
"Novelty": null,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper presents a novel idea with a compelling experimental study. Good paper, accept."
}
}
]
] | [
"This is an interesting paper which proposes a novel angle on the problem of learning long-term dependencies in recurrent nets. The authors argue that most of the action should be in the imaginary part of the eigenvalues of the Jacobian J=F' of the new_state = old_state + epsilon F(old_state, input) incremental typ... | [
[
5,
6,
11
],
[
16
],
[
20
],
[
22
],
[
7
],
[
14
],
[
0
],
[
1
],
[
3
],
[
8
],
[
4
],
[
9
],
[
23
],
[
15
],
[
2
],
[
13
],
[
17,
24,
25
],
[
10
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
},
{
"role": "Author",
"data"... | benchmark/PDF/ICLR2019_ryxepo0cFX.pdf | openreview | benchmark/MD/ICLR2019_ryxepo0cFX.md | ICLR 2019 |
Syfz6sC9tQ | {
"title": "Generative Feature Matching Networks",
"abstract": "We propose a non-adversarial feature matching-based approach to train generative models. Our approach, Generative Feature Matching Networks (GFMN), leverages pretrained neural networks such as autoencoders and ConvNet classifiers to perform feature ext... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2
]
},
"scores": {
"Solid": null,
"Presentation": null,
"Novelty": null,
"Overa... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper proposes a method of training implicit generative models based on moment matching in the feature spaces of pre-trained feature extractors, derived from autoencoders or classifiers. The authors also propose a trick for tracking the moving a... | [
"The paper introduces Generative Feature Matching Networks (GFMNs) which is a non-adversarial approach to train generative models based on feature matching. GFMN uses pretrained neural networks such as Autoencoders (AE) and Deep Convolutional Neural Networks (DCNN) to extract features. Equation (1) is the proposed ... | [
[
1,
22
],
[
2,
19
],
[
4
],
[
8
],
[
9
],
[
13
],
[
16
],
[
6
],
[
17,
30
],
[
20
],
[
23
],
[
24
],
[
10
],
[
11
],
[
5
],
[
15
],
[
27
],
[
28
],
[
29,
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"CLAR-WRT",
"QUAL... | benchmark/PDF/ICLR2019_Syfz6sC9tQ.pdf | openreview | benchmark/MD/ICLR2019_Syfz6sC9tQ.md | ICLR 2019 |
H1g0Z3A9Fm | {
"title": "Supervised Community Detection with Line Graph Neural Networks",
"abstract": "Community detection in graphs can be solved via spectral methods or posterior inference under certain probabilistic graphical models. Focusing on random graph families such as the stochastic block model, recent research has un... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7
]
},
"scores": {
"Solid": null,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper introduces a new graph convolutional neural network, called LGNN, and applied it to solve the community detection problem. The reviewers think LGNN yields a nice and useful extension of graph CNN, especially in using the line graph of edg... | [
"This paper presents a study of the community detection problem via graph neural networks. The presented results open the possibility that neural networks are able to discover the optimal algorithm for a given task. This is rather convincingly demonstrated on the example of the stochastic block model, where the opt... | [
[
3
],
[
5
],
[
4
],
[
7
],
[
2,
9
],
[
11
],
[
12
],
[
13
],
[
15
],
[
0
],
[
1
],
[
14
],
[
6
],
[
8
],
[
10
],
[
16
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
}
],
"category": [
"QUAL-MET",
"QUAL-EXP"
]
},
{
"sentences": [
{
"role": "Reviewer",
"data": [
1
]
}
],
"category": [
... | benchmark/PDF/ICLR2019_H1g0Z3A9Fm.pdf | openreview | benchmark/MD/ICLR2019_H1g0Z3A9Fm.md | ICLR 2019 |
BJemQ209FQ | {
"title": "Learning to Navigate the Web",
"abstract": "Learning in environments with large state and action spaces, and sparse rewards, can hinder a Reinforcement Learning (RL) agent’s learning through trial-and-error. For instance, following natural language instructions on the Web (such as booking a flight ticke... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "All reviewers (including those with substantial expertise in RL) were solid in their praise for this paper that is also tackling an interesting application that is much less well studied but deserves attention.\n\n"
}
}
]
] | [
"The paper propose a framework to deal with large state and action",
"spaces with sparse rewards in reinforcement learning.",
"In particular,\nthey propose to use a meta-learner to generate experience to the agent",
"and to decompose the learning task into simpler sub-tasks. The authors",
"train a DQN with ... | [
[
15
],
[
18
],
[
3
],
[
4,
22
],
[
5
],
[
11
],
[
13
],
[
14
],
[
16
],
[
19
],
[
7
],
[
8
],
[
9
],
[
12
],
[
1
],
[
2,
20
],
[
10
],
[
21
],
[
17
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2,
3,
4,
5,
6,
7,
8
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1"... | benchmark/PDF/ICLR2019_BJemQ209FQ.pdf | openreview | benchmark/MD/ICLR2019_BJemQ209FQ.md | ICLR 2019 |
Hke20iA9Y7 | {
"title": "Efficient Training on Very Large Corpora via Gramian Estimation",
"abstract": "We study the problem of learning similarity functions over very large corpora using neural network embedding models. These models are typically trained using SGD with random sampling of unobserved pairs, with a sample size th... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6
]
},
"scores": {
"Solid": null,
"Pres... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper presents methods to scale learning of embedding models estimated using neural networks. The main idea is to work with Gram matrices whose sizes depend on the length of the embedding. Building upon existing works like SAG algorithm, the pa... | [
"This paper proposes a method for estimating non-linear similarities between items using Gramian estimation. This is achieved by having two separate neural networks defined for each item to be compared, which are then combined via a dot product. The proposed innovation in this paper is to use Gramian estimation for... | [
[
1,
4
],
[
5,
9
],
[
2
],
[
7
],
[
8
],
[
11
],
[
0
],
[
3
],
[
6
],
[
10
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
2
]
},
{
"role": "Author",
... | benchmark/PDF/ICLR2019_Hke20iA9Y7.pdf | openreview | benchmark/MD/ICLR2019_Hke20iA9Y7.md | ICLR 2019 |
BygNqoR9tm | {
"title": "Sinkhorn AutoEncoders",
"abstract": "Optimal Transport offers an alternative to maximum likelihood for learning generative autoencoding models. We show how this principle dictates the minimization of the Wasserstein distance between the encoder aggregated posterior and the prior, plus a reconstruction e... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6
]
},
"scores": {
"Solid": null,
"Pres... | [
[
{
"role": "AC",
"data": {
"metareview": "The reviewers appreciated the contribution of combining Wasserstein Autoencoders with the Sinkhorn algorithm.\n\nYet R4 as well as the author of the WAE paper (Ilya Tolstikhin) both expressed concerns about the empirical evaluation.\n\nWhile R1-R3 w... | [
"The paper proposes a new representation of Wasserstein AutoEncoder and provides the formal analysis of learning autoencoders with optimal transport theory. The proposed model, SAE, employs the constraints on the equality of prior and posterior latent spaces with a Sinkhorn distance.",
"Moreover, the proposed mod... | [
[
29
],
[
10
],
[
24
],
[
2
],
[
9
],
[
12
],
[
16
],
[
17
],
[
31
],
[
22
],
[
3
],
[
5
],
[
19
],
[
21
],
[
26,
28
],
[
30
],
[
38
],
[
32
],
[
34
],
[
4
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"QUAL-MET"
]
},
... | benchmark/PDF/ICLR2019_BygNqoR9tm.pdf | openreview | benchmark/MD/ICLR2019_BygNqoR9tm.md | ICLR 2019 |
SygD-hCcF7 | {
"title": "Dimensionality Reduction for Representing the Knowledge of Probabilistic Models",
"abstract": "Most deep learning models rely on expressive high-dimensional representations to achieve good performance on tasks such as classification. However, the high dimensionality of these representations makes them d... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4
]
},
"scores": {
"Solid": null,
"Presentation": null,
"No... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper introduces an approach for reducing the dimensionality of training data examples in a way that preserves information about soft target probabilistic representations provided by a teacher model, with applications such as zero-shot learning... | [
"Summary:\nThis paper introduces a new supervised dimensionality reduction model. Supervision is provided in the form of class probabilities and the learning algorithm learns low-dimensional representations such that posterior cluster assignment probabilities given the representations match the observed class proba... | [
[
1
],
[
2
],
[
6
],
[
12
],
[
7
],
[
4,
5,
9,
13
],
[
10
],
[
11
],
[
15
],
[
8
],
[
0
],
[
3
],
[
14
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
2,
3
]
},
{
"role": "Au... | benchmark/PDF/ICLR2019_SygD-hCcF7.pdf | openreview | benchmark/MD/ICLR2019_SygD-hCcF7.md | ICLR 2019 |
HkgqFiAcFm | {
"title": "Marginal Policy Gradients: A Unified Family of Estimators for Bounded Action Spaces with Applications",
"abstract": "Many complex domains, such as robotics control and real-time strategy (RTS) games, require an agent to learn a continuous control. In the former, an agent learns a policy over R^d and in ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper introduces a new variance reduced policy gradient method, for directional and clipped action spaces, with provable guarantees that the gradient is lower variance. The paper is clearly written and the theory an important contribution. The e... | [
"Summary\nThis paper derives a new policy gradient method for when continuous actions are transformed by a",
"normalization step, a process called angular policy gradients (APG). A generalization based on",
"a certain class of transformations is presented. The method is an instance of a",
"Rao-Blackwellizatio... | [
[
6
],
[
7
],
[
9
],
[
13
],
[
20
],
[
2
],
[
5
],
[
10
],
[
12
],
[
26
],
[
23
],
[
3
],
[
8,
21
],
[
25
],
[
4
],
[
11
],
[
16
],
[
17
],
[
18
],
[
24
]... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer",
"data": [
2,
3
]
}
],
"category": [
... | benchmark/PDF/ICLR2019_HkgqFiAcFm.pdf | openreview | benchmark/MD/ICLR2019_HkgqFiAcFm.md | ICLR 2019 |
Syxt5oC5YQ | {
"title": "Aggregated Momentum: Stability Through Passive Damping",
"abstract": "Momentum is a simple and widely used trick which allows gradient-based optimizers to pick up speed along low curvature directions. Its performance depends crucially on a damping coefficient. Largecamping coefficients can potentially ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9
]
},
"scores... | [
[
{
"role": "AC",
"data": {
"metareview": "Dear authors,\n\nReviewers liked the idea of your new optimizer and found the experiments convincing. However, they also would have liked to get better insights on the place of AggMo in the existing optimization literature. Given that the related wo... | [
"The authors combined several update steps together to achieve aggregated momentum. They showed that it is more stable than the other momentum methods.",
"Also, in Auto-encoder and image classification, AggMo outperforms than the other methods.",
"Pros:\n(+) Theoretical result is shown on the quadratic problem... | [
[
14
],
[
7
],
[
6
],
[
12
],
[
13
],
[
16,
18
],
[
1
],
[
11
],
[
15
],
[
5
],
[
9
],
[
2
],
[
3,
17
],
[
4
],
[
0
],
[
8
],
[
10
]
] | [
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"SIGN-SOT"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
2
]
}
],
"category": [
"ORIG... | benchmark/PDF/ICLR2019_Syxt5oC5YQ.pdf | openreview | benchmark/MD/ICLR2019_Syxt5oC5YQ.md | ICLR 2019 |
H1gMCsAqY7 | {
"title": "Slimmable Neural Networks",
"abstract": "We present a simple and general method to train a single neural network executable at different widths (number of channels in a layer), permitting instant and adaptive accuracy-efficiency trade-offs at runtime. Instead of training individual networks with differe... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5
]
},
"scores": {
"Solid": null,
"Presentation": null... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper proposed a method that creates neural networks that can run under different resource constraints. The reviewers have consensus on accept. The pro is that the paper is novel and provides a practical approach to adjust model for different c... | [
"This paper trains a single network executable at different widths. This is implemented by maintaining separate BN parameter and statistics for different width. The problem is well-motivated and the proposed method can be very helpful for deployment of deep models to devices with varying capacity and computational ... | [
[
2
],
[
9
],
[
11
],
[
0,
6
],
[
1
],
[
3
],
[
4
],
[
7
],
[
8
],
[
10
],
[
5
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
}
],
"category": [
"QUAL-MET"
]
},
{
"sentences": [
{
"role": "Reviewer",
"data": [
1,
2
]
}
],
"category": [
"QUAL... | benchmark/PDF/ICLR2019_H1gMCsAqY7.pdf | openreview | benchmark/MD/ICLR2019_H1gMCsAqY7.md | ICLR 2019 |
rke_YiRct7 | {
"title": "Small nonlinearities in activation functions create bad local minima in neural networks",
"abstract": "We investigate the loss surface of neural networks. We prove that even for one-hidden-layer networks with \"slightest\" nonlinearity, the empirical risks have spurious local minima in most cases. Our r... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This is an interesting paper that develops new techniques for analyzing the loss surface of deep networks, allowing the existence of spurious local minima to be established under fairly general conditions. The reviewers responded with uniformly pos... | [
"The authors provide a clean and easily understood sufficient",
"condition for spurious local minima to exist in networks with",
"a hidden layer using ReLUs or leaky ReLUs. This condition,",
"that there is not linear transformation with zero loss,",
"is satisfied for almost all inputs with more examples th... | [
[
2
],
[
6
],
[
10
],
[
17
],
[
3
],
[
9
],
[
5,
11
],
[
7
],
[
13
],
[
19
],
[
8
],
[
14
],
[
21
],
[
0
],
[
1
],
[
4
],
[
12
],
[
15
],
[
16
],
[
18
],
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer",
"data": [
2,
3,
4,
5
]
}
... | benchmark/PDF/ICLR2019_rke_YiRct7.pdf | openreview | benchmark/MD/ICLR2019_rke_YiRct7.md | ICLR 2019 |
B1xJAsA5F7 | {
"title": "Learning Multimodal Graph-to-Graph Translation for Molecule Optimization",
"abstract": "We view molecule optimization as a graph-to-graph translation problem. The goal is to learn to map from one molecular graph to another with better properties based on an available corpus of paired molecules. Since mo... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The revisions made by the authors convinced the reviewers to all recommend accepting this paper. Therefore, I am recommending acceptance as well. I believe the revisions were important to make since I concur with several points in the initial review... | [
"As a reviewer I am expert in learning in structured data domains.",
"The paper proposes a quite complex system, involving many different choices and components, for obtaining chemical compounds with improved properties starting from a given corpora.",
"Overall presentation is good, although some details/explan... | [
[
7
],
[
9
],
[
6
],
[
8
],
[
1
],
[
27
],
[
35
],
[
24
],
[
3
],
[
13
],
[
14
],
[
15
],
[
17
],
[
19
],
[
20
],
[
29
],
[
37
],
[
2
],
[
4
],
[
11
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
2,
3
]
},
{
"role": "Au... | benchmark/PDF/ICLR2019_B1xJAsA5F7.pdf | openreview | benchmark/MD/ICLR2019_B1xJAsA5F7.md | ICLR 2019 |
r1GbfhRqF7 | {
"title": "Kernel Change-point Detection with Auxiliary Deep Generative Models",
"abstract": "Detecting the emergence of abrupt property changes in time series is a challenging problem. Kernel two-sample test has been studied for this task which makes fewer assumptions on the distributions than traditional paramet... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8
]
},
"scores": {
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper proposes a new kernel learning framework for change point detection by using a generative model. The reviewers agree that the paper is interesting and useful for the community. One of the reviewer had some issues with the paper but those ... | [
"The manuscript entitled \"Kernel Change-Point Detection with Auxiliary Deep Generative Models\" describes a novel approach to optimising the choice of kernel towards increased testing power in this challenging machine learning problem. The proposed method is shown to offer improvements over alternatives on a set ... | [
[
7
],
[
3,
4
],
[
0
],
[
1
],
[
6
],
[
8
],
[
9
],
[
5
],
[
2
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0,
1,
2,
3,
4,
5,
6,
7
]
},
{
"role": "Author",
"data": [
10,
11,
12,
13,
... | benchmark/PDF/ICLR2019_r1GbfhRqF7.pdf | openreview | benchmark/MD/ICLR2019_r1GbfhRqF7.md | ICLR 2019 |
S1lqMn05Ym | {
"title": "Information asymmetry in KL-regularized RL",
"abstract": "Many real world tasks exhibit rich structure that is repeated across different parts of the state space or in time. In this work we study the possibility of leveraging such repeated structure to speed up and regularize learning. We start from the... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11
... | [
[
{
"role": "AC",
"data": {
"comment": "Thanks for the detailed review comments thus far.\nDo the authors wish to add anything or respond in any way?\n-- area chair"
}
}
],
[
{
"role": "AC",
"data": {
"metareview": "Strengths\n\nThe paper introduces a pr... | [
"-- Originality --\nThis paper studies how to use KL-regularization with information asymmetry to speed up and improve reinforcement learning (RL). Compared with existing work, the major novelty in the proposed algorithm is that it uses a default policy learned from data, rather than a fixed default policy.",
"Mo... | [
[
2,
16
],
[
8
],
[
10
],
[
15
],
[
0,
1,
4
],
[
9
],
[
13
],
[
17
],
[
6
],
[
12
],
[
18
],
[
7
],
[
3
],
[
5
],
[
14
],
[
11
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"ORIG-MTH"
]... | benchmark/PDF/ICLR2019_S1lqMn05Ym.pdf | openreview | benchmark/MD/ICLR2019_S1lqMn05Ym.md | ICLR 2019 |
S1xq3oR5tQ | {
"title": "A Unified Theory of Early Visual Representations from Retina to Cortex through Anatomically Constrained Deep CNNs",
"abstract": "The vertebrate visual system is hierarchically organized to process visual information in successive stages. Neural representations vary drastically across the first stages of... | Accept (Oral) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper advocates neuroscience-based V1 models to adapt CNNs. The results of the simulations are convincing from a neuroscience-perspective. The reviewers equivocally recommend publication."
}
}
]
] | [
"This paper addresses questions about the representation of visual information in the retina. The authors create a deep neural network model of the visual system in which a single parameter (bandwidth between the “retina” and “visual cortex” parts) is sufficient to qualitatively reproduce retinal receptive fields o... | [
[
4
],
[
5
],
[
11
],
[
14
],
[
1
],
[
10
],
[
2
],
[
3
],
[
6
],
[
15
],
[
7
],
[
8
],
[
9
],
[
16
],
[
12
],
[
13
],
[
0
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1,
2
]
}
],
"category": [
"ORIG-COM"... | benchmark/PDF/ICLR2019_S1xq3oR5tQ.pdf | openreview | benchmark/MD/ICLR2019_S1xq3oR5tQ.md | ICLR 2019 |
S1gOpsCctm | {
"title": "Learning Finite State Representations of Recurrent Policy Networks",
"abstract": "Recurrent neural networks (RNNs) are an effective representation of control policies for a wide range of reinforcement and imitation learning problems. RNN policies, however, are particularly difficult to explain, understa... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper addresses the problem of interpreting recurrent neural networks by quantizing their states an mapping them onto a Moore Machine. The paper presents some interesting results on reinforcement learning and other tasks. I believe the experimen... | [
"RNNs are difficult to explain, understand and analyze due to the continuous-valued memory vectors and observations features they use.",
"Thus, this paper attempts to extract finite representation from RNNs so as to better interpret or understand RNNs. They introduce a new technique called Quantized Bottleneck In... | [
[
4
],
[
16
],
[
25
],
[
3
],
[
15
],
[
2,
7,
23
],
[
6,
28
],
[
8
],
[
29
],
[
10
],
[
11
],
[
13
],
[
14
],
[
17
],
[
27
],
[
30
],
[
0
],
[
18
],
[
24
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
}
],
"category": [
"SIGN-BRD"
]
},
{
"sentences": [
{
"role": "Reviewer",
"data": [
1
]
}
],
"category": [
"N/A"
]
},
... | benchmark/PDF/ICLR2019_S1gOpsCctm.pdf | openreview | benchmark/MD/ICLR2019_S1gOpsCctm.md | ICLR 2019 |
r1Gsk3R9Fm | {
"title": "Shallow Learning For Deep Networks",
"abstract": "Shallow supervised 1-hidden layer neural networks have a number of favorable properties that make them easier to interpret, analyze, and optimize than their deep counterparts, but lack their representational power. Here we use 1-hidden layer learning pr... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper discusses layer-wise training of deep networks. The authors show that it's possible to achieve reasonable performance by training deep nets layer by layer, as opposed to now widely adopted end-to-end training. While such a training procedu... | [
"Summary:\nThis paper proposes layer wise training of neural networks using classification auxiliary tasks for training each layer. Experiments are presented on CIFAR10 and Imagenet. Accuracies close to end to end training are obtained.",
"The layer wise training is repeated for J steps, the auxiliary tasks are t... | [
[
33
],
[
9
],
[
12
],
[
13
],
[
23
],
[
35
],
[
22
],
[
26
],
[
24,
29
],
[
31
],
[
32
],
[
36
],
[
2,
3
],
[
4
],
[
5,
6
],
[
7
],
[
8
],
[
16
],
[
17
]... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2,
3,
4
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
5
]
},
... | benchmark/PDF/ICLR2019_r1Gsk3R9Fm.pdf | openreview | benchmark/MD/ICLR2019_r1Gsk3R9Fm.md | ICLR 2019 |
rkemqsC9Fm | {
"title": "Information Theoretic lower bounds on negative log likelihood",
"abstract": "In this article we use rate-distortion theory, a branch of information theory devoted to the problem of lossy compression, to shed light on an important problem in latent variable modeling of data: is there room to improve the ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9
]
},
"scores... | [
[
{
"role": "AC",
"data": {
"metareview": "Strengths: This paper gives a detailed treatment of the connections between rate distortion theory and variational lower bounds, culminating in a practical diagnostic tool. The paper is well-written.\n\nWeaknesses: Many of the theoretical results... | [
"This is an interesting paper that studies the latent variable modeling from an information theoretic perspective.",
"Specifically, the authors argue that the rate-distortion theory for lossy compression provides a natural toolkit for studying latent variable models, and they propose a lower bound (also a gap fun... | [
[
1
],
[
9
],
[
15
],
[
17
],
[
19
],
[
23
],
[
0
],
[
2
],
[
4
],
[
12
],
[
13
],
[
14
],
[
25
],
[
21
],
[
27
],
[
24
],
[
28,
30
],
[
6
],
[
8
],
[
10
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"incorrect",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2
]
}
],
"category": [
"CLAR-WRT"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
3
]
},
{
"role"... | benchmark/PDF/ICLR2019_rkemqsC9Fm.pdf | openreview | benchmark/MD/ICLR2019_rkemqsC9Fm.md | ICLR 2019 |
BygREjC9YQ | {
"title": "A unified theory of adaptive stochastic gradient descent as Bayesian filtering",
"abstract": "We formulate stochastic gradient descent (SGD) as a novel factorised Bayesian filtering problem, in which each parameter is inferred separately, conditioned on the corresopnding backpropagated gradient. Infere... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The aim of this paper is to interpret various optimizers such as RMSprop, Adam, and NAG, as approximate Kalman filtering of the optimal parameters. These algorithms are derived as inference procedures in various dynamical systems. The main empirical... | [
"In this work, the authors attempt to unify existing adaptive gradient methods under the Bayesian filtering framework with the dynamical prior. In Ollivier, 2017, a framework is proposed to connect Bayesian filtering and natural gradient.",
"On the other hand, in Khan et al., 2018. an approach is proposed to co... | [
[
5
],
[
9
],
[
30,
42
],
[
48,
59,
60,
63,
64,
65,
87
],
[
53,
62,
113,
114
],
[
61
],
[
68,
77
],
[
69
],
[
71
],
[
73
],
[
74
],
[
76
],
[
81
],
[
82
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
2
]
},
{
"role": "Author",
... | benchmark/PDF/ICLR2019_BygREjC9YQ.pdf | openreview | benchmark/MD/ICLR2019_BygREjC9YQ.md | ICLR 2019 |
S1fQSiCcYm | {
"title": "Understanding and Improving Interpolation in Autoencoders via an Adversarial Regularizer",
"abstract": "Autoencoders provide a powerful framework for learning compressed representations by encoding all of the information needed to reconstruct a data point in a latent code. In some cases, autoencoders ca... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9
]
},
"scores... | [
[
{
"role": "AC",
"data": {
"metareview": "The reviewers have reached a consensus that this paper is very interesting and add insights into interpolation in autoencoders."
}
}
]
] | [
"This paper proposed an adversarially regularized AE algorithm that improve interpolation in latent space.",
"Specifically, a critic is used to predict the interpolation weight \\alpha and encourage the interpolated images to be more realistic. The paper verified the method on a newly proposed synthetic line benc... | [
[
5
],
[
21
],
[
4
],
[
10
],
[
20
],
[
2
],
[
1
],
[
19
],
[
22
],
[
23
],
[
27
],
[
32
],
[
36
],
[
37
],
[
38
],
[
13
],
[
14
],
[
17,
29
],
[
30
],
[
33... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"incorrect",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
2
]
}
],
"category": [
"ORIG-MTH"... | benchmark/PDF/ICLR2019_S1fQSiCcYm.pdf | openreview | benchmark/MD/ICLR2019_S1fQSiCcYm.md | ICLR 2019 |
S1lg0jAcYm | {
"title": "ARM: Augment-REINFORCE-Merge Gradient for Stochastic Binary Networks",
"abstract": "To backpropagate the gradients through stochastic binary layers, we propose the augment-REINFORCE-merge (ARM) estimator that is unbiased, exhibits low variance, and has low computational complexity. Exploiting variable a... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7
]
},
"scores": {
"Solid": null,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper introduces a new way to estimate gradients of expectations of discrete random variables by introducing antithetic noise samples for use in a control variate.\n\nQuality: The experiments are mostly appropriate, although I disagree with th... | [
"For binary layers, how to calculate and backpropagate gradients is a big problem, particularly for the binary neural networks. To solve the problem, this paper proposes an unbiased and low variance augment-REINFORCE-merge (ARM) estimator. With the help of an appropriate reparameterization, the antithetic sampling ... | [
[
4
],
[
18
],
[
42
],
[
44
],
[
45
],
[
30
],
[
33
],
[
37
],
[
41
],
[
1,
10,
24
],
[
11
],
[
12
],
[
16
],
[
19
],
[
23
],
[
25
],
[
31
],
[
32
],
[
35
],
... | [
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"correc... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"CLAR-WRT"
]
},
... | benchmark/PDF/ICLR2019_S1lg0jAcYm.pdf | openreview | benchmark/MD/ICLR2019_S1lg0jAcYm.md | ICLR 2019 |
r1laEnA5Ym | {
"title": "A Variational Inequality Perspective on Generative Adversarial Networks",
"abstract": "Generative adversarial networks (GANs) form a generative modeling approach known for producing appealing samples, but they are notably difficult to train. One common way to tackle this issue has been to propose new fo... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8
]
},
"scores": {
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper presents a variational inequality perspective on the optimization problem arising in GANs. Convergence of stochastic gradient descent methods (averaging and extragradient variants) is given under monotonicity (or convex) assumptions. In pa... | [
"Overall, the paper is well-written and of high quality, therefore I recommend acceptance.",
"Pros:\n+ The work gives an accessible but still rigorous introduction to the literature on VIs which I find highly valuable, as it creates a bridge between the classical mathematical programming literature and applicatio... | [
[
0
],
[
1
],
[
4
],
[
5
],
[
13
],
[
2
],
[
15
],
[
10
],
[
16
],
[
3
],
[
14
],
[
7,
8
],
[
12
],
[
17
],
[
6
],
[
9
],
[
11
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
}
],
"category": [
"CLAR-WRT"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"CLAR-WRT",
... | benchmark/PDF/ICLR2019_r1laEnA5Ym.pdf | openreview | benchmark/MD/ICLR2019_r1laEnA5Ym.md | ICLR 2019 |
SJxu5iR9KQ | {
"title": "Learning to Schedule Communication in Multi-agent Reinforcement Learning",
"abstract": "Many real-world reinforcement learning tasks require multiple agents to make sequential decisions under the agents’ interaction, where well-coordinated actions among the agents are crucial to achieve the target goal ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5
]
},
"scores": {
"Solid": null,
"Presentation": null... | [
[
{
"role": "AC",
"data": {
"metareview": "The authors present a learnt scheduling mechanism for managing communications in bandwidth-constrained, contentious multi-agent RL domains. This is well-positioned in the rapidly advancing field of MARL and the contribution of the paper is both nove... | [
"# overview\nThis paper focuses on multi-agent reinforcement learning tasks that require communication between the agents, and further presupposes that the communication protocol is bandwidth constrained and contentious so that a scheduling mechanism is necessary. To address this they introduce a new learned weigh... | [
[
3,
7,
14
],
[
5
],
[
15
],
[
1
],
[
4
],
[
10,
12
],
[
17
],
[
8,
11
],
[
9
],
[
16
],
[
2
],
[
0
],
[
6
],
[
13
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
},
{
"role": "Author",
"data": [
... | benchmark/PDF/ICLR2019_SJxu5iR9KQ.pdf | openreview | benchmark/MD/ICLR2019_SJxu5iR9KQ.md | ICLR 2019 |
rkgpy3C5tX | {
"title": "Amortized Bayesian Meta-Learning",
"abstract": "Meta-learning, or learning-to-learn, has proven to be a successful strategy in attacking problems in supervised learning and reinforcement learning that involve small amounts of data. State-of-the-art solutions involve learning an initialization and/or lea... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper combines two ideas: MAML, and the hierarchical Bayesian inference approach of Amit and Meir (2018). The idea is fairly straightforward but well-motivated, and it seems to work well in practice. The paper is well-written and includes good... | [
"This work proposes an adaptation to MAML-type models that accounts for posterior uncertainty in task specific latent variables. This is achieved via a hierarchical Bayesian view of MAML, employing variational inference for the task-specific parameters. The key intuition of this paper is that one can perform fast a... | [
[
11
],
[
12
],
[
2
],
[
3,
32
],
[
22
],
[
9
],
[
4,
8,
26,
33,
38
],
[
14
],
[
5
],
[
15
],
[
17
],
[
18
],
[
19,
30
],
[
24,
35
],
[
27
],
[
16
],
[
23
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"QUAL-MET"
]
},
... | benchmark/PDF/ICLR2019_rkgpy3C5tX.pdf | openreview | benchmark/MD/ICLR2019_rkgpy3C5tX.md | ICLR 2019 |
ByMHvs0cFQ | {
"title": "Quaternion Recurrent Neural Networks",
"abstract": "Recurrent neural networks (RNNs) are powerful architectures to model sequential data, due to their capability to learn short and long-term dependencies between the basic elements of a sequence. Nonetheless, popular tasks such as speech or images recogn... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The authors derive and experiment with quaternion-based recurrent neural networks, and demonstrate their effectiveness on speech recognition tasks (TIMIT and WSJ), where the authors demonstrate that the proposed models can achieve the same accuracy ... | [
"After the discussion with authors, I am happy to recommend acceptance.",
"1.\tIn “Consequently, for each input vector of size N, output vector of size M, dimensions are split into four parts: the first one equals to r, the second is xi, the third one equals to yj, and the last one to zk to compose a quaternion Q... | [
[
1
],
[
3
],
[
4
],
[
15
],
[
20
],
[
11
],
[
16
],
[
19
],
[
22
],
[
9
],
[
17,
23
],
[
5
],
[
6
],
[
7
],
[
8
],
[
14
],
[
24
],
[
2
],
[
10
],
[
18
],... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
},
{
"role": "Author",
"data": [
... | benchmark/PDF/ICLR2019_ByMHvs0cFQ.pdf | openreview | benchmark/MD/ICLR2019_ByMHvs0cFQ.md | ICLR 2019 |
SkVhlh09tX | {
"title": "Pay Less Attention with Lightweight and Dynamic Convolutions",
"abstract": "Self-attention is a useful mechanism to build generative models for language and images. It determines the importance of context elements by comparing each element to the current time step. In this paper, we show that a very lig... | Accept (Oral) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5
]
},
"scores": {
"Solid": null,
"Presentation": null... | [
[
{
"role": "AC",
"data": {
"metareview": "Very solid work, recognized by all reviewers as worthy of acceptance. Additional readers also commented and there is interest in the open source implementation that the authors promise to provide."
}
}
]
] | [
"The paper proposes a convolutional alternative to self-attention. To achieve this, the number of parameters of a typical convolution operation is first reduced by using a depth-wise approach (i.e. convolving only within each channel), and then further reduced by tying parameters across layers in a round-robin fash... | [
[
1
],
[
2,
3
],
[
5
],
[
0
],
[
13
],
[
17
],
[
6
],
[
8
],
[
9
],
[
10
],
[
11
],
[
14
],
[
15
],
[
16
],
[
18
],
[
7
],
[
4
],
[
12
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
3
]
}
],
"category": ... | benchmark/PDF/ICLR2019_SkVhlh09tX.pdf | openreview | benchmark/MD/ICLR2019_SkVhlh09tX.md | ICLR 2019 |
BygfghAcYX | {
"title": "The role of over-parametrization in generalization of neural networks",
"abstract": "Despite existing work on ensuring generalization of neural networks in terms of scale sensitive complexity measures, such as norms, margin and sharpness, these complexity measures do not offer an explanation of why neur... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "I agree with the reviewers that this is a strong contribution and provides new insights, even if it doesn't quite close the problem. \n\np.s.: It seems that centering the weight matrices at initialization is a key idea. The authors note that Dziugai... | [
"It is shown empirically that common algorithms used in supervised learning (SGD) yield networks for which such upper bound decreases as the number of hidden units increases. This might explain why in some cases overparametrized models have better generalization properties.",
"This paper tackles the important que... | [
[
12
],
[
15
],
[
20
],
[
21
],
[
22
],
[
23
],
[
24
],
[
28
],
[
29
],
[
30
],
[
31
],
[
33
],
[
38
],
[
2
],
[
9
],
[
18
],
[
39
],
[
5
],
[
6
],
[
14
],
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2,
3,
4,
5
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
6,
... | benchmark/PDF/ICLR2019_BygfghAcYX.pdf | openreview | benchmark/MD/ICLR2019_BygfghAcYX.md | ICLR 2019 |
HyGIdiRqtm | {
"title": "Evaluating Robustness of Neural Networks with Mixed Integer Programming",
"abstract": "Neural networks trained only to optimize for training accuracy can often be fooled by adversarial examples --- slightly perturbed inputs misclassified with high confidence. Verification of networks enables us to gauge... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0
]
},
"scores": {
"Solid": null,
"Presentation": null,
"Novelty": null,
"Overall": 7,
"Confidence"... | [
[
{
"role": "AC",
"data": {
"metareview": "\nThe paper investigates mixed-integer linear programming methods for neural net robustness verification in presence of adversarial attckas. The paper addresses and important problem, is well-written, presents a novel approach and demonstrates empir... | [
"This paper presents a mixed integer programming technique for verification of piecewise linear neural networks. This work uses progressive bounds tightening approach to determine bounds for inputs to units. The authors also show that this technique speeds up the bound determination by orders of magnitude as compar... | [
[
2
],
[
4,
10
],
[
11
],
[
14
],
[
12,
15
],
[
16
],
[
18
],
[
8
],
[
9
],
[
17
],
[
0
],
[
5
],
[
1
],
[
6
],
[
13
],
[
3
],
[
7
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect"
] | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
}
],
"category": [
"QUAL-MET",
"QUAL-EXP",
"CLAR-WRT",
"SIGN-SOT"
]
},
{
"sentences": [
{
"role": "Reviewer 2",
"data": [
2
]
... | benchmark/PDF/ICLR2019_HyGIdiRqtm.pdf | openreview | benchmark/MD/ICLR2019_HyGIdiRqtm.md | ICLR 2019 |
HyNbtiR9YX | {
"title": "Unsupervised Document Representation using Partition Word-Vectors Averaging",
"abstract": "Learning effective document-level representation is essential in many important NLP tasks such as document classification, summarization, etc. Recent research has shown that simple weighted averaging of word vecto... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8
]
},
"scores": {
... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper proposes a document classification algorithm based on partitioned word vector averaging.\nI agree with even the most positive reviewer. More experiments would be good. This is a very developed old area."
}
},
{
"role":... | [
"Paper overview: The paper extends the method proposed by Arora 2017 for sentence embeddings to longer document embeddings. The main idea is that, averaging word embedding vectors mixes all the different topics on the document, and therefore is not expressive enough.",
"Instead they propose to estimate the topic ... | [
[
23
],
[
24
],
[
34
],
[
40
],
[
4,
41,
42
],
[
6
],
[
18
],
[
19
],
[
20
],
[
21
],
[
22
],
[
12
],
[
30
],
[
33
],
[
8
],
[
5,
7
],
[
10
],
[
9,
35
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
3
]
},
{
"role": "Au... | benchmark/PDF/ICLR2019_HyNbtiR9YX.pdf | openreview | benchmark/MD/ICLR2019_HyNbtiR9YX.md | ICLR 2019 |
rkxusjRctQ | {
"title": "Learning models for visual 3D localization with implicit mapping",
"abstract": "We consider learning based methods for visual localization that do not require the construction of explicit maps in the form of point clouds or voxels. The goal is to learn an implicit representation of the environment at a ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9
]
},
"scores... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper proposes a method that learns mapping implicitly, by using a generative query network of Eslami et al. with an attention mechanism to learn to predict egomotion. The empirical findings is that training for egomotion estimation alongside th... | [
"Summary:\nEslami et al. (2018) proposed a deep neuronal framework for a scene representation and renderer (the Generative Query Networks: GQN), which generate an image from a scene representation and a query camera pose. In this work, the authors use the GQN to estimate the camera pose from a target image. Existin... | [
[
1
],
[
16
],
[
41
],
[
12
],
[
13
],
[
40
],
[
11
],
[
5,
21
],
[
24
],
[
28
],
[
25
],
[
33
],
[
6
],
[
15
],
[
22
],
[
29
],
[
35
],
[
32
],
[
3
],
[
7
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"CLAR-FIG"
]
},
... | benchmark/PDF/ICLR2019_rkxusjRctQ.pdf | openreview | benchmark/MD/ICLR2019_rkxusjRctQ.md | ICLR 2019 |
SygLehCqtm | {
"title": "Learning protein sequence embeddings using information from structure",
"abstract": "Inferring the structural properties of a protein from its amino acid sequence is a challenging yet important problem in biology. Structures are not known for the vast majority of protein sequences, but structure is crit... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The reviewers and authors had a productive conversation, leading to an improvement in the paper quality. The strengths of the paper highlighted by reviewers are a novel learning set-up and new loss functions that seem to help in the task of protein ... | [
"General comment\nThe authors describe two loss functions for learning embeddings of protein amino acids based on i) predicting the global structural similarity of two proteins, and ii) predicting amino acid contacts within proteins. As far as I know, these loss functions are novel and the authors show clear improv... | [
[
2
],
[
13
],
[
19
],
[
9
],
[
24
],
[
25
],
[
26
],
[
28
],
[
31
],
[
3
],
[
4,
17
],
[
29
],
[
34
],
[
37
],
[
8
],
[
14
],
[
15
],
[
21
],
[
27
],
[
30
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
3
]
},
{
"role"... | benchmark/PDF/ICLR2019_SygLehCqtm.pdf | openreview | benchmark/MD/ICLR2019_SygLehCqtm.md | ICLR 2019 |
ByxZX20qFQ | {
"title": "Adaptive Input Representations for Neural Language Modeling",
"abstract": "We introduce adaptive input representations for neural language modeling which extend the adaptive softmax of Grave et al. (2017) to input representations of variable capacity. There are several choices on how to factorize the in... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3
]
},
"scores": {
"Solid": null,
"Presentation": null,
"Novelty": null,
... | [
[
{
"role": "AC",
"data": {
"metareview": "There is a clear consensus among the reviews to accept this submission thus I am recommending acceptance. The paper makes a clear, if modest, contribution to language modeling that is likely to be valuable to many other researchers."
}
}
... | [
"This article presents experiments on medium- and large-scale language modeling when the ideas of adaptive softmax (Grave et al., 2017) are extended to input representations.",
"The article is well written and I find the contribution simple, but interesting. It is a reasonable and well supported increment from ad... | [
[
13
],
[
1
],
[
4
],
[
12
],
[
10
],
[
11
],
[
5
],
[
6
],
[
7
],
[
8
],
[
9
],
[
14
],
[
2
],
[
0
],
[
3
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"CLAR-WRT",
"QUAL... | benchmark/PDF/ICLR2019_ByxZX20qFQ.pdf | openreview | benchmark/MD/ICLR2019_ByxZX20qFQ.md | ICLR 2019 |
HyxPx3R9tm | {
"title": "Variational Discriminator Bottleneck: Improving Imitation Learning, Inverse RL, and GANs by Constraining Information Flow",
"abstract": "Adversarial learning methods have been proposed for a wide range of applications, but the training of adversarial models can be notoriously unstable. Effectively balan... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper proposes a simple and general technique based on the information bottleneck to constrain the information flow in the discriminator of adversarial models. It helps to train by maintaining informative gradients. While the information bottlen... | [
"Summary:\nThe authors propose to apply the Deep Variational Information Bottleneck (VIB) method of [1] on discriminator networks in various adversarial-learning-based scenarios. They propose a way to adaptively update the value for the bêta hyper-parameter to respect the constraint on I(X,Z). Their technique is sh... | [
[
8
],
[
12
],
[
4
],
[
9
],
[
5,
10
],
[
7
],
[
18
],
[
1
],
[
14
],
[
16
],
[
17
],
[
2
],
[
3
],
[
6
],
[
15
],
[
0
],
[
11
],
[
13
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"QUAL-EXP"
]
},
... | benchmark/PDF/ICLR2019_HyxPx3R9tm.pdf | openreview | benchmark/MD/ICLR2019_HyxPx3R9tm.md | ICLR 2019 |
r1lYRjC9F7 | {
"title": "Enabling Factorized Piano Music Modeling and Generation with the MAESTRO Dataset",
"abstract": "Generating musical audio directly with neural networks is notoriously difficult because it requires coherently modeling structure at many different timescales. Fortunately, most music is also highly structure... | Accept (Oral) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9
]
},
"scores... | [
[
{
"role": "AC",
"data": {
"metareview": "All reviewers agree that the presented audio data augmentation is very interesting, well presented, and clearly advancing the state of the art in the field. The authors’ rebuttal clarified the remaining questions by the reviewers. All reviewers reco... | [
"This paper combines state of the art models for piano transcription, symbolic music synthesis, and waveform generation all using a shared piano-roll representation. It also introduces a new dataset of 172 hours of aligned MIDI and audio from real performances recorded on Yamaha Disklavier pianos in the context of... | [
[
8
],
[
9
],
[
6
],
[
5
],
[
15
],
[
4
],
[
0,
11
],
[
12
],
[
13
],
[
3
],
[
7
],
[
14
],
[
16
],
[
10
],
[
1
],
[
2
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
}
],
"category": [
"ORIG-EXP"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"SIGN-SOT"
]... | benchmark/PDF/ICLR2019_r1lYRjC9F7.pdf | openreview | benchmark/MD/ICLR2019_r1lYRjC9F7.md | ICLR 2019 |
S1x4ghC9tQ | {
"title": "Temporal Difference Variational Auto-Encoder",
"abstract": "To act and plan in complex environments, we posit that agents should have a mental simulator of the world with three characteristics: (a) it should build an abstract state representing the condition of the world; (b) it should form a belief whi... | Accept (Oral) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6
]
},
"scores": {
"Solid": null,
"Pres... | [
[
{
"role": "AC",
"data": {
"metareview": "The reviewers agree that this is a novel paper with a convincing evaluation."
}
}
]
] | [
"There are several ingredients in this paper that I really liked.",
"For example, (1) the notion that an agent should build a deterministic function of the past which implicitly captures the belief (the uncertainty or probability distribution about the state), by opposition for example to sampling trajectories to... | [
[
13
],
[
15
],
[
2
],
[
2
],
[
2
],
[
9
],
[
6
],
[
1
],
[
11
],
[
12
],
[
5
],
[
14
],
[
16
],
[
17
],
[
10
],
[
7
],
[
8
],
[
18
],
[
4
],
[
0
],
[
3... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
}
],
"category": [
"ORIG-MTH"
]
},
... | benchmark/PDF/ICLR2019_S1x4ghC9tQ.pdf | openreview | benchmark/MD/ICLR2019_S1x4ghC9tQ.md | ICLR 2019 |
H1goBoR9F7 | {
"title": "Dynamic Sparse Graph for Efficient Deep Learning",
"abstract": "We propose to execute deep neural networks (DNNs) with dynamic and sparse graph (DSG) structure for compressive memory and accelerative execution during both training and inference. The great success of DNNs motivates the pursuing of lightw... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10
]
},... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper proposes a novel approach for network pruning in both training and inference. This paper received a consensus of acceptance. Compared with previous work that focus and model compression on training, this paper saves memory and accelerates... | [
"This manuscript introduces a computational method to speed up training and inference in deep neural networks: the method is based on dynamic pruning of the compute graph at each iteration of the SGD to approximate computations with a sparse graph. To select which neurons can be zeros and ignored at a given iterati... | [
[
2,
12
],
[
18,
19,
25
],
[
10
],
[
13
],
[
14
],
[
15
],
[
5
],
[
7
],
[
9
],
[
21
],
[
22
],
[
23
],
[
24
],
[
6
],
[
8
],
[
20
],
[
0
],
[
3
],
[
4
],... | [
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0
]
},
{
"role": "Author",
"data": [
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
... | benchmark/PDF/ICLR2019_H1goBoR9F7.pdf | openreview | benchmark/MD/ICLR2019_H1goBoR9F7.md | ICLR 2019 |
rJfW5oA5KQ | {
"title": "Approximability of Discriminators Implies Diversity in GANs",
"abstract": "While Generative Adversarial Networks (GANs) have empirically produced impressive results on learning complex real-world distributions, recent works have shown that they suffer from lack of diversity or mode collapse. The theoret... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6
]
},
"scores": {
"Solid": null,
"Pres... | [
[
{
"role": "AC",
"data": {
"metareview": "The paper presents an interesting theoretical analysis by deriving polynomial sample complexity bounds for the training of GANs that depend on the approximator properties of the discriminator.\nEven if it is not clear if the theory will help to pick... | [
"This paper analyzes that the Integral Probability Metric (IPM) can be a good approximation of Wasserstein distance under some mild assumptions. They first showed two theorems based on simple cases (Gaussian Distribution and Exponential Families).",
"Then, they proved that, for an invertible generator, a special ... | [
[
17
],
[
6,
9
],
[
8
],
[
11
],
[
14
],
[
15
],
[
16
],
[
18
],
[
19
],
[
20
],
[
22
],
[
23
],
[
24
],
[
4,
12
],
[
21
],
[
3
],
[
7
],
[
1
],
[
2
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
3,
4
]
},
{
... | benchmark/PDF/ICLR2019_rJfW5oA5KQ.pdf | openreview | benchmark/MD/ICLR2019_rJfW5oA5KQ.md | ICLR 2019 |
BkMiWhR5K7 | {
"title": "Prior Convictions: Black-box Adversarial Attacks with Bandits and Priors",
"abstract": "We study the problem of generating adversarial examples in a black-box setting in which only loss-oracle access to a model is available. We introduce a framework that conceptually unifies much of the existing work on... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6
]
},
"scores": {
"Solid": null,
"Pres... | [
[
{
"role": "AC",
"data": {
"metareview": "This paper is on the problem of adversarial example generation in the setting where the predictor is only accessible via function evaluations with no gradients available. The associated problem can be cast as a blackbox optimization problem wherein ... | [
"UPDATE:\nI've read the revised version of this paper, I think the concernings have been clarified.",
"This paper proposes to employ the bandit optimization based approach for the generation of adversarial examples under the loss accessible black-box situation. The authors examine the feasibility of using the ste... | [
[
7
],
[
8
],
[
9
],
[
11
],
[
13
],
[
19
],
[
1
],
[
2
],
[
3
],
[
4
],
[
6
],
[
10
],
[
12
],
[
18
],
[
16
],
[
15
],
[
17
],
[
21,
22
],
[
26
],
[
20
]... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1,
2
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
3
]
},
{
"role": "Au... | benchmark/PDF/ICLR2019_BkMiWhR5K7.pdf | openreview | benchmark/MD/ICLR2019_BkMiWhR5K7.md | ICLR 2019 |
rJVorjCcKQ | {
"title": "Slalom: Fast, Verifiable and Private Execution of Neural Networks in Trusted Hardware",
"abstract": "As Machine Learning (ML) gets applied to security-critical or sensitive domains, there is a growing need for integrity and privacy for outsourced ML computations. A pragmatic solution comes from Trusted ... | Accept (Oral) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7
]
},
"scores": {
"Solid": null,
... | [
[
{
"role": "AC",
"data": {
"metareview": "The authors propose a new method of securely evaluating neural networks. \n\nThe reviewers were unanimous in their vote to accept. The paper is very well written, the idea is relatively simple, and so it is likely that this would make a nice present... | [
"Given the growing interest in building trust worthy and privacy protecting AI systems, this paper demonstrates a novel approach to achieve these important goals by allowing a trusted, but slow, computation engine to leverage a fast but untrusted computation engine. For the sake of protecting privacy, this is done ... | [
[
20
],
[
3
],
[
4
],
[
9
],
[
12
],
[
17
],
[
19
],
[
2
],
[
5
],
[
6
],
[
8
],
[
13
],
[
14
],
[
15
],
[
11
],
[
21
],
[
10
],
[
18
],
[
0
],
[
1
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer",
"data": [
2
]
}
],
"category": [
"N/A"
]
... | benchmark/PDF/ICLR2019_rJVorjCcKQ.pdf | openreview | benchmark/MD/ICLR2019_rJVorjCcKQ.md | ICLR 2019 |
H1z-PsR5KX | {
"title": "Identifying and Controlling Important Neurons in Neural Machine Translation",
"abstract": "Neural machine translation (NMT) models learn representations containing substantial linguistic information. However, it is not clear if such information is fully distributed or if some of it can be attributed to ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | [
[
{
"role": "AC",
"data": {
"metareview": "Strong points:\n\n-- Interesting, fairly systematic and novel analyses of recurrent NMT models, revealing individual neurons responsible for specific type of information (e.g., verb tense or gender)\n\n-- Interesting experiments showing how these ne... | [
"The authors propose a number of methods to identify individual important neurons in a machine translation system. The crucial assumption, drawn from the computer vision literature, is that important neurons are going to be correlated across related models (e.g. models that are trained on different subsets of the d... | [
[
22
],
[
2
],
[
10
],
[
11
],
[
13
],
[
14
],
[
15
],
[
16
],
[
23
],
[
24
],
[
26
],
[
9
],
[
17
],
[
12
],
[
0,
5
],
[
1
],
[
4
],
[
6
],
[
7
],
[
18
]... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
1
]
},
{
"role": "Author",
"data"... | benchmark/PDF/ICLR2019_H1z-PsR5KX.pdf | openreview | benchmark/MD/ICLR2019_H1z-PsR5KX.md | ICLR 2019 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.