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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
ByW2Avqgg | {
"title": "Neural Causal Regularization under the Independence of Mechanisms Assumption",
"abstract": "Neural networks provide a powerful framework for learning the association between input and response variables and making accurate predictions. However, in many applications such as healthcare, it is important to... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
4,
5
... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers pointed out several issues with the paper, and all recommended rejection. The revision seems to not have been enough to change their minds."
}
}
]
] | [
"Algorithm 2 has notation that I find somewhat confusing. Step 1 generates data from $p_{X,Y}$, but in my reading of the paper there is no place that actually indicates what those distributions are: I understand that Fig 1 has the graphical models, but what is the actual form of the distribution? Does this choice... | [
[
9
],
[
0
],
[
13
],
[
14
],
[
15
],
[
16
],
[
10
],
[
11
],
[
3,
17,
19
],
[
6
],
[
1
],
[
2
],
[
7
],
[
5
],
[
4
],
[
8
],
[
12
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
4
]
}
],
"category": [
"CLAR-NOT"
]
},
{
"sentences": [
{
"role": "Reviewer 1 ... | benchmark/PDF/ICLR2017_ByW2Avqgg.pdf | openreview | benchmark/MD/ICLR2017_ByW2Avqgg.md | ICLR 2017 |
ByEPMj5el | {
"TL;DR": "",
"title": "Out-of-class novelty generation: an experimental foundation",
"abstract": "Recent advances in machine learning have brought the field closer to computational creativity research. From a creativity research point of view, this offers the potential to study creativity in relationship with k... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "This paper aims to present an experimental framework for selecting machine learning models that can generate novel objects. As the work is devoted to a relatively subjective area of study, it is not surprising that opinions of the work are mixed.\n \n ... | [
"The definition of novelty is relative. Its hard to draw the line between novel and random garbage.",
"Practical definition of novelty should be task dependent. Is there a tractable and useful task where novelty is required to achieve an objective? Have you tested on any such objectives and measure the end perfor... | [
[
10,
15
],
[
11
],
[
12
],
[
14
],
[
9
],
[
0
],
[
1
],
[
2
],
[
3
],
[
5
],
[
6
],
[
7
],
[
18
],
[
13
],
[
4
],
[
8
],
[
16
],
[
17
],
[
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",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
4,
15,
16,
17,
18,
19,
33
]
}
],
"category": [
... | benchmark/PDF/ICLR2017_ByEPMj5el.pdf | openreview | benchmark/MD/ICLR2017_ByEPMj5el.md | ICLR 2017 |
r1w7Jdqxl | {
"TL;DR": "",
"title": "Collaborative Deep Embedding via Dual Networks",
"abstract": "Despite the long history of research on recommender systems, current approaches still face a number of challenges in practice, e.g. the difficulties in handling new items, the high diversity of user interests, and the noisiness... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4
]
}
... | [
[
{
"role": "PC",
"data": {
"comment": "The paper presents a collaborative filtering method, using dual deep nets for users and items. The nets can take advantage of content in addition to ratings. This contribution is just below the bar, in that its novelty relative to existing methods is l... | [
"Since recall@N is used as metric. It makes sense to do a negative sampling based approach baseline, as opposed to rate prediction models.",
"E.g. Bayesian personalized recommendation",
"Thanks for the suggestion. The use of rate prediction is a common practice in existing methods, like WMF and CDL. We did perf... | [
[
8
],
[
12
],
[
0
],
[
1,
2,
3
],
[
4
],
[
6
],
[
7
],
[
9
],
[
10
],
[
5
],
[
11
],
[
13
]
] | [
"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"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
2,
3,
4
]
}
],
"category": [
"QUAL-CMP"
]
},
{
"sentences":... | benchmark/PDF/ICLR2017_r1w7Jdqxl.pdf | openreview | benchmark/MD/ICLR2017_r1w7Jdqxl.md | ICLR 2017 |
rky3QW9le | {
"title": "Transformational Sparse Coding",
"abstract": "\nA fundamental problem faced by object recognition systems is that\nobjects and their features can appear in different locations, scales\nand orientations. Current deep learning methods attempt to achieve\ninvariance to local translations via pooling, disca... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "This paper learns affine transformations from images jointly with object features. The motivation is interesting and sound, but the experiments fail to deliver and demonstrate the validity of the claims advanced -- they are restricted to toy settings. ... | [
"How is sparsity defined in Table 1? I'm unsure how to interpret these numbers.",
"Could you further explain the regularizers in Section 2.3? It was especially unclear to me how to interpret the action of the first regularizer, S_in. (what is the role of the outer square in the definition of S_in?)",
"Also, wha... | [
[
0
],
[
1,
5
],
[
28
],
[
30
],
[
31
],
[
7
],
[
20
],
[
21
],
[
22
],
[
23
],
[
29
],
[
33
],
[
12
],
[
11
],
[
27
],
[
9
],
[
10
],
[
14
],
[
32
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
3,
8
]
}
],
"category": [
"CLAR-FIG"
]
},
{
"sentences": [
{
"role":... | benchmark/PDF/ICLR2017_rky3QW9le.pdf | openreview | benchmark/MD/ICLR2017_rky3QW9le.md | ICLR 2017 |
ryEGFD9gl | {
"title": "Submodular Sum-product Networks for Scene Understanding",
"abstract": "Sum-product networks (SPNs) are an expressive class of deep probabilistic models in which inference takes time linear in their size, enabling them to be learned effectively. However, for certain challenging problems, such as scene un... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5,
6
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"commen... | [
[
{
"role": "PC",
"data": {
"comment": "This paper was reviewed by three experts. While they all find merits in the paper (interesting new model class SSPN, new MAP inference algorithm), they all consistently point to deficiencies in the current manuscript (lack of parameter learning, emphas... | [
"InferSSPN looks to be a good algorithm for solving MAP inference problems in Submodular Sum-Product Networks (SSPN).",
"However, I'm missing the experimental justification for why SSPN is a good model. This leads to some questions:",
"- Why evaluate on training data?",
"- How do the energy values achieved by... | [
[
14
],
[
18
],
[
21
],
[
23
],
[
12,
17
],
[
1
],
[
2
],
[
4
],
[
5
],
[
9
],
[
10,
15,
22
],
[
11
],
[
20
],
[
0
],
[
3
],
[
8
],
[
16
],
[
6
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
}
],
"category": [
"QUAL-MET"
]
},
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
1
]
}
],
"categor... | benchmark/PDF/ICLR2017_ryEGFD9gl.pdf | openreview | benchmark/MD/ICLR2017_ryEGFD9gl.md | ICLR 2017 |
HJ5PIaseg | {
"title": "Towards an automatic Turing test: Learning to evaluate dialogue responses",
"abstract": "Automatically evaluating the quality of dialogue responses for unstructured domains is a challenging problem.\nUnfortunately, existing automatic evaluation metrics are biased and correlate very poorly with human jud... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3
]
}
}
}
],... | [
[
{
"role": "PC",
"data": {
"comment": "Noting the authors' concern about one of the reviewers, I read the paper myself and offer my own brief review.\n \n Evaluation is an extremely important question that does not get enough attention in the machine learning community, so the authors' effo... | [
"Learning a more powerful evaluation model is a nice idea. Could you please comment on the (expected) generalization ability of ADEM beyond the particular dataset that you happen to use in this paper? Ideally you would conduct experiments with other datasets without retraining ADEM on those datasets.",
"Thanks fo... | [
[
11
],
[
6
],
[
7
],
[
0
],
[
1,
5
],
[
8
],
[
9
],
[
10
],
[
12
],
[
13
],
[
15
],
[
4
],
[
2
],
[
3
],
[
14
]
] | [
"correct",
"correct",
"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",
"incorrect"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1,
2,
3
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
... | benchmark/PDF/ICLR2017_HJ5PIaseg.pdf | openreview | benchmark/MD/ICLR2017_HJ5PIaseg.md | ICLR 2017 |
BJuysoFeg | {
"title": "Revisiting Batch Normalization For Practical Domain Adaptation",
"abstract": "Deep neural networks (DNN) have shown unprecedented success in various computer vision applications such as image classification and object detection. However, it is still a common annoyance during the training phase, that one... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The paper performs domain adaptation using a very simple trick inspired by BatchNorm. The paper received below margin scores. The reviewers both, liked the simplicity of the approach, and at the same time felt that the contribution was too thin. Given ... | [
"1. p. 5, 3.3, What is the purpose of the second paragraph? I'm not sure I understand the link between this one and the previous one.",
"2. p. 6, Table 1, What is the evaluation protocol for the Office dataset? DAN results are for the transductive setting; RevGrad results differ from the ones in the ICML paper (a... | [
[
0
],
[
12
],
[
15
],
[
18
],
[
17
],
[
4,
11
],
[
6
],
[
9
],
[
13
],
[
16
],
[
1
],
[
2
],
[
7
],
[
8
],
[
5
],
[
3
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
4,
5,
6
]
}
],
"category": [
"CLAR-WRT"
]
},
{
"sentences": [
{
... | benchmark/PDF/ICLR2017_BJuysoFeg.pdf | openreview | benchmark/MD/ICLR2017_BJuysoFeg.md | ICLR 2017 |
rJbPBt9lg | {
"TL;DR": "",
"title": "Neural Code Completion",
"abstract": "Code completion, an essential part of modern software development, yet can bechallenging for dynamically typed programming languages. In this paper we ex-plore the use of neural network techniques to automatically learn code completionfrom a large ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
... | [
[
{
"role": "PC",
"data": {
"comment": "The paper extends existing code completion methods over discrete symbols with an LSTM-based neural network. This constitutes a novel application of neural networks to this domain, but is rather incremental. Alone, I don't think this would be a bad thin... | [
"A few questions:",
"- It seems a strange to have all the EMPTY terminal tokens in the encoding described in section 4. What is the justification for using the described encoding of trees instead of an encoding like in Grammar as a Foreign Language [1]?",
"- What is the point of predicting non-terminals in a co... | [
[
7
],
[
1
],
[
15
],
[
16
],
[
29
],
[
23
],
[
0
],
[
18
],
[
22
],
[
30
],
[
4
],
[
6
],
[
9
],
[
10,
26
],
[
11
],
[
13
],
[
19
],
[
25
],
[
27
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
6
]
}
],
"category": [
"QUAL-CMP",
"CLAR-WRT"
]
},
{
"sentences": [
... | benchmark/PDF/ICLR2017_rJbPBt9lg.pdf | openreview | benchmark/MD/ICLR2017_rJbPBt9lg.md | ICLR 2017 |
SJk01vogl | {
"TL;DR": "Exploration of ways to attack generative models with adversarial examples and why someone might want to do that.",
"title": "Adversarial examples for generative models",
"abstract": "We explore methods of producing adversarial examples on deep generative models such as the variational autoencoder (VAE... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2
]
}
}
}
],
[
{
... | [
[
{
"role": "PC",
"data": {
"comment": "The main idea in this paper is interesting, of considering various forms of adversarial examples in generative models. The paper has been considerably revised since the original submission. The results showing the susceptibility of generative models to... | [
"Figure 7 suggests that your adversarial approach is systematically distorting the representation such that the reconstructed images belong to a different class. Have you compared the nearest neighbor of your perturbed embeddings to the embeddings in your training data set?",
"Additionally, have you quantified th... | [
[
6
],
[
10,
14,
29
],
[
12,
31
],
[
15
],
[
16
],
[
19
],
[
20
],
[
21
],
[
24
],
[
30
],
[
32
],
[
11
],
[
3
],
[
26
],
[
5
],
[
8,
13,
23
],
[
28
],
[
0
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"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",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
2
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role":... | benchmark/PDF/ICLR2017_SJk01vogl.pdf | openreview | benchmark/MD/ICLR2017_SJk01vogl.md | ICLR 2017 |
H1Go7Koex | {
"title": "Character-aware Attention Residual Network for Sentence Representation",
"abstract": "Text classification in general is a well studied area. However, classifying short and noisy text remains challenging. Feature sparsity is a major issue. The quality of document representation here has a great impact on... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2
]
}
}
}
],
[
{
... | [
[
{
"role": "PC",
"data": {
"comment": "The paper introduces some interesting architectural ideas for character-aware sequence modelling. However, as pointed out by reviewers and from my own reading of the paper, this paper fails badly on the evaluation front. First, some of the evaluation t... | [
"Sentiment analysis is one of most popular tasks on text classification. It would be interesting to see the evaluation of the proposed model on some sentence-level datasets, for example, the Stanford Sentiment Treebank. Is there any reason that kind of experiment was not included?",
"Thanks for the feedback. We h... | [
[
10
],
[
5,
11
],
[
7
],
[
6
],
[
2
],
[
13
],
[
15
],
[
0
],
[
1
],
[
3
],
[
9,
14
],
[
8
],
[
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 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1,
2
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role":... | benchmark/PDF/ICLR2017_H1Go7Koex.pdf | openreview | benchmark/MD/ICLR2017_H1Go7Koex.md | ICLR 2017 |
SywUHFcge | {
"title": " A Theoretical Framework for Robustness of (Deep) Classifiers against Adversarial Samples",
"abstract": "Most machine learning classifiers, including deep neural networks, are vulnerable to adversarial examples. Such inputs are typically generated by adding small but purposeful modifications that lead t... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5,
6,
7,
8
]
}
}
},
{
"role": "Author",
"data": {
... | [
[
{
"role": "PC",
"data": {
"comment": "The authors propose a framework to analyze \"robustness\" to adversarial perturbations using topological concepts. The authors conduct an empirical study using a siamese networks. \n \n The paper generated extensive discussions. The authors implemented... | [
"The theory provides some interesting perspectives for analyzing the adversarial phenomenon. The important thing is to link the theory with novel concrete algorithms.",
"I have following questions for bettering understanding the paper:",
"1.\tIn Corollary 3.4 and Section 3.6, what are features and what are clas... | [
[
29
],
[
7
],
[
8
],
[
2
],
[
3
],
[
10
],
[
16
],
[
21,
23,
25,
26,
27
],
[
24
],
[
0
],
[
14
],
[
18
],
[
4
],
[
6
],
[
19
],
[
9
],
[
12
],
[
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",
"incorrect",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"incorrect",
"incorrec... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
31,
207,
227
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
... | benchmark/PDF/ICLR2017_SywUHFcge.pdf | openreview | benchmark/MD/ICLR2017_SywUHFcge.md | ICLR 2017 |
rJqBEPcxe | {
"title": "Zoneout: Regularizing RNNs by Randomly Preserving Hidden Activations",
"abstract": "We propose zoneout, a novel method for regularizing RNNs.\nAt each timestep, zoneout stochastically forces some hidden units to maintain their previous values.\nLike dropout, zoneout uses random noise to train a pseudo-e... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "Very nice paper, with simple, intuitive idea that works quite well, solving the problem of how to do recurrent dropout.\n \n Pros:\n - Improved results\n - Very simple method\n \n Cons:\n - Almost the best results (aside from Variational Dropout)"
... | [
"I'm glad to see that the setup for the word level PTB task is similar to many other related works, making it easy to compare at a glance. Do you have any intuition as to the different between a zoneout LSTM and the variational LSTM when it comes to perplexity? You mention elsewhere (conclusion?) that you don't per... | [
[
5
],
[
9
],
[
14
],
[
20
],
[
2,
8
],
[
26
],
[
15
],
[
19
],
[
28
],
[
0
],
[
3
],
[
4
],
[
6,
7,
24
],
[
12
],
[
18
],
[
27
],
[
29
],
[
11
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
4
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role": "Reviewer 1 ... | benchmark/PDF/ICLR2017_rJqBEPcxe.pdf | openreview | benchmark/MD/ICLR2017_rJqBEPcxe.md | ICLR 2017 |
HkNKFiGex | {
"title": "Neural Photo Editing with Introspective Adversarial Networks",
"abstract": "The increasingly photorealistic sample quality of generative image models suggests their feasibility in applications beyond image generation. We present the Neural Photo Editor, an interface that leverages the power of generativ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "Here is a summary of strengths and weaknesses as per the reviews:\n \n Strengths\n Work/application is exciting (R3)\n Enough detail for reproducibility (R3)\n May provide a useful analysis tool for generative models (R1)\n \n Weaknesses\n Clarity of t... | [
"1.\nRegarding the masking technique, the paper is vague as to what the gradient step is minimizing. Could you elaborate?",
"2. Does the IAF loss function include a KL-divergence term between the approximate posterior distribution and the prior distribution like is done for VAEs?",
"Hi,",
"Thanks for the ques... | [
[
9
],
[
18
],
[
30
],
[
0
],
[
6
],
[
27
],
[
1
],
[
17,
37,
40
],
[
23
],
[
42
],
[
16
],
[
38
],
[
11
],
[
3,
14,
19
],
[
5
],
[
15
],
[
24
],
[
26
],
[
... | [
"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",
"incorrect",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
4,
6
]
}
],
"category": [
"CLAR-NOT"
]
},
{
"sentences": [
{
"role":... | benchmark/PDF/ICLR2017_HkNKFiGex.pdf | openreview | benchmark/MD/ICLR2017_HkNKFiGex.md | ICLR 2017 |
BJlxmAKlg | {
"title": "ReasoNet: Learning to Stop Reading in Machine Comprehension",
"abstract": "Teaching a computer to read a document and answer general questions pertaining to the document is a challenging yet unsolved problem. In this paper, we describe a novel neural network architecture called Reasoning Network ({Reaso... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "This paper introduces a method for estimating the number of iterations of an attention mechanism in a neural machine reading module using REINFORCE and a custom baseline, which is estimated on the data. Estimating a baseline from data, multi-hop attent... | [
"The graph reachability task is an interesting contribution to highlight the reasoning ability of the model. Is this the first time such a task has been proposed for highlighting the behaviour of memory network-like models? If not, how have previous models done on this task? If so, would you consider testing out pr... | [
[
10
],
[
12
],
[
15
],
[
13
],
[
7
],
[
5
],
[
0,
4,
17
],
[
1,
9
],
[
2
],
[
3
],
[
6
],
[
14
],
[
11
],
[
8
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
1... | benchmark/PDF/ICLR2017_BJlxmAKlg.pdf | openreview | benchmark/MD/ICLR2017_BJlxmAKlg.md | ICLR 2017 |
ryF7rTqgl | {
"title": "Understanding intermediate layers using linear classifier probes",
"abstract": "Neural network models have a reputation for being black boxes. We propose a new method to better understand the roles and dynamics of the intermediate layers. This has direct consequences on the design of such models and it ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers generally agreed that the research direction pursued in the paper is a valuable one, but all reviewers expressed strong reservations about the value of a linear probe on intermediate features. The lack of experiments on more complex state... | [
"The experimental results depicted in Figure 8 are very intriguing. Have you tried using addition instead of concatenation and running the same experiment? It would be very interesting to compare the two scenarios, especially since the concatenation operation seems to have made the first layers of the network seem ... | [
[
5
],
[
16
],
[
3
],
[
11
],
[
14
],
[
2
],
[
13
],
[
19
],
[
10
],
[
0
],
[
1
],
[
4
],
[
17
],
[
6
],
[
7
],
[
8
],
[
9
],
[
18
],
[
12
],
[
15
]
] | [
"correct",
"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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
3
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role": "Reviewer 1 ... | benchmark/PDF/ICLR2017_ryF7rTqgl.pdf | openreview | benchmark/MD/ICLR2017_ryF7rTqgl.md | ICLR 2017 |
r1Bjj8qge | {
"title": "Encoding and Decoding Representations with Sum- and Max-Product Networks",
"abstract": "Sum-Product networks (SPNs) are expressive deep architectures for representing probability distributions, yet allowing exact and efficient inference. SPNs have been successfully applied in several domains, however al... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "AC",
"data": {
"comment": "Dear authors, the reviewers brought up an interesting point in their reviews. They would like to understand the choice of benchmarks and baselines. Specifically, the comparisons in this paper focus on MADE, NADE and MANIAC. MANIAC seems like a curio... | [
"It would be helpful if the authors would explicitly list would the novel and key contributions of this paper are. I appears that previous work from the authors and others is covering a large portion of the approach employed here.",
"What prevents the authors to apply their approach to more interesting datasets?"... | [
[
17
],
[
2,
9
],
[
3
],
[
4
],
[
6
],
[
7
],
[
16
],
[
19
],
[
0
],
[
11
],
[
25
],
[
14
],
[
27
],
[
29
],
[
1
],
[
5
],
[
13
],
[
20
],
[
22
],
[
26
],... | [
"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 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
... | benchmark/PDF/ICLR2017_r1Bjj8qge.pdf | openreview | benchmark/MD/ICLR2017_r1Bjj8qge.md | ICLR 2017 |
HkNEuToge | {
"TL;DR": "",
"title": "Energy-Based Spherical Sparse Coding",
"abstract": "In this paper, we explore an efficient variant of convolutional sparse coding with unit norm code vectors and reconstructions are evaluated using an inner product (cosine distance). To use these codes for discriminative classification, w... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "This paper proposes a variant of convolutional sparse coding with unit norm code vectors using cosine distance to evaluate reconstructions. The performance gains over baseline networks are quite minimal and demonstrated on limited datasets, therefore t... | [
"If I understand correctly, the main point of the paper is that by using this energy based model, with a smart sparse-coding-type formulation, one can involve a feedback strategy into a DNN without the need of a iterative procedure, just running a feed-forward pass. I find the idea interesting.",
"The optimisatio... | [
[
21
],
[
8
],
[
3
],
[
24
],
[
0
],
[
11
],
[
12
],
[
25
],
[
6
],
[
18
],
[
4
],
[
7
],
[
14
],
[
17
],
[
19
],
[
20
],
[
30
],
[
31
],
[
1,
15
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
1,
2
]
},
{
"role... | benchmark/PDF/ICLR2017_HkNEuToge.pdf | openreview | benchmark/MD/ICLR2017_HkNEuToge.md | ICLR 2017 |
HJStZKqel | {
"title": "Lifelong Perceptual Programming By Example",
"abstract": "We introduce and develop solutions for the problem of Lifelong Perceptual Programming By Example (LPPBE). The problem is to induce a series of programs that require understanding perceptual data like images or text. LPPBE systems learn from weak ... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3
]
}
}
},
{
"role": "Unknown",
"data": {
"value": {
"comment": [
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers were quite divided on this submission, which proposes a method for lifelong learning in the context of program generation. While a novel idea, the experiments and baselines are simply not clear enough or convincing enough, and the method ... | [
"This is a wonderful paper on program induction, with a very nice framework for combining code with neural networks.",
"I'd like to ask the authors a simple question, which in no way attempts to diminish the great value of this paper, but rather prompts for clarification.",
"Can you learn more complex programs ... | [
[
2
],
[
7
],
[
10
],
[
16
],
[
1
],
[
4
],
[
5,
15
],
[
8
],
[
12
],
[
17
],
[
11
],
[
13
],
[
14
],
[
0
],
[
3
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
2
]
},
{
"role": "U... | benchmark/PDF/ICLR2017_HJStZKqel.pdf | openreview | benchmark/MD/ICLR2017_HJStZKqel.md | ICLR 2017 |
Sk36NgFeg | {
"title": "Filling in the details: Perceiving from low fidelity visual input",
"abstract": "Humans perceive their surroundings in great detail even though most of our visual field is reduced to low-fidelity color-deprived (e.g., dichromatic) input by the retina. In contrast, most deep learning architectures deploy... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The program committee appreciates the authors' response to concerns raised in the reviews. Unfortunately, reviews are not leaning sufficiently towards acceptance. Reviewers find this direction of exploration to be interesting, but a bit preliminary at ... | [
"Hi,",
"I have some questions regarding your paper:",
"- How does your model relate to state-of-the-art super resolution algorithms (e.g. https://arxiv.org/abs/1609.04802)? In your examples, you compare only to bilinear interpolation.",
"- What does the first row of Figure 3 show (\"original\")?",
"Thanks f... | [
[
2
],
[
21
],
[
22
],
[
23
],
[
28
],
[
1,
14
],
[
6
],
[
17
],
[
5
],
[
7
],
[
8
],
[
9
],
[
10
],
[
15
],
[
16
],
[
19
],
[
20
],
[
25
],
[
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",
"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 Further Reply",
"data": [
2
]
},
{
"role": "A... | benchmark/PDF/ICLR2017_Sk36NgFeg.pdf | openreview | benchmark/MD/ICLR2017_Sk36NgFeg.md | ICLR 2017 |
SygvTcYee | {
"TL;DR": "",
"title": "ParMAC: distributed optimisation of nested functions, with application to binary autoencoders",
"abstract": "Many powerful machine learning models are based on the composition of multiple processing layers, such as deep nets, which gives rise to nonconvex objective functions. A general, r... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The work proposes a parallel/distributed variant of the MAC decomposition method. In presents some theoretical and experimental results supporting the parallelization strategy. The reviews are mixed and indeed a common concern among the reviewers was t... | [
"As described, it seems like this is a parallelized EM-like algorithm that works really well when you have two stages that do not support differentiation.",
"The authors point to prior work that applies to general deep models.",
"However, it's not clear to me what the benefits of parallelization will be there..... | [
[
20
],
[
25
],
[
1
],
[
7
],
[
8
],
[
10,
11,
12
],
[
13
],
[
15
],
[
19
],
[
21
],
[
16
],
[
2
],
[
4
],
[
5
],
[
9
],
[
18
],
[
22,
27,
32,
33
],
[
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",
"incorrect",
"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 Further Reply",
"data": [
2
]
},
{
"role": "A... | benchmark/PDF/ICLR2017_SygvTcYee.pdf | openreview | benchmark/MD/ICLR2017_SygvTcYee.md | ICLR 2017 |
Bkul3t9ee | {
"title": "Unsupervised Perceptual Rewards for Imitation Learning",
"abstract": "Reward function design and exploration time are arguably the biggest obstacles to the deployment of reinforcement learning (RL) agents in the real world. In many real-world tasks, designing a suitable reward function takes considerabl... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "Quality, Clarity:\n \n The work is well motivated and clearly written -- no issues there.\n \n Originality, Significance:\n \n The idea is simple and well motivated, i.e., the learning of reward functions based on feature selection from identified subt... | [
"It would be great if the authors could provide more thorough ablation analysis on the parameter chosen (e.g, number of splits, number of discriminative features selected, mixing coefficients alpha in Eqn. 1). It would be great if the authors could provide with the reasoning process how they are chosen.",
"In Sec... | [
[
1
],
[
15
],
[
3
],
[
13
],
[
14
],
[
16
],
[
30
],
[
34
],
[
35
],
[
36
],
[
39
],
[
0
],
[
2
],
[
6,
38
],
[
8
],
[
9
],
[
10,
20
],
[
11
],
[
12,
22,
... | [
"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 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
4,
5
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role":... | benchmark/PDF/ICLR2017_Bkul3t9ee.pdf | openreview | benchmark/MD/ICLR2017_Bkul3t9ee.md | ICLR 2017 |
ByToKu9ll | {
"title": "Evaluation of Defensive Methods for DNNs against Multiple Adversarial Evasion Models",
"abstract": "Due to deep cascades of nonlinear units, deep neural networks (DNNs) can automatically learn non-local generalization priors from data and have achieved high performance in various applications.\nHowever,... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3
]
}
}
}
],... | [
[
{
"role": "PC",
"data": {
"comment": "The paper investigates several retraining approached based upon adversarial data. While the experimental evaluation looks reasonable, the actual contribution of this paper is quite small. The approaches being evaluated, for the most part, are already p... | [
"Adversarial examples for one network may be adversarial examples for different networks (trained with different initial conditions or even different network architectures). Have you examined how each defense method performs when an adversarial example is calculated on a different model then where it is evaluated o... | [
[
4
],
[
7
],
[
9
],
[
16
],
[
24
],
[
13
],
[
18
],
[
10
],
[
17
],
[
20
],
[
25
],
[
0
],
[
1
],
[
6
],
[
11
],
[
15
],
[
19,
22
],
[
2
],
[
3
],
[
8
],... | [
"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",
"incorrect",
"c... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
2,
3
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role":... | benchmark/PDF/ICLR2017_ByToKu9ll.pdf | openreview | benchmark/MD/ICLR2017_ByToKu9ll.md | ICLR 2017 |
SJ-uGHcee | {
"TL;DR": "",
"title": "Efficient iterative policy optimization",
"abstract": "We tackle the issue of finding a good policy when the number of policy updates is limited. This is done by approximating the expected policy reward as a sequence of concave lower bounds which can be efficiently maximized, drastically ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers generally agreed that exploring policy search methods of this type is interesting, but the results presented in the paper are not at the standard required for publication. There are no comparisons of any sort, and the only task that is te... | [
"It proposes an interesting modification to PoWER. Questions:",
"- Could you elaborate on the statement in second paragraph of section 2, where you state since the “event is highly unpredictable... carefully crafted Q-functions are unlikely to yield improvement”, and why Iterative PoWER has edge over those method... | [
[
1
],
[
6
],
[
7
],
[
12
],
[
15
],
[
0
],
[
14
],
[
3,
5
],
[
8
],
[
2
],
[
11
],
[
4
],
[
10
],
[
9
],
[
13
]
] | [
"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"
] | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
1
]
},
{
"role": "Author",
... | benchmark/PDF/ICLR2017_SJ-uGHcee.pdf | openreview | benchmark/MD/ICLR2017_SJ-uGHcee.md | ICLR 2017 |
HysBZSqlx | {
"title": "Playing SNES in the Retro Learning Environment",
"abstract": "Mastering a video game requires skill, tactics and strategy. While these attributes may be acquired naturally by human players, teaching them to a computer program is a far more challenging task. In recent years, extensive research was carrie... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The authors present a new set of environments, similar to ALE but based on Super Nintendo rather than Atari. This is a great asset and could be important for RL research, but it doesn't merit ICLR publication because of the lack of novel research ideas... | [
"I'd appreciate a bit more context on the motivation for why SNES is an appropriate next benchmark collection, apart from \"more games\". Do you foresee novel AI challenges that could not be studied on any of the existing benchmark domains? In the same vein, maybe cast a somewhat wider net on related work: there ar... | [
[
23
],
[
2
],
[
4
],
[
11
],
[
12
],
[
13,
24
],
[
14
],
[
19
],
[
0
],
[
27
],
[
30
],
[
6,
16,
20
],
[
7
],
[
10,
17
],
[
22
],
[
1
],
[
3
],
[
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 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1,
2,
3,
4,
5,
6,
7,
8,
9
]
}
],
... | benchmark/PDF/ICLR2017_HysBZSqlx.pdf | openreview | benchmark/MD/ICLR2017_HysBZSqlx.md | ICLR 2017 |
S13wCE9xx | {
"title": "Riemannian Optimization for Skip-Gram Negative Sampling",
"abstract": "Skip-Gram Negative Sampling (SGNS) word embedding model, well known by its implementation in \"word2vec\" software, is usually optimized by stochastic gradient descent. It can be shown that optimizing for SGNS objective can be viewed... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5
... | [
[
{
"role": "PC",
"data": {
"comment": "The paper is mostly clearly written. The observation made in the paper that word-embedding models based on optimizing skip-gram negative sampling objective function can be formulated as a low-rank matrix estimation problem, and solved using manifold op... | [
"Does using Riemannian optimization allow the model to converge faster than the alternatives?",
"I'm asking this because the evaluation doesn't show a dramatic advantage to RO-SGNS; the 1% difference on the word similarity benchmarks is within the range of hyperparameter effects (see \"Improving Distributional Si... | [
[
3
],
[
4
],
[
15
],
[
2
],
[
5
],
[
12
],
[
0,
7
],
[
10
],
[
11
],
[
9
],
[
16
],
[
1,
14
],
[
6
],
[
8
],
[
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"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
1
]
},
{
"r... | benchmark/PDF/ICLR2017_S13wCE9xx.pdf | openreview | benchmark/MD/ICLR2017_S13wCE9xx.md | ICLR 2017 |
SkwSJ99ex | {
"TL;DR": "",
"title": "DeepRebirth: A General Approach for Accelerating Deep Neural Network Execution on Mobile Devices",
"abstract": "Deploying deep neural networks on mobile devices is a challenging task due to computation complexity and memory intensity. Existing works solve this problem by reducing model si... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The proposed method doesn't have enough novelty to be accepted to ICLR."
}
}
]
] | [
"It is not always clear in the write-up how the depth (number of feature maps) of the merged layers relate to the depth of the layers before merging. In the case of branch merging, you suggest that the final depth is the sum total of the merged layers' depth. In streamline merging, it's not clear what happens.",
... | [
[
4
],
[
14
],
[
16,
17
],
[
2
],
[
5
],
[
12
],
[
15
],
[
0,
7
],
[
1
],
[
6
],
[
8
],
[
10
],
[
13
],
[
3
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
3,
4
]
}
],
"category": [
"QUAL-REP"
]
},
{
"sentences": [
{
... | benchmark/PDF/ICLR2017_SkwSJ99ex.pdf | openreview | benchmark/MD/ICLR2017_SkwSJ99ex.md | ICLR 2017 |
BJ46w6Ule | {
"title": "Dynamic Partition Models",
"abstract": "We present a new approach for learning compact and intuitive distributed representations with binary encoding. Rather than summing up expert votes as in products of experts, we employ for each variable the opinion of the most reliable expert. Data points are hence... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3
]
}
}
}
],... | [
[
{
"role": "PC",
"data": {
"comment": "This paper is about learning distributed representations. All reviewers agreed that the first draft was not clear enough for acceptance.\n \n Reviewer time is limited and a paper that needed a complete overhaul after the reviews were written is not goi... | [
"Could your approach be extended to have level of expertise be temporally dependent? If not in the most general setting of non-stationarity, then even in a special case in which different experts specialize in different time periods (for different variables), or even in a periodic fashion (e.g. \"sleeping experts\"... | [
[
3
],
[
6
],
[
9
],
[
10
],
[
11
],
[
14
],
[
2
],
[
1
],
[
5
],
[
12
],
[
7
],
[
13
],
[
0
],
[
4
],
[
8
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"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 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1,
2,
3
]
}
],
"category": [
"QUAL-MET"
]
},
{
"sentences": [
{
... | benchmark/PDF/ICLR2017_BJ46w6Ule.pdf | openreview | benchmark/MD/ICLR2017_BJ46w6Ule.md | ICLR 2017 |
BJ0Ee8cxx | {
"title": "Hierarchical Memory Networks",
"abstract": "Memory networks are neural networks with an explicit memory component that can be both read and written to by the network. The memory is often addressed in a soft way using a softmax function, making end-to-end training with backpropagation possible. However, ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "This paper was reviewed by three experts. While they find interesting ideas in the manuscript, all three point to deficiencies (unconvincing results, etc) and unanimously recommend rejection."
}
}
]
] | [
"1. Any reason to not learn the memories, as in the case of standard memory networks? Is the model hard to learn, given that it also has to learn MIPS?",
"2. Why not report the wall clock time in the experiments to get a sense of speedup? Comparing the number of epochs across models is not fair because each epoch... | [
[
17
],
[
5
],
[
7
],
[
15
],
[
1
],
[
3
],
[
8
],
[
14
],
[
16
],
[
18
],
[
0,
2,
9,
11
],
[
4
],
[
12
],
[
13
],
[
6
],
[
10
]
] | [
"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",
"incorrect",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
2,
3,
4,
5
]
}
],
"category": [
"QUAL-MET"
]
},
{
"sentences":... | benchmark/PDF/ICLR2017_BJ0Ee8cxx.pdf | openreview | benchmark/MD/ICLR2017_BJ0Ee8cxx.md | ICLR 2017 |
SJiFvr9el | {
"title": "Linear Time Complexity Deep Fourier Scattering Network and Extension to Nonlinear Invariants",
"abstract": "In this paper we propose a scalable version of a state-of-the-art deterministic time-\ninvariant feature extraction approach based on consecutive changes of basis and\nnonlinearities, namely, the ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
... | [
[
{
"role": "PC",
"data": {
"comment": "This paper proposes a to use squared modulus nonlinearities within convolutional architectures. Because point-wise squaring can be written as a convolution in the Fourier domain, when doing all the operations in the Fourier this architecture becomes 'd... | [
"I have few questions:",
"I do not understand the notion of “higher order non-linearity”. In (Waldspurger 2016), I might be wrong, but I understand “higher order non-linearity” as scattering (or deep features) coefficients of order (e.g. the number of intermediary non-linearity) equal to or more than 3. Could you... | [
[
10
],
[
25
],
[
53
],
[
3
],
[
4
],
[
5
],
[
6
],
[
18
],
[
19
],
[
29
],
[
42
],
[
0
],
[
11
],
[
14
],
[
20
],
[
24
],
[
32
],
[
33
],
[
34
],
[
41
],
[... | [
"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 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
15,
16
]
}
],
"category": [
"CLAR-WRT"
]
},
{
"sentences": [
{
... | benchmark/PDF/ICLR2017_SJiFvr9el.pdf | openreview | benchmark/MD/ICLR2017_SJiFvr9el.md | ICLR 2017 |
Hk4kQHceg | {
"title": "Multiplicative LSTM for sequence modelling",
"abstract": "We introduce multiplicative LSTM (mLSTM), a novel recurrent neural network architecture for sequence modelling that combines the long short-term memory (LSTM) and multiplicative recurrent neural network architectures. mLSTM is characterised by it... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5,
6,
7,
8
]
}
}
},
{
"role": "Author",
"data": {
... | [
[
{
"role": "PC",
"data": {
"comment": "The paper presents a new way of doing multiplicative / tensored recurrent weights in RNNs. The multiplicative weights are input dependent. Results are presented on language modeling (PTB and Hutter). We found the paper to be clearly written, and the id... | [
"What is the reference to tensor RNN in section 1.2?",
"I think you can compare mRNN and mLSTM also to Multi-Function Recurrent Unit (MuFuRU) (https://arxiv.org/abs/1606.03002)?",
"Why do you chose the variant in equation (16) to use tanh after the output gating?",
"How does it differ in performance to the us... | [
[
17,
29
],
[
2
],
[
8
],
[
14
],
[
36
],
[
22
],
[
30
],
[
35
],
[
38
],
[
0
],
[
1
],
[
3
],
[
6
],
[
5,
28
],
[
7
],
[
11,
15,
25
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
9,
10
]
}
],
"category": [
"QUAL-CMP"
]
},
{
"sentences": [
{
"role"... | benchmark/PDF/ICLR2017_Hk4kQHceg.pdf | openreview | benchmark/MD/ICLR2017_Hk4kQHceg.md | ICLR 2017 |
ryUPiRvge | {
"title": "Extrapolation and learning equations",
"abstract": "In classical machine learning, regression is treated as a black box process of identifying a\nsuitable function from a hypothesis set without attempting to gain insight into the mechanism connecting inputs and outputs.\nIn the natural sciences, howeve... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "This paper proposes using functions such as sin and cos as basis functions, then training a neural network with L1 regularization to obtain a simple estimate of functions that can extrapolate under some circumstances.\n \n Pros:\n - the paper has a wid... | [
"Hi,",
"I have a few questions:",
"1. Can you comment/speculate about the space of functions in which the EQL would work well? The sin, cos and other mappings seem to fit perfectly well for the functions considered in the paper, but I wonder how these will generalise to general equations, which could make the c... | [
[
26
],
[
20
],
[
16
],
[
17
],
[
10
],
[
3
],
[
12
],
[
14
],
[
18
],
[
6
],
[
23
],
[
1
],
[
8,
21
],
[
11
],
[
13
],
[
22
],
[
24
],
[
25
],
[
2
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0,
1
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
2
]
},
{
"role": "A... | benchmark/PDF/ICLR2017_ryUPiRvge.pdf | openreview | benchmark/MD/ICLR2017_ryUPiRvge.md | ICLR 2017 |
BJ8fyHceg | {
"title": "Tuning Recurrent Neural Networks with Reinforcement Learning",
"abstract": "The approach of training sequence models using supervised learning and next-step prediction suffers from known failure modes. For example, it is notoriously difficult to ensure multi-step generated sequences have coherent globa... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1
]
}
}
}
],
[
{
"role": "Revi... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers generally liked the application; there were a number of technical points raised that leave doubt about the novelty of the approach. However, this may be an interesting avenue in the future, thus the PCs are accepting it to the workshop tr... | [
"It would be nice to add a link to some of the results, is that possible?",
"Thanks for your interest! Actually, on page 8 we provide a link to the samples generated for the user study: goo.gl/XIYt9m.",
"The paper addresses music generation from MIDI tracks. The authors suggest some technical ideas for improvin... | [
[
8
],
[
17,
25
],
[
26
],
[
22
],
[
14,
20
],
[
5
],
[
6
],
[
7,
10
],
[
24
],
[
2
],
[
12
],
[
15
],
[
27
],
[
9
],
[
11
],
[
18
],
[
19
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1
]
}
],
"category": [
"QUAL-REP"
]
},
{
"sentences": [
{
"role": "Reviewer",
... | benchmark/PDF/ICLR2017_BJ8fyHceg.pdf | openreview | benchmark/MD/ICLR2017_BJ8fyHceg.md | ICLR 2017 |
B16dGcqlx | {
"TL;DR": "Agent watches another agent at a different camera angle completing the task and learns via raw pixels how to imitate. ",
"title": "Third Person Imitation Learning",
"abstract": "Reinforcement learning (RL) makes it possible to train agents capable of achieving\nsophisticated goals in complex and uncer... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "pros:\n - new problem\n - huge number of experimental evaluations, based in part on open-review comments\n \n cons:\n - the main critiques related to not enough experiments being run; this has been addressed in the revised version\n \n The current revi... | [
"In this paper, I didn't see how the expert (teacher) data is acquired. Is this demonstrated from human or optimized (given the known physical parameters)?",
"Also, there is no detailed specification of network architecture (e.g., what layers did you use? How complicated is D_F, D_R and D_D) in the paper besides ... | [
[
3
],
[
16,
27
],
[
17
],
[
28
],
[
19
],
[
25
],
[
9
],
[
2
],
[
5,
10
],
[
6,
11
],
[
7,
12
],
[
8,
13
],
[
20,
30,
32
],
[
26,
29
],
[
34
],
[
21
],
[
... | [
"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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct"... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
2,
3,
4
]
}
],
"category": [
"QUAL-REP",
"QUAL-EXP"
]
},
{
"senten... | benchmark/PDF/ICLR2017_B16dGcqlx.pdf | openreview | benchmark/MD/ICLR2017_B16dGcqlx.md | ICLR 2017 |
Hy-lMNqex | {
"title": "Tartan: Accelerating Fully-Connected and Convolutional Layers in Deep Learning Networks by Exploiting Numerical Precision Variability",
"abstract": "Tartan {TRT} a hardware accelerator for inference with Deep Neural Networks (DNNs) is presented and evaluated on Convolutional Neural Networks. TRT exploit... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "AC",
"data": {
"comment": "Hardware is listed on the call-for-papers as relevant topic for ICLR 2017, and so the paper is on-topic.\n\nWe worked hard to improve on the initial reviewer assignment for this paper in ensure that we would get hardware-knowledgeable reviewers on boar... | [
"The paper provides insights into hardware and circuit design. My question is what is the take-away for the machine learning community? How to train such low precision networks that work well on the specialized hardware? Is there any plan to open source the modified caffe framework?",
"On the hardware perspective... | [
[
11
],
[
3,
17
],
[
6
],
[
13
],
[
14,
19,
21
],
[
2
],
[
8
],
[
16
],
[
1,
9,
15
],
[
18
],
[
0
],
[
20
],
[
5,
7,
10,
22
],
[
4
],
[
12
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"incorrect",
"incorrect"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
4,
5,
6,
9,
10,
11,
12,
13,
14,
15,
... | benchmark/PDF/ICLR2017_Hy-lMNqex.pdf | openreview | benchmark/MD/ICLR2017_Hy-lMNqex.md | ICLR 2017 |
BkjLkSqxg | {
"title": "LipNet: End-to-End Sentence-level Lipreading",
"abstract": "Lipreading is the task of decoding text from the movement of a speaker's mouth. Traditional approaches separated the problem into two stages: designing or learning visual features, and prediction. More recent deep lipreading approaches are end-... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "Let me start by saying that your area chair does not read Twitter, Reddit/ML, etc. The metareview below is, therefore, based purely on the manuscript and the reviews and rebuttal on OpenReview.\n \n The goal of the ICLR review process is to establish a... | [
"I have not yet read the paper completely, but just a quick note about references to prior work on lipreading. The paper claims that \"All existing works, however, perform only word classification, not sentence-level sequence",
"prediction.\" This is not true -- prior work on audio-visual speech recognition oft... | [
[
21
],
[
2
],
[
8
],
[
16
],
[
17
],
[
18
],
[
19
],
[
20
],
[
22
],
[
23
],
[
30
],
[
7,
10
],
[
13
],
[
0,
11,
28
],
[
1
],
[
27
],
[
3
],
[
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",
"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 Further Reply",
"data": [
0,
1,
2
]
},
{
"role": "Author",
"data": [
4,
5,
6,
7
]
}
],
"category": [
"QUAL-CMP"
]... | benchmark/PDF/ICLR2017_BkjLkSqxg.pdf | openreview | benchmark/MD/ICLR2017_BkjLkSqxg.md | ICLR 2017 |
rJXTf9Bxg | {
"title": "Conditional Image Synthesis With Auxiliary Classifier GANs",
"abstract": "Synthesizing high resolution photorealistic images has been a long-standing challenge in machine learning. In this paper we introduce new methods for the improved training of generative adversarial networks (GANs) for image synthe... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "Ratings summary:\n 3: Clear rejection\n 6: Marginally above acceptance threshold\n 6: Marginally above acceptance threshold\n \n Clear easy to read paper focusing on generating higher quality higher resolution (128x128) pixel imagery with GANs. There w... | [
"1. Did you try a comparison of AC-GAN to other class-conditional models?",
"Specifically, plotting Inception accuracy vs. MS-SSIM for AC-GAN compared to e.g. conditional GANs would be insightful. A qualitative comparison of AC-GAN samples vs. other models would be helpful as well.",
"2. Do you think that the M... | [
[
31
],
[
4
],
[
22
],
[
5
],
[
15
],
[
7,
21
],
[
17
],
[
26
],
[
0,
9,
10
],
[
1,
8,
19,
20,
33,
36,
37
],
[
2
],
[
13
],
[
14
],
[
23,
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",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
6,
7,
8,
9,
10,
11,
12,
13,
14
... | benchmark/PDF/ICLR2017_rJXTf9Bxg.pdf | openreview | benchmark/MD/ICLR2017_rJXTf9Bxg.md | ICLR 2017 |
HJTXaw9gx | {
"title": "Recursive Regression with Neural Networks: Approximating the HJI PDE Solution",
"abstract": "Most machine learning applications using neural networks seek to approximate some function g(x) by minimizing some cost criterion. In the simplest case, if one has access to pairs of the form (x, y) where y = g(... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"summary_of_the_paper": null,
"value": {
"review": [
0,
1,
2,
3,
4,
5
]
},
"scores": {
"Solid": null,
"Presentation": null... | [
[
{
"role": "PC",
"data": {
"comment": "The basic approach of this paper is to use a neural net to sequentially generate points that can be used as the basis points in a PDE solver. The idea is definitely an interesting one, and all three reviewers are in agreement that the approach does see... | [
"I have no familiarity with the HJI PDE (I've only dealt with parabolic PDE's such as diffusion in the past).",
"So the details of transforming this problem into a supervised loss escape me.",
"Therefore, as indicated below, my review should be taken as an \"educated guess\". I imagine that many readers of ICLR... | [
[
0
],
[
6
],
[
7
],
[
4
],
[
14
],
[
2
],
[
5,
15
],
[
8
],
[
1
],
[
3
],
[
13
],
[
12
],
[
9
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
0,
1,
2
]
}
],
"category": [
"CLAR-NOT"
]
},
{
"sentences": [
{
"role": "Reviewer 1",
"data": [
3
]
},
{
"rol... | benchmark/PDF/ICLR2017_HJTXaw9gx.pdf | openreview | benchmark/MD/ICLR2017_HJTXaw9gx.md | ICLR 2017 |
H1hoFU9xe | {
"title": "Generative Adversarial Networks for Image Steganography",
"abstract": "Steganography is collection of methods to hide secret information (\"payload\") within non-secret information (\"container\"). Its counterpart, Steganalysis, is the practice of determining if a message contains a hidden payload, and ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1
]
}
}
}
],
[
{
"role": "Revi... | [
[
{
"role": "PC",
"data": {
"comment": "This paper examines an application of that deviates from the usual applications presented at ICLR. The idea seems very interesting to the reviewers, but a number of reviewers had trouble really understanding why the proposed SGAN would be attractive fo... | [
"Have you explored how many bits may be stored in an adversarial perturbation that might not be detected? I would imagine that if the perturbation were larger, more bits might be communicated but at a risk of being detected easier.",
"In steganographic articles the payload bit rate is usually set to 0.4-0.6 bits ... | [
[
4,
13,
15
],
[
6
],
[
18
],
[
8
],
[
3
],
[
0
],
[
1
],
[
5,
7,
10,
17
],
[
11
],
[
12
],
[
9
],
[
16
],
[
2
],
[
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",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role": "Reviewer 2 ... | benchmark/PDF/ICLR2017_H1hoFU9xe.pdf | openreview | benchmark/MD/ICLR2017_H1hoFU9xe.md | ICLR 2017 |
r1br_2Kge | {
"title": "Short and Deep: Sketching and Neural Networks",
"abstract": "Data-independent methods for dimensionality reduction such as random projections, sketches, and feature hashing have become increasingly popular in recent years. These methods often seek to reduce dimensionality while preserving the hypothesi... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10
]
}
}
},
{
"role": "... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers present a detailed set of concerns regarding the paper. In particular, the paper lacks comparison to other sketching works. The sketches used in the paper are rudimentary and in practice, there are more sophisticated sketches employed.\n ... | [
"Overall, I like the paper and the ideas.",
"The paper is about a better dimensionality reduction techniques for sparse binary data using count sketches. The nice thing about sketching is the possibility of reconstruction, and hence any function approximation can be build using NN (as NN are universal approximato... | [
[
11
],
[
1
],
[
8
],
[
12
],
[
17
],
[
27,
36
],
[
16
],
[
0
],
[
37
],
[
38
],
[
2,
3,
4,
5,
6,
7,
9
],
[
13,
14
],
[
19
],
[
23
],
[
30
],
[
42
],
[
... | [
"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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
1
]
}
],
"categor... | benchmark/PDF/ICLR2017_r1br_2Kge.pdf | openreview | benchmark/MD/ICLR2017_r1br_2Kge.md | ICLR 2017 |
Sy1rwtKxg | {
"title": "Parallel Stochastic Gradient Descent with Sound Combiners",
"abstract": "Stochastic gradient descent (SGD) is a well-known method for regression and classification tasks. However, it is an inherently sequential algorithm — at each step, the processing of the current example depends on the parameters lea... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers largely agree that this paper is well written and presents an interesting, novel approach to parallelizing Stochastic Gradient Descent. However, the current formulation is restricted to linear regression models and requires sketching tech... | [
"Specifically, it is unclear from the paper whether does the gradient calculation utilizes the sparsity of the dataset.",
"It would be helpful to give a detailed description on how gradient is calculated, and time complexity analysis, given N X M a sparse input data, with only Z entries",
"We do utilize the sp... | [
[
10
],
[
19
],
[
5
],
[
13
],
[
7
],
[
15
],
[
3
],
[
16
],
[
18
],
[
21
],
[
22
],
[
0
],
[
1
],
[
2
],
[
6,
9
],
[
8
],
[
14
],
[
12
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
2,
3,
4
]
}
],
"category": [
"QUAL-REP"
]
},
{
"sentences": [
{
... | benchmark/PDF/ICLR2017_Sy1rwtKxg.pdf | openreview | benchmark/MD/ICLR2017_Sy1rwtKxg.md | ICLR 2017 |
rkaRFYcgl | {
"title": "Low-rank passthrough neural networks",
"abstract": "Deep learning consists in training neural networks to perform computations that sequentially unfold in many steps over a time dimension or an intrinsic depth dimension. For large depths, this is usually accomplished by specialized network architectures... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5
... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers seem to agree that the framework presented is not very novel, something I agree with.\n The experiments show that the low rank + diagonal parameterization can be useful, however. The paper could be improved by making a more tightened mess... | [
"- Two proposals are made (passthrough and lowrank) and experiments are shown with lowrank only, lowrank and passthrough, but not passthrough only. Why?",
"- Figure 2 only shows curves from the proposed models, but not for baselines; why?",
"- In Table 1 and Table 2, what happens when the state size is higher f... | [
[
9,
15
],
[
4
],
[
7
],
[
12,
13
],
[
18
],
[
21
],
[
22
],
[
0
],
[
1,
23
],
[
2
],
[
5
],
[
17
],
[
20
],
[
24
],
[
3
],
[
16
],
[
25
],
[
6
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
3
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role": "Reviewer 1 ... | benchmark/PDF/ICLR2017_rkaRFYcgl.pdf | openreview | benchmark/MD/ICLR2017_rkaRFYcgl.md | ICLR 2017 |
Bk_zTU5eg | {
"title": "Inefficiency of stochastic gradient descent with larger mini-batches (and more learners)",
"abstract": "Stochastic Gradient Descent (SGD) and its variants are the most important optimization algorithms used in large scale machine learning. Mini-batch version of stochastic gradient is often used in pract... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2
]
}
}
},
{
"role"... | [
[
{
"role": "PC",
"data": {
"comment": "The work addresses the question of whether mini-batching improves the convergence of stochastic gradient methods, in terms of the number of examples, in the general non-asymptotic/non-convex setting of Ghadimi and Lan. Similar results are already known... | [
"Your theoretical results suggest that for SGD, not using mini-batch (i.e., mini-batch size = 1) is the best. On asynchronous SGD, your results suggest that using only 1 learner is best. These seem contradictory to the common practice. Can you please explain?",
"What our analysis tries to show is that in terms of... | [
[
1
],
[
7
],
[
0
],
[
2
],
[
5
],
[
3
],
[
4
],
[
8
],
[
10
],
[
6
],
[
9
]
] | [
"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 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
1... | benchmark/PDF/ICLR2017_Bk_zTU5eg.pdf | openreview | benchmark/MD/ICLR2017_Bk_zTU5eg.md | ICLR 2017 |
r1osyr_xg | {
"title": "Fuzzy paraphrases in learning word representations with a lexicon",
"abstract": "A synonym of a polysemous word is usually only the paraphrase of one sense among many. When lexicons are used to improve vector-space word representations, such paraphrases are unreliable and bring noise to the vector-space... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers agree that the paper's clarity and experimental evaluation can be improved."
}
}
]
] | [
"Compared to JointREPs, your method achieves slightly better overall accuracy on the analogy task, but it performs much better on the semantic category than it does on the syntactic one. Do you have an explanation for this imbalance?",
"Dear AnonReviewer1:\nHello. Thanks for your comment.",
"Each question in th... | [
[
14
],
[
7,
13
],
[
1
],
[
4
],
[
16
],
[
17
],
[
0
],
[
3,
6
],
[
5
],
[
8
],
[
9
],
[
10
],
[
11
],
[
15
],
[
2
],
[
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"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
2,
3,
4,
5,
6,
7,
8,
9
]
}
],
"category"... | benchmark/PDF/ICLR2017_r1osyr_xg.pdf | openreview | benchmark/MD/ICLR2017_r1osyr_xg.md | ICLR 2017 |
ryh_8f9lg | {
"title": "Classless Association using Neural Networks",
"abstract": "The goal of this paper is to train a model based on the relation between two instances that represent the same unknown class. This scenario is inspired by the Symbol Grounding Problem and the association learning in infants. We propose a novel... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The paper explores neural-network learning on pairs of samples that are labeled as either similar or dissimilar. The proposed model appears to be different from standard siamese architectures, but it is poorly motivated. The experimental evaluation of ... | [
"Interesting paper, but still i do not understand how to define the \"target distribution (E[z1 , . . . , zm ] ∼ φ ∈ Rc )\" why it should be \"uniform\", should we know the prior of the different classes?",
"Thank you very much for your question.",
"Our model relies on the distribution of the classes i.e., fo... | [
[
4
],
[
6
],
[
0,
3,
8
],
[
5
],
[
9
],
[
11
],
[
12
],
[
15
],
[
1
],
[
14
],
[
7
],
[
13
],
[
2
],
[
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"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
2,
3,
4,
5,
6
]
}
],
"category": [
"CLAR-WRT"
]
},
{
... | benchmark/PDF/ICLR2017_ryh_8f9lg.pdf | openreview | benchmark/MD/ICLR2017_ryh_8f9lg.md | ICLR 2017 |
ryCcJaqgl | {
"TL;DR": "",
"title": "TreNet: Hybrid Neural Networks for Learning the Local Trend in Time Series",
"abstract": "Local trends of time series characterize the intermediate upward and downward patterns of time series. Learning and forecasting the local trend in time series data play an important role in many real... | Reject | [
[
{
"role": "Author",
"data": {
"value": {
"comment": [
0,
1,
2,
3
]
}
}
}
],
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
4,
5
... | [
[
{
"role": "PC",
"data": {
"comment": "I appreciate the authors putting a lot of effort into the rebuttal. But it seems that all the reviewers agree that the local trend features segmentation and computation is adhoc, and the support for accepting the paper is lukewarm.\n \n As an additiona... | [
"Dear Reviewers,\nWe upload a new version of the paper and list the content updated as follows:",
"1. Refine Figure 1 and Figure 2.",
"2. In the experiment section, update some experiment results and the result discussion.",
"3. In the appendix section, add data pre-processing subsection to explain the local ... | [
[
4,
5,
25
],
[
14
],
[
15
],
[
16,
27
],
[
18
],
[
26
],
[
20
],
[
32
],
[
1,
7
],
[
12
],
[
13
],
[
21
],
[
23,
28
],
[
34
],
[
37
],
[
39
],
[
42
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer 2 Further Reply",
"data": [
4
]
},
{
"role": "Author",
"data": [
7,
8,
9,
10,
11,
12,
13
]
}
],
"category": [
... | benchmark/PDF/ICLR2017_ryCcJaqgl.pdf | openreview | benchmark/MD/ICLR2017_ryCcJaqgl.md | ICLR 2017 |
BkIqod5ll | {
"title": "Convolutional Neural Networks Generalization Utilizing the Data Graph Structure",
"abstract": "Convolutional Neural Networks have proved to be very efficient in image and audio processing. Their success is mostly attributed to the convolutions which utilize the geometric properties of a low - dimensiona... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "This work studies the problem of generalizing a convolutional neural network to data lacking grid-structure. \n \n The authors consider the Random Walk Normalized Laplacian and its finite powers to define a convolutional layer in a general graph. Exper... | [
"1. p. 8, 4.2, How does the proposed method compare to regular CNNs with the same number of parameters?",
"2. p. 8, 4.2, How one would ensure consistent ordering of pixels for all the spatial locations? For some pixels, the expected number of visits is the same and it's essential that the ties are broken consiste... | [
[
9
],
[
5
],
[
7
],
[
8
],
[
10
],
[
13
],
[
0
],
[
12
],
[
14
],
[
16
],
[
1
],
[
2
],
[
4
],
[
3
],
[
6
],
[
11
],
[
15
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"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 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
3,
4
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role":... | benchmark/PDF/ICLR2017_BkIqod5ll.pdf | openreview | benchmark/MD/ICLR2017_BkIqod5ll.md | ICLR 2017 |
By1snw5gl | {
"title": "L-SR1: A Second Order Optimization Method for Deep Learning",
"abstract": "We describe L-SR1, a new second order method to train deep neural networks. Second order methods hold great promise for distributed training of deep networks. Unfortunately, they have not proven practical. Two significant barrier... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The paper proposes an interesting approach, in that (unlike many second-order methods) SR1 updates can potentially take advantage of negative curvature in the Hessian. However, all reviewers had some significant concerns about the utility of the method... | [
"Dear Authors,",
"1) How does the experimental results look in terms of the wallclock time?",
"2) The generalized trust region method that overcomes the saddle point problem is presented in the work \"Identifying and attacking the saddle point problem in",
"high-dimensional non-convex optimization\", where th... | [
[
3
],
[
9
],
[
1
],
[
4
],
[
10
],
[
0
],
[
6
],
[
7
],
[
8
],
[
2
],
[
5
]
] | [
"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 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
4,
5,
6
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences":... | benchmark/PDF/ICLR2017_By1snw5gl.pdf | openreview | benchmark/MD/ICLR2017_By1snw5gl.md | ICLR 2017 |
Sywh5KYex | {
"title": "Learning Identity Mappings with Residual Gates",
"abstract": "We propose a layer augmentation technique that adds shortcut connections with a linear gating mechanism, and can be applied to almost any network model. By using a scalar parameter to control each gate, we provide a way to learn identity mapp... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "Although this was a borderline paper, the reviewers ultimately concluded that, given how easy it would be for a practitioner to independently devise the methodological trick of the paper, the paper did not demonstrate that the idea was sufficiently use... | [
"Can you please elaborate on how SDI is assumed to be a credible measure?",
"Assuming that random pruning in case of Augmentation drops layers irrespective of corresponding k value. It is surprising to see that performance does not degrade even when layers with high k might have been dropped. Can you please expla... | [
[
7
],
[
4
],
[
17
],
[
16
],
[
26
],
[
3
],
[
9
],
[
10
],
[
12,
25
],
[
18
],
[
19
],
[
1
],
[
2
],
[
5
],
[
8
],
[
11
],
[
14
],
[
15
],
[
20
],
[
21
]... | [
"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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
3
]
}
],
"category": [
"QUAL-MET"
]
},
{
"sentences": [
{
"role": "Reviewer 1 ... | benchmark/PDF/ICLR2017_Sywh5KYex.pdf | openreview | benchmark/MD/ICLR2017_Sywh5KYex.md | ICLR 2017 |
SJx7Jrtgl | {
"title": "Deep Unsupervised Clustering with Gaussian Mixture Variational Autoencoders",
"abstract": "We study a variant of the variational autoencoder model (VAE) with a Gaussian mixture as a prior distribution, with the goal of performing unsupervised clustering through deep generative models. We observe that th... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"comment": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5
]... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers have looked through both the responses, updates, and had much discussion. We agree that the paper is well executed and exposes ideas that are of value and interest. At the same same, the extent to which the methods can be applied in pract... | [
"I fail to understand what is the advantage of using a learned approximate posterior q_{\\phi_z}(z|y) when in equation (6), you derive the *exact* form of p(z|x,w). Why don't you use the approximate posterior q_{\\phi_w}(w|y)q_{\\phi_x}(x|y)p(z|x,w) instead of q_{\\phi_w}(w|y)q_{\\phi_x}(x|y)q_{\\phi_z}(z|y) ? Both... | [
[
22
],
[
12
],
[
26
],
[
3
],
[
10
],
[
4
],
[
5
],
[
6
],
[
11
],
[
19
],
[
7
],
[
8
],
[
15
],
[
18
],
[
24
],
[
0,
16
],
[
1
],
[
13
],
[
14
],
[
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",
"incorrect",
"c... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
2,
3,
4,
5,
71,
76,
77
]
}
],
"catego... | benchmark/PDF/ICLR2017_SJx7Jrtgl.pdf | openreview | benchmark/MD/ICLR2017_SJx7Jrtgl.md | ICLR 2017 |
HyQWFOVge | {
"title": "Significance of Softmax-Based Features over Metric Learning-Based Features",
"abstract": "The extraction of useful deep features is important for many computer vision tasks.\nDeep features extracted from classification networks have proved to perform well in those tasks.\nTo obtain features of greater u... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The paper aims to compare the representations learnt by metric learning and classification objectives. While this is an interesting topic, the presented evaluation is not sufficiently clear for the paper to be accepted."
}
}
]
] | [
"L2 normalization seemed to very important for experiments in S4.4.",
"Yet in table 1 the experiments show little difference w and w/o normalization. Why? (Did you use cosine distance in any of the experiments, or was it all L2?)",
"Could you please report the pool5 / pool5+L2 results in table 2? (Or is there a... | [
[
19
],
[
20
],
[
6
],
[
7
],
[
8,
29
],
[
9
],
[
13
],
[
21
],
[
27
],
[
0
],
[
1
],
[
4,
5,
26,
28,
30
],
[
11
],
[
14
],
[
12
],
[
22
],
[
17
],
[
15
]... | [
"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",
"incorrect",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
6,
7,
8,
9,
10,
11,
12,
13
]
}
... | benchmark/PDF/ICLR2017_HyQWFOVge.pdf | openreview | benchmark/MD/ICLR2017_HyQWFOVge.md | ICLR 2017 |
ry7O1ssex | {
"TL;DR": "",
"title": "Generative Adversarial Networks as Variational Training of Energy Based Models",
"abstract": "In this paper, we study deep generative models for effective unsupervised learning. We propose VGAN, which works by minimizing a variational lower bound of the negative log likelihood (NLL) of an... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2
]
}
}
}
],
[
{
... | [
[
{
"role": "PC",
"data": {
"comment": "This paper is timely since it addresses the connections between energy-based models, GANS, and the general space of generative models. The two principal concerns about the paper are: lack of clarity and coherence in the paper; inability to effectively ... | [
"Do you have learning the curves from multiple runs? That could help comparing the stability of the models.",
"Hi,\nWe did not record multiple runs in our experiments, but we experienced very similar behaviors across different random initializations. You are also welcome to try out our code, which is available th... | [
[
2
],
[
15
],
[
16
],
[
13
],
[
14
],
[
3
],
[
6
],
[
8
],
[
9
],
[
1
],
[
5,
10
],
[
11
],
[
12
],
[
0
],
[
4
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1
]
}
],
"category": [
"QUAL-STA"
]
},
{
"sentences": [
{
"role": "Reviewer 2 ... | benchmark/PDF/ICLR2017_ry7O1ssex.pdf | openreview | benchmark/MD/ICLR2017_ry7O1ssex.md | ICLR 2017 |
S1Bm3T_lg | {
"title": "Compositional Kernel Machines",
"abstract": "Convolutional neural networks (convnets) have achieved impressive results on recent computer vision benchmarks. While they benefit from multiple layers that encode nonlinear decision boundaries and a degree of translation invariance, training convnets is a le... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1
]
}
}
}
],
[
{
"role": "Revi... | [
[
{
"role": "PC",
"data": {
"comment": "There is consensus among the reviewers that the proposed method has potential merit, but that the experimental evaluation is too preliminary to warrant publication of the current manuscript. The paper also appears to make broad claims that are not full... | [
"NORB is a very toy-ish dataset by today's standard. Have you tried the method on more realistic datasets, such as ImageNet?",
"Yes, we are in the process of scaling this method to ImageNet.",
"This paper proposes a new learning framework called \"compositional kernel machines\" (CKM). It combines two ideas: ke... | [
[
6
],
[
7
],
[
8
],
[
25
],
[
27
],
[
31
],
[
35
],
[
2
],
[
15
],
[
23
],
[
24,
26
],
[
28
],
[
33
],
[
37
],
[
0,
16
],
[
4
],
[
5,
18
],
[
19
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
90,
105
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"rol... | benchmark/PDF/ICLR2017_S1Bm3T_lg.pdf | openreview | benchmark/MD/ICLR2017_S1Bm3T_lg.md | ICLR 2017 |
B1TTpYKgx | {
"title": "On the Expressive Power of Deep Neural Networks",
"abstract": "We study the expressive power of deep neural networks before and after\ntraining. Considering neural nets after random initialization, we show that\nthree natural measures of expressivity all display an exponential dependence\non the depth o... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "While the reviewers saw some value in your contribution, there were also serious issues, so the paper does not reach the acceptance threshold."
}
}
]
] | [
"Kindly explain ``circular trajectory is chosen between two random vectors''.",
"Also ``non-zero magnitude perpendicular to x(t)''. Is this saying that the curve is not a ray from the origin?",
"In Theorem 1, random neural network means a fixed network with random weights?",
"The one dimensional trajectory is... | [
[
42
],
[
45
],
[
46
],
[
47
],
[
13
],
[
15
],
[
32
],
[
33
],
[
34
],
[
35
],
[
40
],
[
0
],
[
1
],
[
2
],
[
3
],
[
8
],
[
10
],
[
11
],
[
14
],
[
18,
36
... | [
"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",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
5
]
}
],
"category": [
"CLAR-WRT"
]
},
{
"sentences": [
{
"role": "Reviewer 1 ... | benchmark/PDF/ICLR2017_B1TTpYKgx.pdf | openreview | benchmark/MD/ICLR2017_B1TTpYKgx.md | ICLR 2017 |
BJ_MGwqlg | {
"title": "Rethinking Numerical Representations for Deep Neural Networks",
"abstract": "With ever-increasing computational demand for deep learning, it is critical to investigate the implications of the numeric representation and precision of DNN model weights and activations on computational efficiency. In this w... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1
]
}
}
}
],
[
{
"role": "Revi... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers feel that this is a well written paper on floating and fixed point representations for inference with several state of the art deep learning architectures. At the same time, in order for results to be more convincing, they recommend using... | [
"Did you try training these networks using these customized precisions too ? It would be a bonus if the same hardware could be used for training.",
"We briefly explored training DNNs with customized precision and found that training requires more precision than inference, since training makes minute adjustments t... | [
[
3,
20
],
[
14
],
[
5
],
[
1
],
[
9
],
[
15
],
[
19
],
[
0
],
[
4
],
[
10
],
[
11
],
[
18
],
[
21
],
[
12
],
[
7
],
[
8
],
[
16
],
[
17
],
[
2
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role": "Reviewer 2 ... | benchmark/PDF/ICLR2017_BJ_MGwqlg.pdf | openreview | benchmark/MD/ICLR2017_BJ_MGwqlg.md | ICLR 2017 |
HyCRyS9gx | {
"title": "Fast Adaptation in Generative Models with Generative Matching Networks",
"abstract": "Despite recent advances, the remaining bottlenecks in deep generative models are necessity of extensive training and difficulties with generalization from small number of training examples.\nBoth problems may be addres... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3,
4,
5
... | [
[
{
"role": "PC",
"data": {
"comment": "This work extends variational autoencoders to adapt to a new dataset containing a small number of examples. While this work is promising, two of the reviewers had serious concerns about clarity. A new version of the paper has been submitted, however I ... | [
"Can you expand on the relationship between your approach and the matching networks of Vinyals et al. (2016)? You seem to be building on that work, but it's only discussed briefly in the introduction.",
"Thank you for the question.",
"That is totally correct, as we mention in the paper our approach is inspired... | [
[
14
],
[
8
],
[
11
],
[
13
],
[
0,
21
],
[
1
],
[
6
],
[
9
],
[
10
],
[
12
],
[
19
],
[
22
],
[
3
],
[
4,
23
],
[
7
],
[
15
],
[
16
],
[
17,
18
],
[
2
],... | [
"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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
2,
3,
4,
5
]
}
],
"category": [
"CLAR-WRT",
"QUAL-CMP"
]
},
... | benchmark/PDF/ICLR2017_HyCRyS9gx.pdf | openreview | benchmark/MD/ICLR2017_HyCRyS9gx.md | ICLR 2017 |
HJ1JBJ5gl | {
"title": "Representing inferential uncertainty in deep neural networks through sampling",
"abstract": "As deep neural networks (DNNs) are applied to increasingly challenging problems, they will need to be able to represent their own uncertainty. Modelling uncertainty is one of the key features of Bayesian methods... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers unanimously recommend rejecting this paper."
}
}
]
] | [
"The results of this analysis are based on a very specific scenario of MNIST with dropout 0.5. Do these results hold across other, more complicated datasets?",
"Even though 0.5 is common for dropout, this is not always the value that is used. What do these results look like at different values of the dropout para... | [
[
13
],
[
3,
11,
19
],
[
4
],
[
5
],
[
10
],
[
14
],
[
17
],
[
0,
7,
9
],
[
18
],
[
12
],
[
1
],
[
15
],
[
8
],
[
16
],
[
2
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
3,
4,
5,
6,
7,
8
]
}
],
"category": [
"QU... | benchmark/PDF/ICLR2017_HJ1JBJ5gl.pdf | openreview | benchmark/MD/ICLR2017_HJ1JBJ5gl.md | ICLR 2017 |
Hk85q85ee | {
"title": "Symmetry-Breaking Convergence Analysis of Certain Two-layered Neural Networks with ReLU nonlinearity",
"abstract": "In this paper, we use dynamical system to analyze the nonlinear weight dynamics of two-layered bias-free networks in the form of $g(x; w) = \\sum_{j=1}^K \\sigma(w_j \\cdot x)$, where $\\s... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The paper analyzes the dynamics of learning under Gaussian input using dynamical systems theory. As two of the reviewers have pointed out, the paper is hard to read, and not written in a way which is accessible to the wider ICLR community. Hence, I can... | [
"The assumption of Gaussian input X seems rather strong. Do you have a sense of how the analysis would change if the input features have non-Gaussian or multi-modal distributions (as is most likely the case with real-world input features)? Are there any other implicit or explicit assumptions made in the paper?",
... | [
[
3
],
[
4
],
[
5
],
[
6
],
[
8
],
[
15
],
[
1
],
[
7
],
[
9
],
[
17
],
[
18
],
[
19
],
[
21
],
[
22
],
[
14
],
[
20
],
[
12
],
[
0
],
[
2
],
[
11
],
[
... | [
"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",
"incorrect",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
2,
3,
4,
5,
6,
59
]
}
],
"category": [
"QUAL-MET"
... | benchmark/PDF/ICLR2017_Hk85q85ee.pdf | openreview | benchmark/MD/ICLR2017_Hk85q85ee.md | ICLR 2017 |
HyY4Owjll | {
"TL;DR": "",
"title": "Boosted Generative Models",
"abstract": "We propose a new approach for using boosting to create an ensemble of generative models, where models are trained in sequence to correct earlier mistakes. Our algorithm can leverage many existing base learners, including recent latent variable mod... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The idea of boosting has recently seen a revival, and the ideas presented here are stimulating. After discussion, the reviewers agreed that the latest updates and clarifications have improved the paper, but overall they still felt that the paper is not... | [
"Could the authors provide a more detailed comparison to the work of Tu (2007) ? There seems to be a significant overlap with Sec. 2.2 (including Theorem 2 and its proof), which is not fairly represented in the related works section IMO. I believe more credit and transparency in this regard would benefit the paper,... | [
[
8
],
[
18
],
[
0
],
[
2
],
[
15
],
[
17
],
[
1,
14
],
[
10
],
[
4
],
[
5
],
[
6
],
[
7
],
[
13
],
[
3
],
[
9,
20
],
[
12,
19
],
[
22
],
[
16
],
[
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"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
4,
5,
6,
7,
8,
9
]
}
],
"category": [
"QUAL-CMP"
]... | benchmark/PDF/ICLR2017_HyY4Owjll.pdf | openreview | benchmark/MD/ICLR2017_HyY4Owjll.md | ICLR 2017 |
SJZAb5cel | {
"title": "A Joint Many-Task Model: Growing a Neural Network for Multiple NLP Tasks",
"abstract": "Transfer and multi-task learning have traditionally focused on either a single source-target pair or very few, similar tasks. Ideally, the linguistic levels of morphology, syntax and semantics would benefit each othe... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5,
6
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"commen... | [
[
{
"role": "PC",
"data": {
"comment": "There is a bit of spread in the reviewer scores, but ultimately the paper does not meet the high bar for acceptance to ICLR. The lack of author responses to the reviews does not help either."
}
}
]
] | [
"questions:",
"- without output label embeddings how do you connect two tasks?",
"- depth and shortcut connections effect relatedness and entailment tasks too much. how do you explain this? could this be related to using the same corpus (PTB) for all the previous tasks?",
"- for the first three tasks section ... | [
[
0
],
[
19
],
[
11
],
[
12
],
[
13
],
[
22
],
[
4
],
[
1
],
[
3
],
[
5,
18
],
[
6
],
[
7
],
[
8
],
[
9,
25
],
[
10,
24
],
[
14
],
[
15
],
[
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",
"incorrect",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
9,
10,
11,
12,
13
]
}
],
"category": [
"CLAR-WRT"
... | benchmark/PDF/ICLR2017_SJZAb5cel.pdf | openreview | benchmark/MD/ICLR2017_SJZAb5cel.md | ICLR 2017 |
BkXMikqxx | {
"title": "Cortical-Inspired Open-Bigram Representation for Handwritten Word Recognition",
"abstract": "Recent research in the cognitive process of reading hypothesized that we do\nnot read words by sequentially recognizing letters, but rather by identifing\nopen-bigrams, i.e. couple of letters that are not necess... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11
]
}
}
},
{... | [
[
{
"role": "PC",
"data": {
"comment": "There is consistent agreement towards the originality of this work and that the topic here is \"interesting\". Additionally there is consensus that the work is \"clearly written\", and (excepting questions of the word \"cortical\") all would be primed ... | [
"The performance numbers provided for IAM and RIMES in Table 1 do not seem to be optimal. The best results known to me were published in [Poznansky & Wolf, CVPR 2016], with 3.9% WER for RIMES, and 6.44% WER for IAM.",
"On the other hand, the model complexity chosen here, as described in Appendix A.3, seems to be ... | [
[
9,
18
],
[
27
],
[
17
],
[
16
],
[
19,
25
],
[
21
],
[
22
],
[
7,
14
],
[
29
],
[
0
],
[
10
],
[
24
],
[
30,
34
],
[
32
],
[
1
],
[
3
],
[
4,
12
],
[
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... | [
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct"... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
13,
14,
15,
16
]
}
],
"category": [
"QUAL-CMP"
]
},
{
"sentenc... | benchmark/PDF/ICLR2017_BkXMikqxx.pdf | openreview | benchmark/MD/ICLR2017_BkXMikqxx.md | ICLR 2017 |
H1Fk2Iqex | {
"title": "Fast Chirplet Transform to Enhance CNN Machine Listening - Validation on Animal calls and Speech",
"abstract": "The scattering framework offers an optimal hierarchical convolutional decomposition according to its kernels. Convolutional Neural Net (CNN) can be seen asan optimal kernel decomposition, nev... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11
]
}
}
},
{... | [
[
{
"role": "PC",
"data": {
"comment": "This paper studies efficient signal representations to perform bioacoustic classification based on CNNs. Contrary to image classification, where most useful information can be extracted with spatially localized kernels, bioacoustic signatures are more ... | [
"First, few typos:\n2.2: \" A wavelet is an atom with compact support in time and frequency domain which integrates to 0\"",
"You mean \"often localized\". (Haar wavelets are not and the compact support in fourier and spatial domain is mathematically not possible)",
"2.2: Otherwise, few remarks: The wavelet sho... | [
[
0,
10,
28
],
[
18
],
[
23,
29
],
[
25
],
[
26
],
[
30
],
[
21
],
[
22
],
[
4
],
[
13
],
[
19
],
[
20
],
[
1
],
[
3,
12
],
[
5,
15
],
[
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",
"cor... | [
"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 1 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
12,
13,
14,
61
]
}
],
"category": [
"CLAR-WRT"
]
},
{... | benchmark/PDF/ICLR2017_H1Fk2Iqex.pdf | openreview | benchmark/MD/ICLR2017_H1Fk2Iqex.md | ICLR 2017 |
HJV1zP5xg | {
"title": "Diverse Beam Search: Decoding Diverse Solutions from Neural Sequence Models",
"abstract": "Neural sequence models are widely used to model time-series data. Equally ubiquitous is the usage of beam search (BS) as an approximate inference algorithm to decode output sequences from these models. BS explores... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "Unfortunately, even after the reviewers adjusted their scores, this paper remains very close to the decision boundary. It presents a thorough empirical evaluation, but the improvements are fairly models. The area chair is also not convinced the idea it... | [
"Is it possible to apply the technique on the applications in Jiwei Li's papers, so you can do a fair comparison with Li?",
"Thank you for the question.",
"To be sure we are talking about the same papers, we believe you are referring to:",
"\"Mutual Information and Diverse Decoding Improve Neural Machine Tran... | [
[
4
],
[
5
],
[
3
],
[
0
],
[
9
],
[
10
],
[
14,
16
],
[
17
],
[
1
],
[
6
],
[
8
],
[
11
],
[
13,
15,
18
],
[
2
],
[
7
],
[
12
],
[
19
]
] | [
"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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
1... | benchmark/PDF/ICLR2017_HJV1zP5xg.pdf | openreview | benchmark/MD/ICLR2017_HJV1zP5xg.md | ICLR 2017 |
S1Y0td9ee | {
"title": "Shift Aggregate Extract Networks",
"abstract": "The Shift Aggregate Extract Network SAEN is an architecture for learning representations on social network data.\nSAEN decomposes input graphs into hierarchies made of multiple strata of objects.\nVector representations of each object are learnt by applyin... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
4,
5
... | [
[
{
"role": "PC",
"data": {
"comment": "The authors present a novel architecture, called Shift Aggregate Extract Network (SAEN), for learning representations on social network data. SAEN decomposes input graphs into hierarchies made of multiple strata of objects. The proposed approach gives ... | [
"1. I found the second sentence in the last paragraph of page 2 quite confusing. Could you double check all the subscripts (i's and j's) and make sure they're correct?",
"In particular \"o_j is part of o_j\" and then the definition of the relation inverse... is this supposed to describe the set of composite objec... | [
[
0
],
[
3
],
[
2,
14,
17,
18
],
[
4
],
[
5
],
[
6
],
[
7
],
[
9
],
[
15
],
[
11
],
[
1
],
[
20
],
[
8
],
[
10
],
[
12
],
[
13
],
[
16
],
[
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",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
4
]
}
],
"category": [
"CLAR-NOT"
]
},
{
"sentences": [
{
"role":... | benchmark/PDF/ICLR2017_S1Y0td9ee.pdf | openreview | benchmark/MD/ICLR2017_S1Y0td9ee.md | ICLR 2017 |
B1mAJI9gl | {
"TL;DR": "",
"title": "Towards Understanding the Invertibility of Convolutional Neural Networks",
"abstract": "Several recent works have empirically observed that Convolutional Neural Nets (CNNs) are (approximately) invertible. To understand this approximate invertibility phenomenon and how to leverage it more ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5,
6
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"commen... | [
[
{
"role": "PC",
"data": {
"comment": "While the reviewers found some interest in this work, I'm afraid I have to agree with the critique that the model studied is too simple that its relevance for deep learning is questionable."
}
}
]
] | [
"the paper presents theoretical results on inverting CNN in a random model .",
"There are several confusing notations in the paper, it is not clear at which level in the CNN the invertibility is considered. Seems no pooling or non linearity is considered which is in fine a linear convolution invertibility in the... | [
[
1,
2,
9,
13
],
[
5
],
[
11
],
[
12
],
[
21
],
[
16,
19
],
[
18
],
[
20
],
[
15,
23
],
[
24
],
[
14
],
[
4
],
[
3
],
[
6
],
[
7
],
[
0
],
[
8
],
[
10
],
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
1
]
},
{
"role": "Author",
... | benchmark/PDF/ICLR2017_B1mAJI9gl.pdf | openreview | benchmark/MD/ICLR2017_B1mAJI9gl.md | ICLR 2017 |
HkSOlP9lg | {
"TL;DR": "",
"title": "Recurrent Inference Machines for Solving Inverse Problems",
"abstract": "Inverse problems are typically solved by first defining a model and then choosing an inference procedure. With this separation of modeling from inference, inverse problems can be framed in a modular way. For example,... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
... | [
[
{
"role": "PC",
"data": {
"comment": "This paper presents an approach for learning both a model and inference procedure at the same time using RNNs. The reviewers agree that the idea is interesting, but discussion and considering the responses to the reviews, still felt that more is needed... | [
"Interesting work. The idea of unrolling inference into a deep trainable net has been re-discovered and proved efficient and effective.",
"Could you please also discuss the relationship with two very related papers? It seems to me that MAP inference as a recurrent net and learning the energy models have been cove... | [
[
20
],
[
22
],
[
0
],
[
1,
23
],
[
2,
11,
15
],
[
3,
4,
5,
6
],
[
12,
17
],
[
14
],
[
16
],
[
9
],
[
13
],
[
8
],
[
10
],
[
7
],
[
18
],
[
19
],
[
21
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"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",
"incorrect",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct"... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
1,
2,
4,
... | benchmark/PDF/ICLR2017_HkSOlP9lg.pdf | openreview | benchmark/MD/ICLR2017_HkSOlP9lg.md | ICLR 2017 |
HJSCGD9ex | {
"title": "Beyond Bilingual: Multi-sense Word Embeddings using Multilingual Context",
"abstract": "Word embeddings, which represent a word as a point in a vector space, have become ubiquitous to several NLP tasks. A recent line of work uses bilingual (two languages) corpora to learn a different vector for each sen... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
... | [
[
{
"role": "PC",
"data": {
"comment": "While the reviewers find the core idea intriguing, the method needs a clearer explanation and a more thorough comparison to related work."
}
}
]
] | [
"- you do not model the polysemy in the foreign language. how can the model be extended in that direction?",
"- have you considered exploiting the morphology in foreign languages?",
"- why is table 2 missing Russian and Spanish?\nMINOR issues:",
"- dirichlet process reference has a problem",
"- page 6 engli... | [
[
10
],
[
6
],
[
3
],
[
4
],
[
5,
24
],
[
8
],
[
9
],
[
12,
14
],
[
13,
18
],
[
21
],
[
17
],
[
23
],
[
2
],
[
15
],
[
19
],
[
20
],
[
0
],
[
1
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"c... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
7,
8,
9
]
}
],
"category": [
"QUAL-MET"
]
},
{
"sentences": [
{
... | benchmark/PDF/ICLR2017_HJSCGD9ex.pdf | openreview | benchmark/MD/ICLR2017_HJSCGD9ex.md | ICLR 2017 |
rJo9n9Feg | {
"title": "Chess Game Concepts Emerge under Weak Supervision: A Case Study of Tic-tac-toe",
"abstract": "This paper explores the possibility of learning chess game concepts under weak supervision with convolutional neural networks, which is a topic that has not been visited to the best of our knowledge. We put thi... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The program committee appreciates the authors' response to concerns raised in the reviews. Unfortunately, all reviewers are leaning against accepting the paper. Authors are encouraged to incorporate reviewer feedback in future iterations of this work."... | [
"* Could more quantitative analysis be provided to support the claims about model behavior? How often does the CAM localization pick the correct winning square? Can the learned features be used to predict whose turn it is, where they should move, and if anyone can win?",
"* The chessboard domain is much simpler t... | [
[
8
],
[
13
],
[
30
],
[
31
],
[
34
],
[
36
],
[
39
],
[
40
],
[
3
],
[
4
],
[
10
],
[
24
],
[
32
],
[
15
],
[
6
],
[
9
],
[
35
],
[
37
],
[
47
],
[
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",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
4,
5
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role":... | benchmark/PDF/ICLR2017_rJo9n9Feg.pdf | openreview | benchmark/MD/ICLR2017_rJo9n9Feg.md | ICLR 2017 |
H1_EDpogx | {
"TL;DR": "",
"title": "Near-Data Processing for Machine Learning",
"abstract": "In computer architecture, near-data processing (NDP) refers to augmenting the memory or the storage with processing power so that it can process the data stored therein. By offloading the computational burden of CPU and saving the n... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "This paper is well motivated and clearly written, and is representative of the rapidly growing interdisciplinary area of hardware-software co-design for handling large-scale Machine Learning workloads. In particular, the paper develops a detailed simul... | [
"Bringing computation as close to data as possible is certainly an interesting and highly relevant topic.",
"As a simple pre-review question, I'd like to ask how you see this (SSD-based) method evolving alongside other approaches based on, say, technological advances in data interconnects?",
"For example, HP ju... | [
[
4
],
[
1
],
[
12
],
[
2
],
[
3
],
[
5
],
[
8
],
[
14
],
[
15
],
[
9
],
[
0,
7
],
[
13
],
[
10
],
[
6
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
}
],
"category": [
"SIGN-BRD"
]
},
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
1,
2
]
},
... | benchmark/PDF/ICLR2017_H1_EDpogx.pdf | openreview | benchmark/MD/ICLR2017_H1_EDpogx.md | ICLR 2017 |
rJg_1L5gg | {
"title": "Incremental Sequence Learning",
"abstract": "Deep learning research over the past years has shown that by increasing the scope or difficulty of the learning problem over time, increasingly complex learning problems can be addressed. We study incremental learning in the context of sequence learning, usin... | Reject | [
[
{
"role": "Author",
"data": {
"value": {
"comment": [
0,
1,
2
]
}
}
}
],
[
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "This is an empirical paper which compares three different instantiations of a kind of incremental/curriculum learning for sequences.\n \n The reviews from R1 and R3 (which gave confidence scores of 4) were negative. The main concerns addressed by the r... | [
"Updated version:",
"- added a section with generative results, including movies showing what the network has learned over the course of training",
"- used the ICLR style file",
"- Clarified the contribution (see abstract, intro, and conclusion)",
"- Added figures to illustrate the architecture of the netwo... | [
[
11
],
[
22
],
[
21
],
[
6
],
[
7,
18,
23,
29
],
[
9
],
[
20
],
[
10
],
[
28
],
[
26
],
[
8
],
[
0,
4,
17,
27
],
[
1
],
[
5,
14
],
[
12
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 3 Further Reply",
"data": [
8
]
},
{
"role": "Author",
"data": [
11,
12,
13,
14,
15,
16,
17,
18,
19,
20
... | benchmark/PDF/ICLR2017_rJg_1L5gg.pdf | openreview | benchmark/MD/ICLR2017_rJg_1L5gg.md | ICLR 2017 |
SJCscQcge | {
"title": "Simple Black-Box Adversarial Perturbations for Deep Networks",
"abstract": "Deep neural networks are powerful and popular learning models that achieve state-of-the-art pattern recognition performance on many computer vision, speech, and language processing tasks. However, these networks have also been s... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "While this is an interesting topic, both the method description and experimental setup could be improved."
}
}
]
] | [
"The paper seems to be an incremental twist on the already large field of adversarial images. What's the big advantage?",
"At top of page for (end of first paragraph) you mention that one advantage is that you don't rely on the transferability assumption and that that makes it much more applicable. How so? Can yo... | [
[
39
],
[
38
],
[
1
],
[
2
],
[
5,
13,
17,
21,
24
],
[
14
],
[
29
],
[
37
],
[
40
],
[
0
],
[
4
],
[
11
],
[
20
],
[
22
],
[
23
],
[
35
],
[
6,
25
],
[
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... | [
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
3,
4,
9
]
}
],
"category": [
"ORIG-MTH"
]
},
{
"sentences": [
{
... | benchmark/PDF/ICLR2017_SJCscQcge.pdf | openreview | benchmark/MD/ICLR2017_SJCscQcge.md | ICLR 2017 |
HkCjNI5ex | {
"title": "Regularizing Neural Networks by Penalizing Confident Output Distributions",
"abstract": "We propose regularizing neural networks by penalizing low entropy output distributions. We show that penalizing low entropy output distributions, which has been shown to improve exploration in reinforcement learning... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers agreed that the idea proposed in this paper is sensible and possibly very useful, and that the experiments are thorough with good results. However, they share strong doubts regarding the novelty of the proposed approach. Hopefully the dis... | [
"In this draft, the models are only evaluated (as is standard) by accuracy.",
"But often model probabilities are used as input to decision theory.",
"It seems one consequence of the entropy regularization is that it could un-calibrate probabilities.",
"Consider a problem with two classes. The first class has ... | [
[
4
],
[
7
],
[
34
],
[
18
],
[
23
],
[
19
],
[
32
],
[
26
],
[
3,
29
],
[
10,
11,
12,
13
],
[
14,
15
],
[
33
],
[
5
],
[
21
],
[
27
],
[
0,
2
],
[
1,
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0,
1,
2,
3,
4,
5,
6,
7,
8
]
},
{
"role": "Author",
"data": [
19,
20
]
... | benchmark/PDF/ICLR2017_HkCjNI5ex.pdf | openreview | benchmark/MD/ICLR2017_HkCjNI5ex.md | ICLR 2017 |
Bygq-H9eg | {
"title": "An Analysis of Deep Neural Network Models for Practical Applications",
"abstract": "Since the emergence of Deep Neural Networks (DNNs) as a prominent technique in the field of computer vision, the ImageNet classification challenge has played a major role in advancing the state-of-the-art. While accuracy... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1
]
}
}
},
{
"role": "Reviewer",
... | [
[
{
"role": "PC",
"data": {
"comment": "The paper presents an evaluation of off-the-shelf image classification architectures. The findings are not too surprising and don't provide much new insight."
}
}
]
] | [
"Would any curve that is 1) continuous 2) monotonic 3) bounded up asymptotically (here by the 100% accuracy ceiling) have to naturally be 'hyperbolic' by your definition? This statement pretty much mechanically derives from the nature of the quantity you're measuring, not from the data, so doesn't add any informati... | [
[
12
],
[
22
],
[
0
],
[
6
],
[
10
],
[
13
],
[
18
],
[
19
],
[
25
],
[
14
],
[
2
],
[
21,
27
],
[
1,
17
],
[
3
],
[
7
],
[
8
],
[
15
],
[
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... | [
"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... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1
]
}
],
"category": [
"ORIG-ANL"
]
},
{
"sentences": [
{
"role": "Reviewer 1 ... | benchmark/PDF/ICLR2017_Bygq-H9eg.pdf | openreview | benchmark/MD/ICLR2017_Bygq-H9eg.md | ICLR 2017 |
B186cP9gx | {
"title": "Eigenvalues of the Hessian in Deep Learning: Singularity and Beyond",
"abstract": "We look at the eigenvalues of the Hessian of a loss function before and after training. The eigenvalue distribution is seen to be composed of two parts, the bulk which is concentrated around zero, and the edges which are ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5
... | [
[
{
"role": "PC",
"data": {
"comment": "This is quite an important topic to understand, and I think the spectrum of the Hessian in deep learning deserves more attention. However, all 3 official reviewers (and the public reviewer) comment that the paper needs more work. In particular, there a... | [
"1- In Section 3, why did you look at the eigenvalues of the Hessian at the end of the training and not at different points during the training which could tell us more about the optimization?",
"2- In Figure 7, one could guess that in more complex cases, the magnitude of the learned weights is higher and that is... | [
[
2,
16
],
[
5
],
[
8
],
[
15
],
[
17
],
[
22
],
[
3
],
[
21
],
[
6
],
[
24
],
[
0,
14
],
[
1,
12
],
[
11
],
[
13
],
[
19
],
[
20
],
[
10
],
[
4
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
2,
3
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role":... | benchmark/PDF/ICLR2017_B186cP9gx.pdf | openreview | benchmark/MD/ICLR2017_B186cP9gx.md | ICLR 2017 |
rJM69B5xx | {
"title": "Finding a Jack-of-All-Trades: An Examination of Semi-supervised Learning in Reading Comprehension",
"abstract": "Deep learning has proven useful on many NLP tasks including reading\ncomprehension. However it requires a lot of training data which are not\navailable in some domains of application. Hence w... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5
... | [
[
{
"role": "AC",
"data": {
"comment": "Dear authors and reviewers, this paper is currently very close to the decision boundary for acceptance and would benefit from a bit more discussion.\n\n"
}
},
{
"role": "Author",
"data": {
"value": {
"comment... | [
"- have you tried characterizing the errors made by models with/without pretraining? it would be helpful to observe if what percentage of the errors are the same/different.",
"- can you add a horizontal line to figure2a and 2b so that we can see the performance of 0-shot model and the model trained on only full t... | [
[
1
],
[
19
],
[
20
],
[
8
],
[
9,
12
],
[
10
],
[
13
],
[
4
],
[
5
],
[
0
],
[
2
],
[
3
],
[
6
],
[
7
],
[
11
],
[
15
],
[
18
],
[
14
],
[
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",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
2,
3
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role":... | benchmark/PDF/ICLR2017_rJM69B5xx.pdf | openreview | benchmark/MD/ICLR2017_rJM69B5xx.md | ICLR 2017 |
H1GEvHcee | {
"title": "Annealing Gaussian into ReLU: a New Sampling Strategy for Leaky-ReLU RBM",
"abstract": "Restricted Boltzmann Machine (RBM) is a bipartite graphical model that is used as the building block in energy-based deep generative models. Due to numerical stability and quantifiability of the likelihood, RBM is c... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3
]
}
}
}
],... | [
[
{
"role": "PC",
"data": {
"comment": "This paper identifies a joint distribution for an RBM variant based on leaky-ReLU activations. It also proposes using a sequence of distributions, both as an annealing-based training method, or to estimate log(Z) with AIS.\n \n This paper was borderlin... | [
"Please plot some samples generated by Leaky-ReLU RBM.",
"We will include the sampled images in the appendix. The revision will come in the next week.",
"However, one should note that single layer RBM does not adequately model CIFAR10 and SVHN (considered here) when compared to multilayer models.",
"Moreover ... | [
[
1
],
[
4
],
[
8,
10,
14
],
[
15
],
[
21
],
[
26
],
[
6
],
[
12
],
[
16
],
[
17
],
[
24
],
[
0
],
[
3
],
[
5
],
[
7
],
[
11,
18
],
[
23
],
[
19
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1,
2,
3
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
... | benchmark/PDF/ICLR2017_H1GEvHcee.pdf | openreview | benchmark/MD/ICLR2017_H1GEvHcee.md | ICLR 2017 |
r1IRctqxg | {
"TL;DR": "",
"title": "Sample Importance in Training Deep Neural Networks",
"abstract": "The contribution of each sample during model training varies across training iterations and the model's parameters. We define the concept of sample importance as the change in parameters induced by a sample. In this paper, ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers provided detailed, confident reviews and there was significant discussion between the parties. \n \n Reviewer 2 and 3 felt quite strongly that the paper was a clear reject. Reviewer 1 thought the paper should be accepted.\n \n I was conce... | [
"Can the authors convince me that the overall importance is a sensible measure ? If g_i^t > g_j^t, then clearly example i has a greater impact on \\theta_{t+1} then x_j (ignoring issues of Euclidian vs Fisher metric). The magnitude of gradients will change significantly during learning however, and I am not sure wh... | [
[
3
],
[
25
],
[
5
],
[
18
],
[
27
],
[
15
],
[
23,
30
],
[
4
],
[
11
],
[
20
],
[
21
],
[
22
],
[
9,
12
],
[
14
],
[
17
],
[
19,
24,
29
],
[
26
],
[
0,
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 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
5,
6,
7,
8,
9,
10,
11
]
}
],
"category": [
"... | benchmark/PDF/ICLR2017_r1IRctqxg.pdf | openreview | benchmark/MD/ICLR2017_r1IRctqxg.md | ICLR 2017 |
SkC_7v5gx | {
"title": "The Power of Sparsity in Convolutional Neural Networks",
"abstract": "Deep convolutional networks are well-known for their high computational and memory demands. Given limited resources, how does one design a network that balances its size, training time, and prediction accuracy? A surprisingly effectiv... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1
]
}
}
}
],
[
{
"role": "Revi... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers agreed that the main contribution is the first empirical analysis on large-scale convolutional networks concerning layer-to-layer sparsity. The main concerns were that of novelty (connection-wise sparsity being explored previously but not... | [
"In the conclusion section, the paper says: \"For example, this simple method achieves AlexNet-level accuracy with fewer than 400K parameters and VGG level with roughly 1M parameters.\" I couldn't see these results in any of the graphs. By looking at the graphs, my understanding is there was no experiment conducted... | [
[
10
],
[
3
],
[
0
],
[
1
],
[
7
],
[
12
],
[
8
],
[
6,
13
],
[
11
],
[
4
],
[
2
],
[
5
],
[
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"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
"role": "Reviewer 2 ... | benchmark/PDF/ICLR2017_SkC_7v5gx.pdf | openreview | benchmark/MD/ICLR2017_SkC_7v5gx.md | ICLR 2017 |
Hkz6aNqle | {
"TL;DR": "",
"title": "Deep Error-Correcting Output Codes",
"abstract": "Existing deep networks are generally initialized with unsupervised methods, such as random assignments and greedy layerwise pre-training. This may result in the whole training process (initialization/pre-training + fine-tuning) to be very ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers unanimously recommend rejecting this paper."
}
}
]
] | [
"1. In your experiments, can you clarify that you use the same training procedure for DeepECOC and DAE, do you first pre-train and then fine tune the network with dropout for both methods ?",
"2. Why do you use SVM as the binary classifier and not another method, such as (regularized) logistic regression as this ... | [
[
2
],
[
4
],
[
7,
18
],
[
11
],
[
9
],
[
10
],
[
16
],
[
8
],
[
3
],
[
6
],
[
13
],
[
14
],
[
17
],
[
20
],
[
12
],
[
0
],
[
1
],
[
19
],
[
5
],
[
15
]
] | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"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 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
6,
7
]
}
],
"category": [
"QUAL-REP"
]
},
{
"sentences": [
{
"role":... | benchmark/PDF/ICLR2017_Hkz6aNqle.pdf | openreview | benchmark/MD/ICLR2017_Hkz6aNqle.md | ICLR 2017 |
ByvJuTigl | {
"title": "End-to-End Learnable Histogram Filters",
"abstract": "Problem-specific algorithms and generic machine learning approaches have complementary strengths and weaknesses, trading-off data efficiency and generality. To find the right balance between these, we propose to use problem-specific information encod... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "In many respects, this is a strong paper, in my opinion better than the reviews thus far in the system suggest. The idea of learning the parameters of a state estimation system, even if it is a simple example like a histogram filter, is an interesting ... | [
"1. A strong prior on the motion model is assumed -- linear motion with Gaussian noise. This is applicable only in simple scenarios: such motion model is not directly applicable in settings, where either: (1) the actions relate to the resulting motion in a more complicated way, or (2) the environment effects the tr... | [
[
18
],
[
16
],
[
7
],
[
11
],
[
12
],
[
17
],
[
5
],
[
19
],
[
2
],
[
14
],
[
20
],
[
0,
4,
10
],
[
1
],
[
8
],
[
3
],
[
13
],
[
9
],
[
6
],
[
15
]
] | [
"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",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
4,
5,
7,
8,
9,
10,
11
]
}
],
"category": [
"... | benchmark/PDF/ICLR2017_ByvJuTigl.pdf | openreview | benchmark/MD/ICLR2017_ByvJuTigl.md | ICLR 2017 |
By5e2L9gl | {
"title": "Trusting SVM for Piecewise Linear CNNs",
"abstract": "We present a novel layerwise optimization algorithm for the learning objective of Piecewise-Linear Convolutional Neural Networks (PL-CNNs), a large class of convolutional neural networks. Specifically, PL-CNNs employ piecewise linear non-linearities ... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The authors present a novel layer-wise optimization approach for learning convolutional neural networks with piecewise linear nonlinearities. The proposed approach trains piecewise linear ConvNets layer by layer, reduces the sub-problem into latent str... | [
"1) The baseline in the experiment is a strawman variant of backprop (without batch norm, drop out, etc). I think this is unfair. Batch norm and dropout have been proved to be very useful for CNN models. If the proposed method cannot take advantage of batch norm/dropout, then this is an advantage of backprop.",
"... | [
[
12
],
[
3
],
[
9
],
[
11,
17
],
[
25
],
[
13,
24
],
[
7
],
[
16
],
[
0
],
[
1,
6,
28
],
[
5
],
[
10,
15
],
[
14
],
[
20
],
[
23
],
[
26
],
[
27
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"corr... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
3,
4,
5,
6
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences":... | benchmark/PDF/ICLR2017_By5e2L9gl.pdf | openreview | benchmark/MD/ICLR2017_By5e2L9gl.md | ICLR 2017 |
ByG8A7cee | {
"title": "Reference-Aware Language Models",
"abstract": "We propose a general class of language models that treat reference as an explicit stochastic latent variable. This architecture allows models to create mentions of entities and their attributes by accessing external databases (required by, e.g., dialogue ge... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "All of the reviewers point out clarity problems; while these may have been resolved in an updated version, the reviewers have not expressed that the matter is resolved. There are several questions raised about the use of perplexity, both whether the co... | [
"Hi, I have a few (mostly clarification) questions:",
"(1) Does any of the models presented in Table 4, 5, 6 assume that the mention boundaries are given to the model during test?",
"(2) Do you provide gold co-reference chains for training for the coref-based LM?",
"(3) Do you provide gold co-reference chains... | [
[
11
],
[
12
],
[
16
],
[
17
],
[
20
],
[
21
],
[
13
],
[
14
],
[
15
],
[
18
],
[
19
],
[
30
],
[
37
],
[
7
],
[
8,
32
],
[
23
],
[
24
],
[
27
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"cor... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
1
]
},
{
"role": "Author",
... | benchmark/PDF/ICLR2017_ByG8A7cee.pdf | openreview | benchmark/MD/ICLR2017_ByG8A7cee.md | ICLR 2017 |
B1KBHtcel | {
"title": "Here's My Point: Argumentation Mining with Pointer Networks",
"abstract": "One of the major goals in automated argumentation mining is to uncover the argument structure present in argumentative text. In order to determine this structure, one must understand how different individual components of the ove... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The paper presents an interesting application of pointer networks to the argumentation mining task, and the reviewers found it generally solid. The reviewers generally agree (and I share their concerns) that the contribution on the machine learning sid... | [
"Is it the case that every example in your corpora is structured as a tree, or are there examples of other structures as well (such as forests of non-connected trees)?",
"If it is the case that the data is guaranteed to be tree-structured, why did you only weakly/indirectly enforce this constraint? What tradeoffs... | [
[
8
],
[
9
],
[
4
],
[
6
],
[
11
],
[
17
],
[
2,
7
],
[
15
],
[
3
],
[
12
],
[
13
],
[
1
],
[
0
],
[
14,
16
],
[
5
],
[
10
]
] | [
"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",
"incorrect",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"incorrect"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
2,
4,
5,
7,
9
]
}
],
"category": [
"QUAL-REP"
]
},
{
... | benchmark/PDF/ICLR2017_B1KBHtcel.pdf | openreview | benchmark/MD/ICLR2017_B1KBHtcel.md | ICLR 2017 |
r1GKzP5xx | {
"title": "Recurrent Normalization Propagation",
"abstract": "We propose a LSTM parametrization that preserves the means and variances of the hidden states and memory cells across time. While having training benefits similar to Recurrent Batch Normalization and Layer Normalization, it does not need to estimate st... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "Paper proposes a modification of batch normalization. After the revisions the paper is a much better read. However it still needs more diverse experiments to show the success of the method.\n \n Pros:\n - interesting idea with interesting analysis of t... | [
"Hi,\nSorry for the delay in posting this. One argument you make is that due to this new parametrization, the new method might be cheaper (computationally) then recurrent batch normalization.",
"But you have to compute the norms of each column of the weight.. can you give a comparison about how much you save?",
... | [
[
15
],
[
9
],
[
11
],
[
12,
14
],
[
13
],
[
7
],
[
6
],
[
1
],
[
0
],
[
3
],
[
4
],
[
10
],
[
2,
8
],
[
5
]
] | [
"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"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
3,
4
]
}
],
"category": [
"QUAL-EXP"
]
},
{
"sentences": [
{
... | benchmark/PDF/ICLR2017_r1GKzP5xx.pdf | openreview | benchmark/MD/ICLR2017_r1GKzP5xx.md | ICLR 2017 |
Sk8J83oee | {
"title": "Generative Adversarial Parallelization",
"abstract": "Generative Adversarial Networks (GAN) have become one of the most studied frameworks for unsupervised learning due to their intuitive formulation. They have also been shown to be capable of generating convincing examples in limited domains, such as l... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "This paper was reviewed by three experts. While they find interesting ideas in the manuscript, all three point to deficiencies (problems with the use of GAM metric, lack of convincing results) and unanimously recommend rejection. I do not see a reason ... | [
"Thank you for your submission.",
"One experiment to measure mode coverage would be to synthesize simple data, such as MNIST digits, and then use a ConvNet to count if I get ~10% samples from each digit.",
"Did you think about doing such an experiment?",
"One potential issue with the evaluation proposed in th... | [
[
12
],
[
13
],
[
30
],
[
33
],
[
9
],
[
10
],
[
21
],
[
22
],
[
23
],
[
1
],
[
2
],
[
3
],
[
4
],
[
5
],
[
8
],
[
11
],
[
15,
25
],
[
16
],
[
17
],
[
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",
"incorrect",
"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 Further Reply",
"data": [
1,
2
]
},
{
"role": "A... | benchmark/PDF/ICLR2017_Sk8J83oee.pdf | openreview | benchmark/MD/ICLR2017_Sk8J83oee.md | ICLR 2017 |
S19eAF9ee | {
"title": "Structured Sequence Modeling with Graph Convolutional Recurrent Networks",
"abstract": "This paper introduces Graph Convolutional Recurrent Network (GCRN), a deep learning model able to predict structured sequences of data. Precisely, GCRN is a generalization of classical recurrent neural networks (RNN)... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5,
6
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"commen... | [
[
{
"role": "PC",
"data": {
"comment": "While graph structures are an interesting problem, as the reviewers observed, the paper extends previous work incrementally and the results are not very moving.\n \n pros\n - interesting problem space that has not been thoroughly explored\n cons\n - ex... | [
"I have some questions about section 5.2",
"1. It seems that you used the same dataset (and split) as in Zaremba et al. (2014).",
"But in Zaremba et al. experiment there are 929k training words, 73k validation words, and 82k test words whereas you reported 887k/70k/78k. Can you explain the difference?",
"2. I... | [
[
17
],
[
19
],
[
6,
13
],
[
15,
21
],
[
28
],
[
31,
33
],
[
25
],
[
27
],
[
29
],
[
7
],
[
9
],
[
10
],
[
12
],
[
16
],
[
23
],
[
24
],
[
32,
34
],
[
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",
"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 Further Reply",
"data": [
1,
2
]
},
{
"role": "A... | benchmark/PDF/ICLR2017_S19eAF9ee.pdf | openreview | benchmark/MD/ICLR2017_S19eAF9ee.md | ICLR 2017 |
SkJeEtclx | {
"title": "Memory-augmented Attention Modelling for Videos",
"abstract": "Recent works on neural architectures have demonstrated the utility of attention mechanisms for a wide variety of tasks. Attention models used for problems such as image captioning typically depend on the image under consideration, as well as... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "There appears to be consensus among the reviewers that the paper appears the overstate its contributions: the originality of the proposed temporal modeler (TEM) is limited, and the experimental evaluation (which itself is of good quality!) does not dem... | [
"1) can the authors clarify why Table 1 shows better results for both HAM and TEM? It seems to me that the results are not conclusive as the takeaways from different metrics are in conflict.",
"2) can the authors clarify, in english, what the HAM module is doing? It seems that this is a core contribution of the p... | [
[
11
],
[
14
],
[
29
],
[
10
],
[
27
],
[
28
],
[
39
],
[
1,
8
],
[
9
],
[
23
],
[
40
],
[
43
],
[
44
],
[
47
],
[
13,
17,
19
],
[
41
],
[
45
],
[
2
],
[
20... | [
"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... | [
"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 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
3,
4,
5,
6,
7,
8,
9
]
}
],
"category": [
"QU... | benchmark/PDF/ICLR2017_SkJeEtclx.pdf | openreview | benchmark/MD/ICLR2017_SkJeEtclx.md | ICLR 2017 |
SJJN38cge | {
"TL;DR": "",
"title": "Distributed Transfer Learning for Deep Convolutional Neural Networks by Basic Probability Assignment",
"abstract": "Transfer learning is a popular practice in deep neural networks, but fine-tuning of a large number of parameters is a hard challenge due to the complex wiring of neurons bet... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "All three reviewers appeared to have substantial difficulties understanding the proposed approach due to unclear presentation. This makes it hard for the reviewers to evaluate the originality and potential merits of the proposed approach, and to assess... | [
"1. p. 3, 2.1, paragraph 2, In which cases could\\mathcal{C} and \\mathcal{L} be different?",
"2. p. 4, 2.2, paragraph 3, It's not clear what the authors mean by \"we start by learning a classifier \\phi\". What features are used to train this classifier? Which subset of data is \\phi trained on?",
"3. Do I und... | [
[
17
],
[
0,
20
],
[
8
],
[
9
],
[
12
],
[
21
],
[
7
],
[
14
],
[
16,
24
],
[
22
],
[
27
],
[
11
],
[
18
],
[
13
],
[
26
],
[
1
],
[
2,
4
],
[
3
],
[
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",
"incorrect",
"correct",
"incorrect",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
5,
6,
7,
8
]
}
],
"category": [
"CLAR-NOT"
]
},
{
"sentences":... | benchmark/PDF/ICLR2017_SJJN38cge.pdf | openreview | benchmark/MD/ICLR2017_SJJN38cge.md | ICLR 2017 |
SyCSsUDee | {
"title": "Semantic Noise Modeling for Better Representation Learning",
"abstract": "Latent representation learned from multi-layered neural networks via hierarchical feature abstraction enables recent success of deep learning. Under the deep learning framework, generalization performance highly depends on the lea... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers all expressed concerns with the technical quality of this work. In particular, the reviewers are concerned that ignoring certain entropy terms in the objective is problematic and would require significantly more justification theoreticall... | [
"The paper proposes a new regularization method for neural network based on maximization of the total correlation between input, hidden variables and output. How does this regularization technique compares to other recently proposed techniques such as adversarial training?",
"The method is tested for two specific... | [
[
14
],
[
22
],
[
10
],
[
13,
15
],
[
20
],
[
0
],
[
2
],
[
25
],
[
1
],
[
4,
8,
18
],
[
5,
11
],
[
6,
7,
16,
27
],
[
9
],
[
19
],
[
23
],
[
12
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"inc... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
3,
4,
5,
6,
7
]
}
],
"category": [
"QUAL-CMP"
]
},
{
... | benchmark/PDF/ICLR2017_SyCSsUDee.pdf | openreview | benchmark/MD/ICLR2017_SyCSsUDee.md | ICLR 2017 |
ryhqQFKgl | {
"TL;DR": "",
"title": "Towards Deep Interpretability (MUS-ROVER II): Learning Hierarchical Representations of Tonal Music",
"abstract": "Music theory studies the regularity of patterns in music to capture concepts underlying music styles and composers' decisions. This paper continues the study of building \\emp... | Accept (Poster) | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "Given that all reviewers were positive aobut this paper and given the unusual application domain, we recommend to accept this paper for poster presentation at the main conference."
}
}
]
] | [
"In the paper, I didn't see detailed introductions about the student optimization (right part of Fig. 1). What is S_q? Could the authors elaborate on it? I assume this is part of the author's previous system. It would be great to mention it briefly in the text.",
"Also, what is the meaning of mod_12, diff, order,... | [
[
8
],
[
1,
12
],
[
0,
19
],
[
4
],
[
14
],
[
18,
23
],
[
24
],
[
10
],
[
13
],
[
20
],
[
11
],
[
22
],
[
17
],
[
2
],
[
3
],
[
5
],
[
6
],
[
16
],
[
7
],... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"c... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
3,
4,
5,
64
]
}
],
"category": [
"CLAR-WRT"
]
},
{
"sentences"... | benchmark/PDF/ICLR2017_ryhqQFKgl.pdf | openreview | benchmark/MD/ICLR2017_ryhqQFKgl.md | ICLR 2017 |
HJPmdP9le | {
"TL;DR": "",
"title": "Efficient Summarization with Read-Again and Copy Mechanism",
"abstract": "Encoder-decoder models have been widely used to solve sequence to sequence prediction tasks. However current approaches suffer from two shortcomings. First, the encoders compute a representation of each word taking ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3,
4,
5
... | [
[
{
"role": "AC",
"data": {
"comment": "Reviewers are currently in discussion. Please post a rebuttal to any comments or questions in their reviews.\n\nThanks!"
}
}
],
[
{
"role": "PC",
"data": {
"comment": "This work presents a method for reducing the t... | [
"What is the difference between the copy mechanism in this paper and the other baselines‘? Have you compared the baseline models (with copy mechanism) with your model on the smaller vocabularies?",
"Thank you for the valuable comments and questions.",
"->What is the difference between the copy mechanism in this... | [
[
2
],
[
4
],
[
1
],
[
7,
22,
26,
33
],
[
34
],
[
35
],
[
15,
29
],
[
16
],
[
23
],
[
27,
31
],
[
36
],
[
0
],
[
3
],
[
8
],
[
11
],
[
13,
17
],
[
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",
"correct",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
2,
3,
4,
5,
87,
88,
93,
94,
95,
96,
... | benchmark/PDF/ICLR2017_HJPmdP9le.pdf | openreview | benchmark/MD/ICLR2017_HJPmdP9le.md | ICLR 2017 |
SJttqw5ge | {
"TL;DR": "",
"title": "Communicating Hierarchical Neural Controllers for Learning Zero-shot Task Generalization",
"abstract": "The ability to generalize from past experience to solve previously unseen tasks is a key research challenge in reinforcement learning (RL). In this paper, we consider RL tasks defined a... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12
]
}
... | [
[
{
"role": "PC",
"data": {
"comment": "The paper looks at how natural language instructions can be decomposed into sub-tasks for as-yet-unseen new tasks\n hence the zero-shot generalization, which is considered to be the primary challenge to be solved. \n The precise problem being solved by... | [
"Thank you for writing this nice paper. I have a few questions:",
"1. Please explain the notation in section 3.1. What are the parameters of the convolution layer? only b? Is the fully connected layer using W twice? the term fully connected seems to be an abuse of notation, otherwise please explain. Do you use th... | [
[
8
],
[
1
],
[
2
],
[
4
],
[
9,
15,
16
],
[
18
],
[
20
],
[
24
],
[
27
],
[
33
],
[
34,
35
],
[
36
],
[
12
],
[
22
],
[
13
],
[
6
],
[
10
],
[
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",
"cor... | [
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"... | [
{
"sentences": [
{
"role": "Reviewer",
"data": [
0
]
}
],
"category": [
"N/A"
]
},
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
1
]
},
{
"role": "Author",
... | benchmark/PDF/ICLR2017_SJttqw5ge.pdf | openreview | benchmark/MD/ICLR2017_SJttqw5ge.md | ICLR 2017 |
HkLXCE9lx | {
"title": "RL^2: Fast Reinforcement Learning via Slow Reinforcement Learning",
"abstract": "Deep reinforcement learning (deep RL) has been successful in learning sophisticated behaviors automatically; however, the learning process requires a huge number of trials. In contrast, animals can learn new tasks in just a... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5
... | [
[
{
"role": "PC",
"data": {
"comment": "This paper fits into the mold of a lot of recent work on \"learning to learn\". The idea here is to use an RNN to solve tasks in an MDP, and then use a higher level RL algorithm (TRPO) to learn the parameters of this RNN. It's an interesting idea, and ... | [
"Although the tasks are distinctly flavored (bandit and other classic regret minimization tasks), isn't the algorithm (learning an RNN to solve POMDPs through a standard RL method) similar to e.g. (https://arxiv.org/pdf/1602.01783.pdf [LSTM layer trained with A3C]), https://arxiv.org/pdf/1512.04455.pdf [RNN trained... | [
[
2,
10,
11,
14
],
[
19
],
[
23
],
[
12
],
[
17
],
[
21
],
[
0
],
[
5
],
[
7
],
[
8
],
[
20
],
[
3
],
[
4
],
[
13
],
[
15
],
[
6
],
[
18
],
[
1
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0,
1
]
},
{
"role": "Author",
"data": [
3,
4,
5
]
}
],
"category": [
"QUAL-CMP"
]
},
{
"sentences":... | benchmark/PDF/ICLR2017_HkLXCE9lx.pdf | openreview | benchmark/MD/ICLR2017_HkLXCE9lx.md | ICLR 2017 |
BJ9fZNqle | {
"title": "Multi-modal Variational Encoder-Decoders",
"abstract": "Recent advances in neural variational inference have facilitated efficient training of powerful directed graphical models with continuous latent variables, such as variational autoencoders. However, these models usually assume simple, uni-modal pri... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3,
4
]
}
... | [
[
{
"role": "PC",
"data": {
"comment": "This paper explores a variational autoencoder variant.\n \n ICLR gives authors some respect that other conferences don't. It is flexible about the length of the paper, and allows revisions to be submitted. The understanding should be that authors shoul... | [
"Since it gives such dramatically better results on 20-NG, what is the exact difference between G-NVDM and NVDM? Is it the interpolation/gating between prior and posterior? Or is it the learned prior? Why does learning a variance and mean for the Gaussian prior help, when the latent space is already transformed by ... | [
[
17
],
[
6
],
[
8,
10,
32
],
[
12
],
[
13
],
[
18
],
[
19
],
[
26,
28
],
[
27
],
[
25
],
[
31
],
[
1
],
[
4
],
[
7
],
[
15
],
[
16
],
[
23
],
[
29
],
[
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",
"incorrect",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1,
2,
3
]
}
],
"category": [
"QUAL-MET"
]
},
{
"sentences": [
{
... | benchmark/PDF/ICLR2017_BJ9fZNqle.pdf | openreview | benchmark/MD/ICLR2017_BJ9fZNqle.md | ICLR 2017 |
BkCPyXm1l | {
"TL;DR": "",
"title": "SoftTarget Regularization: An Effective Technique to Reduce Over-Fitting in Neural Networks",
"abstract": "Deep neural networks are learning models with a very high capacity and therefore prone to over-fitting. Many regularization techniques such as Dropout, DropConnect, and weight decay ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers unanimously recommend rejection."
}
}
]
] | [
"- The method as proposed requires memory that grows linearly both with the number of training samples and the output dimensionality (i.e. number of target classes). How would you propose to scale this to large datasets?",
"- A commonly used regularization scheme is early stopping, whereby you monitor a validatio... | [
[
25
],
[
32
],
[
16,
29,
31,
34
],
[
17,
24
],
[
1
],
[
5
],
[
7
],
[
11
],
[
14
],
[
35
],
[
2,
6,
8,
15
],
[
10
],
[
13
],
[
18
],
[
19
],
[
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... | [
"incorrect",
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
4,
5,
6
]
}
],
"category": [
"QUAL-MET"
]
},
{
"sentences": [
{
... | benchmark/PDF/ICLR2017_BkCPyXm1l.pdf | openreview | benchmark/MD/ICLR2017_BkCPyXm1l.md | ICLR 2017 |
BkV4VS9ll | {
"TL;DR": "Pruning algorithms reveal fundamental insights into neural network learning representations",
"title": "The Incredible Shrinking Neural Network: New Perspectives on Learning Representations Through The Lens of Pruning",
"abstract": "How much can pruning algorithms teach us about the fundamentals of le... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3
]
}
}
}
],... | [
[
{
"role": "PC",
"data": {
"comment": "The paper does not seem to have enough novelty, and the contribution is not clear enough due to presentation issues."
}
}
]
] | [
"The page limit are 8 pages, you have 23.",
"Hi,\nThank you for your question. It is mentioned in the call for papers (http://www.iclr.cc/doku.php?id=iclr2017:callforpapers) that there is no strict page limit enforced. 8 pages (without References) seems to be the recommended length but is not seemingly a hard lim... | [
[
2
],
[
5,
13
],
[
10
],
[
12,
17
],
[
22
],
[
24
],
[
25
],
[
14
],
[
26
],
[
28
],
[
0
],
[
19
],
[
4,
15
],
[
6,
27
],
[
16,
21
],
[
20
],
[
8
],
[
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... | [
"incorrect",
"correct",
"correct",
"correct",
"incorrect",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"... | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1,
2,
3
]
}
],
"category": [
"POL-IMP"
]
},
{
"sentences": [
{
... | benchmark/PDF/ICLR2017_BkV4VS9ll.pdf | openreview | benchmark/MD/ICLR2017_BkV4VS9ll.md | ICLR 2017 |
ByG4hz5le | {
"TL;DR": "",
"title": "Adaptive Feature Abstraction for Translating Video to Language",
"abstract": "Previous models for video captioning often use the output from a specific layer of a Convolutional Neural Network (CNN) as video representations, preventing them from modeling rich, varying context-dependent sem... | Invite to Workshop Track | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5,
6
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"commen... | [
[
{
"role": "PC",
"data": {
"comment": "Reviewers feel the work is well executed and that the model makes sense, but two of the reviewers were not convinced that the proposed method contains enough novelty in light of prior work. The comparison of the soft vs hard attention model variations ... | [
"Nice paper and convincing results! Here are some pre-review questions:",
"1) As a significant part of the problems you have to solve are due to different dimensions of the feature layers, have you tried a simple baseline consisting of simply reshaping video tensors to the same size? You are already doing it for ... | [
[
3
],
[
8
],
[
19
],
[
4,
22
],
[
9
],
[
25
],
[
17
],
[
10
],
[
13
],
[
20
],
[
26
],
[
27
],
[
2,
7,
18
],
[
11
],
[
12
],
[
1
],
[
6
],
[
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",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"incorrect",
"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 Further Reply",
"data": [
1,
2
]
},
{
"role": "A... | benchmark/PDF/ICLR2017_ByG4hz5le.pdf | openreview | benchmark/MD/ICLR2017_ByG4hz5le.md | ICLR 2017 |
BysZhEqee | {
"TL;DR": "",
"title": "Marginal Deep Architectures: Deep learning for Small and Middle Scale Applications",
"abstract": "In recent years, many deep architectures have been proposed in different fields. However, to obtain good results, most of the previous deep models need a large number of training data. In thi... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "The reviewers unanimously recommend rejection."
}
}
]
] | [
"What is the A matrix in equation 3?",
"For classification on Cifar-10, wouldn't the simplest baseline be to compare MFA to state-of-the-art CNNs trained from random weights?",
"Many thanks for your comments!",
"1. In equation 3, \"a^k\" is the activation output of the k-th hidden layer.",
"By the way, do y... | [
[
22
],
[
0
],
[
6,
18
],
[
12
],
[
10,
13
],
[
17
],
[
21
],
[
1
],
[
4
],
[
5
],
[
15
],
[
9
],
[
20
],
[
7
],
[
19
],
[
2
],
[
3
],
[
11
],
[
8
],
[
... | [
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"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 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
3,
4
]
}
],
"category": [
"CLAR-NOT"
]
},
{
"sentences": [
{
"role":... | benchmark/PDF/ICLR2017_BysZhEqee.pdf | openreview | benchmark/MD/ICLR2017_BysZhEqee.md | ICLR 2017 |
ryXZmzNeg | {
"title": "Improving Sampling from Generative Autoencoders with Markov Chains",
"abstract": "We focus on generative autoencoders, such as variational or adversarial autoencoders, which jointly learn a generative model alongside an inference model. Generative autoencoders are those which are trained to softly enfor... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
1,
2,
3,
4,
5,
... | [
[
{
"role": "AC",
"data": {
"comment": "While you cite Rezende et al. 2014 when referring to VAEs, you claim the main contribution of your work is the generation of posterior samples from a Markov Chain. However, Rezende et al. 2014 presented a very similar idea. Let me quote them directly:\... | [
"For the sake of understanding, can you position your paper relative to https://arxiv.org/abs/1410.6460 and https://arxiv.org/abs/1612.00005 ? Note that I'm not discussing precedence. I'd like to understand the contributions of this paper compared to the above two and GSNs.",
"1) Markov Chain Monte Carlo and Vari... | [
[
6
],
[
7
],
[
12
],
[
2
],
[
3
],
[
4
],
[
8
],
[
9,
15
],
[
11
],
[
1
],
[
0
],
[
5,
14
],
[
17
],
[
10
],
[
16
],
[
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 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
1... | benchmark/PDF/ICLR2017_ryXZmzNeg.pdf | openreview | benchmark/MD/ICLR2017_ryXZmzNeg.md | ICLR 2017 |
SJqaCVLxx | {
"title": "New Learning Approach By Genetic Algorithm In A Convolutional Neural Network For Pattern Recognition",
"abstract": "Almost all of the presented articles in the CNN are based on the error backpropagation algorithm and calculation of derivations of error, our innovative proposal refers to engaging TICA ... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"comment": [
2,
3,
4,
5,
... | [
[
{
"role": "PC",
"data": {
"comment": "This paper is unfortunately quite unclear and unreadable and nowhere near ready for any conference.\n I would advise the authors to 1) restructure their paper to present first some type of context and identify a problem that they are trying to solve, 2... | [
"I do not understand the algorithm and the motivation behind the algorithm proposed? Why is back propagation a bad idea? Is it the computational speed? or convergence issues? If so then why not compare your algorithm with the standard backprop algorithm in the experiments section?",
"More generally the paper is e... | [
[
19
],
[
20
],
[
0
],
[
1
],
[
2,
6
],
[
3,
5,
7,
8,
15,
21
],
[
12
],
[
13
],
[
18
],
[
4,
10
],
[
11
],
[
14
],
[
17
],
[
9
],
[
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",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0
]
},
{
"role": "Author",
"data": [
2,
3,
4,
5,
6
]
}
],
"category": [
"CLAR-WRT",
"QUAL-CMP",... | benchmark/PDF/ICLR2017_SJqaCVLxx.pdf | openreview | benchmark/MD/ICLR2017_SJqaCVLxx.md | ICLR 2017 |
Bkp_y7qxe | {
"title": "Unsupervised Deep Learning of State Representation Using Robotic Priors ",
"abstract": "Our understanding of the world depends highly on how we represent it. Using background knowledge about its complex underlying physical rules, our brain can produce intuitive and simplified representations which it c... | Reject | [
[
{
"role": "Reviewer",
"data": {
"value": {
"question": [
0,
1,
2,
3,
4,
5,
6
]
}
}
},
{
"role": "Author",
"data": {
"value": {
"commen... | [
[
{
"role": "PC",
"data": {
"comment": "The authors apply an already-published method for state representation learning in a very simple experimental scenario. They give no additional contribution or comparison, nor do they offer any empirical or analytical study."
}
}
]
] | [
"Hello,",
"A few questions:",
"1. Why do you use so few sequences for training and only one (!) for validation? Why do you not simulate more data? Are you sure the training and the validation sets are sufficiently different?",
"2. You argue deep network is better because it can learn invariance to illuminatio... | [
[
9
],
[
10
],
[
12,
15
],
[
2
],
[
5
],
[
7
],
[
16
],
[
8
],
[
0
],
[
1
],
[
3
],
[
4
],
[
11
],
[
13
],
[
14
],
[
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",
"correct",
"correct",
"incorrect",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct",
"correct"
] | [
{
"sentences": [
{
"role": "Reviewer 1 Further Reply",
"data": [
0,
1,
2
]
},
{
"role": "Author",
"data": [
8,
9,
10,
11
]
}
],
"category": [
"QUAL-EXP"
... | benchmark/PDF/ICLR2017_Bkp_y7qxe.pdf | openreview | benchmark/MD/ICLR2017_Bkp_y7qxe.md | ICLR 2017 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.