paper_id stringlengths 19 21 | paper_title stringlengths 8 170 | paper_abstract stringlengths 8 5.01k | paper_acceptance stringclasses 18
values | meta_review stringlengths 29 10k | label stringclasses 3
values | review_ids list | review_writers list | review_contents list | review_ratings list | review_confidences list | review_reply_tos list |
|---|---|---|---|---|---|---|---|---|---|---|---|
iclr_2021_Y5TgO3J_Glc | Neurosymbolic Deep Generative Models for Sequence Data with Relational Constraints | Recently, there has been significant progress designing deep generative models that generate realistic sequence data such as text or music. Nevertheless, it remains difficult to incorporate high-level structure to guide the generative process. We propose a novel approach for incorporating structure in the form of relat... | withdrawn-rejected-submissions | This paper was pretty borderline, but ultimately I am recommending rejection, for the following reason:
The two most negative reviewers (in terms of original score) were concerned about both the quality of the evaluation
and whether the evaluation metrics actually meant what the paper claimed they meant.
The authors ... | train | [
"CWS1rIhHkFY",
"L1zo_hA9zcx",
"f5rv861k2Uf",
"wINdhSo3D1",
"Vj3MDVsydGU"
] | [
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer"
] | [
"**Summary**\n\nThis paper presents a novel method of incorporating structural and relational constraints into the generative process of sequences. This approach is explored in both the music and poetry domains using several generative approaches. An analysis of the resulting sequences finds that they successfully ... | [
4,
6,
-1,
7,
6
] | [
4,
4,
-1,
4,
4
] | [
"iclr_2021_Y5TgO3J_Glc",
"iclr_2021_Y5TgO3J_Glc",
"iclr_2021_Y5TgO3J_Glc",
"iclr_2021_Y5TgO3J_Glc",
"iclr_2021_Y5TgO3J_Glc"
] |
iclr_2021_zFM0Uo_GnYE | On the Importance of Looking at the Manifold | Data rarely lies on uniquely Euclidean spaces. Even data typically represented in regular domains, such as images, can have a higher level of relational information, either between data samples or even relations within samples, e.g., how the objects in an image are linked. With this perspective our data points can be e... | withdrawn-rejected-submissions | While the motivation of the paper is interesting the reviewers expressed concerns about the experimental setup, comparison to related work, and paper framing. For experiments, it was unclear why authors compared such disparate methods instead of more fine-grained adjustments (e.g., such as corrupting graphs as suggeste... | train | [
"gj1W5-VatE",
"ZM9lGrL9s9A",
"cQ9A4lkx6AM",
"EmdvrOVHYId",
"caC1m8LsR8P"
] | [
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"**Summary**:\nThis paper investigates different ways of incorporating topological information about the data in the machine learning models. The paper introduces a novel loss that aims to enforce the relational information between data points into the embedding space learned by a Vae on the node features. The expe... | [
4,
-1,
3,
5,
4
] | [
3,
-1,
4,
4,
4
] | [
"iclr_2021_zFM0Uo_GnYE",
"iclr_2021_zFM0Uo_GnYE",
"iclr_2021_zFM0Uo_GnYE",
"iclr_2021_zFM0Uo_GnYE",
"iclr_2021_zFM0Uo_GnYE"
] |
iclr_2021_4qgEGwOtxU | Importance and Coherence: Methods for Evaluating Modularity in Neural Networks | As deep neural networks become more advanced and widely-used, it is important to understand their inner workings. Toward this goal, modular interpretations are appealing because they offer flexible levels of abstraction aside from standard architectural building blocks (e.g., neurons, channels, layers). In this paper, ... | withdrawn-rejected-submissions | The paper present an approach for defending for, and search for, 'modularity' in neural networks, as a step to better interpretations of their functional structure. This is an interesting, and highly original approach, as recognised by the reviewers. However, there was also some discussion about what exactly can be le... | train | [
"7LXScpW7tW6",
"MaRfTci6mK9",
"eMhFzLsChol",
"-aNZK9b_h7",
"lmbNMjZrG15",
"uSawEznuE-",
"u39tsuibbVy",
"c2TOtugPSdb",
"ZeduGWY7C2",
"fD6fBvPSC6r",
"ghzQRTMGy0y",
"R30mqifim5o",
"NXju7UC4mN6",
"pP96MWqcgLH",
"tfIR4X6HUQG",
"pgsBVKNObcP",
"IKY7SYg-6Cy"
] | [
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"The manuscript introduces an approach, based on importance and coherence, for evaluation whether a partitioning of a network exhibits modular characteristics. \nImportance refers to how crucial is a neuron , or set of neurons, to the performance of a network on a given task, e.g. classification.\nCoherence refers ... | [
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3
] | [
"iclr_2021_4qgEGwOtxU",
"iclr_2021_4qgEGwOtxU",
"lmbNMjZrG15",
"uSawEznuE-",
"NXju7UC4mN6",
"u39tsuibbVy",
"fD6fBvPSC6r",
"ghzQRTMGy0y",
"iclr_2021_4qgEGwOtxU",
"pP96MWqcgLH",
"R30mqifim5o",
"pgsBVKNObcP",
"7LXScpW7tW6",
"IKY7SYg-6Cy",
"iclr_2021_4qgEGwOtxU",
"iclr_2021_4qgEGwOtxU",
... |
iclr_2021_4xzY5yod28y | Scheduled Restart Momentum for Accelerated Stochastic Gradient Descent | Stochastic gradient descent (SGD) algorithms, with constant momentum and its variants such as Adam, are the optimization methods of choice for training deep neural networks (DNNs). There is great interest in speeding up the convergence of these methods due to their high computational expense. Nesterov accelerated gradi... | withdrawn-rejected-submissions | The reviewers acknowledge that the paper has some promising experiments. However, they think that the theoretical contributions are not rigorous, specifically the assumption in Theorem 2. It is true that the main part of the proof relies on this assumption. The main question is whether this assumption holds or not. The... | test | [
"xdUsxnVs_Sd",
"y583XsqbzAD",
"uHY6qP1ufgw",
"EvI5RuAnBkP",
"7vlQcHXweR-",
"P0tXuy50iIB",
"wCkGCDX8q4",
"hZqS5amedYD",
"JYF7f4Zfzqo",
"fZVhCbWMpBa",
"AngOW9d-wJZ",
"W4VqWKOqBMk",
"Hsfu3RI4Tys",
"6WlkO_gBx6u",
"fpOeRYwi1P2",
"VU0je3a14Iu",
"VZ3ym8Dxo6n",
"xNH7Ii0r1Qd",
"Mn33Od5hLg... | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official... | [
"UPDATE:\n\nI find the main contribution of the work to be the empirical analysis. I personally liked this paper, but I must agree with the other reviewers that improving the theoretical results will strengthen the paper's impact. This would require a major revision not suited for a conference rebuttal, and so, for... | [
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
4,
5,
5
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
3,
4
] | [
"iclr_2021_4xzY5yod28y",
"XCvo5cMaQVx",
"JYF7f4Zfzqo",
"hZqS5amedYD",
"AngOW9d-wJZ",
"iclr_2021_4xzY5yod28y",
"P0tXuy50iIB",
"fpOeRYwi1P2",
"fZVhCbWMpBa",
"xNH7Ii0r1Qd",
"Hsfu3RI4Tys",
"6WlkO_gBx6u",
"VZ3ym8Dxo6n",
"XCvo5cMaQVx",
"xdUsxnVs_Sd",
"xdUsxnVs_Sd",
"gfgwxTXfn2t",
"XV_Gu2... |
iclr_2021_-kigPjfTIGd | SSW-GAN: Scalable Stage-wise Training of Video GANs | Current state-of-the-art generative models for videos have high computational requirements that impede high resolution generations beyond a few frames. In this work we propose a stage-wise strategy to train Generative Adversarial Networks (GANs) for videos. We decompose the generative process to first produce a downsa... | withdrawn-rejected-submissions | This paper proposes a GAN for video generation based on stagewise training over different resolutions, addressing scalability issues with previous approaches. Reviewers noted that the paper is clearly written, proposes a method that improves upon the DVD-GAN architecture by reducing training time and memory consumption... | train | [
"oI6oCp8aam6",
"mnnZKAle0kG",
"eHpNf7uVWH",
"EdpQSmZXPAU",
"ewD_R35cSYa",
"YZeNtTYsDOx",
"RpjOZtyK7q",
"gx7x9kW-1T",
"vrBEyU7v0Y",
"QFSTgE_38aF",
"qv2Srjm9JV2",
"yD7phq65eLO",
"Tgvacj42HcS",
"S_P6fy7E5qA",
"EBNqdQHe1k2",
"zpYaN8vCLe",
"sa9JPWHZ-M"
] | [
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"The paper proposes a GAN-based model which generates videos in multiple stages. The main idea is the upsampling of the spatio-temporal resolution upon addition of a stage. This is the key feature of the proposed model allowing the model to generate videos of higher temporal resolution while using significantly les... | [
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
3,
6,
7
] | [
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3,
3,
5
] | [
"iclr_2021_-kigPjfTIGd",
"YZeNtTYsDOx",
"Tgvacj42HcS",
"ewD_R35cSYa",
"RpjOZtyK7q",
"iclr_2021_-kigPjfTIGd",
"yD7phq65eLO",
"oI6oCp8aam6",
"oI6oCp8aam6",
"EBNqdQHe1k2",
"S_P6fy7E5qA",
"zpYaN8vCLe",
"sa9JPWHZ-M",
"iclr_2021_-kigPjfTIGd",
"iclr_2021_-kigPjfTIGd",
"iclr_2021_-kigPjfTIGd",... |
iclr_2021_3-a23gHXQmr | Parametric Density Estimation with Uncertainty using Deep Ensembles | In parametric density estimation, the parameters of a known probability density are typically recovered from measurements by maximizing the log-likelihood. Prior knowledge of measurement uncertainties is not included in this method -- potentially producing degraded or even biased parameter estimates.
We propose a... | withdrawn-rejected-submissions | The reviewers are in consensus that the manuscript is not ready for publication in its current form: more comprehensive evaluation, and careful analysis (either theoretical or empirical) of the simple-but-effective methodology would improve the quality further. The discussion was constructive and helped the authors to ... | train | [
"jgFL6_0pkCt",
"kC9--ElTg8e",
"hTknGesryf3",
"i5htw0dNbUU",
"tcDjYFRYaKs",
"JNx2bWV_5_",
"jLH1lhK0VWz",
"RXzTVJr1zim",
"6nA4AaRtsZX",
"e7vYk5KJv4e",
"l133YASEvL",
"60d_gUUWuqK"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"Update: After reading the other reviews/responses, I'm keeping my score of 5, due to pervasive concerns about the narrow focus of the experiments and incremental novelty of the method.\n\nThis paper proposes a method that uses ensembles of deep neural networks for parametric feature density estimation and enables ... | [
5,
4,
-1,
-1,
-1,
-1,
5,
-1,
-1,
-1,
-1,
5
] | [
3,
3,
-1,
-1,
-1,
-1,
3,
-1,
-1,
-1,
-1,
3
] | [
"iclr_2021_3-a23gHXQmr",
"iclr_2021_3-a23gHXQmr",
"tcDjYFRYaKs",
"iclr_2021_3-a23gHXQmr",
"l133YASEvL",
"60d_gUUWuqK",
"iclr_2021_3-a23gHXQmr",
"6nA4AaRtsZX",
"kC9--ElTg8e",
"jgFL6_0pkCt",
"jLH1lhK0VWz",
"iclr_2021_3-a23gHXQmr"
] |
iclr_2021_HeEzgm-f4g1 | On Batch-size Selection for Stochastic Training for Graph Neural Networks | Batch size is an important hyper-parameter for training deep learning models with stochastic gradient decent (SGD) method, and it has great influence on the training time and model performance. We study the batch size selection problem for training graph neural network (GNN) with SGD method.
To reduce the traini... | withdrawn-rejected-submissions | All the reviewers find the problem studied in the interesting. However all of them raise concerns about the assumptions made in the paper for the analysis. Reviewers find the assumptions very limiting and far from the practical training of GNNs. Improving the analysis by relaxing the assumptions further can significant... | train | [
"s7USH2rCCP",
"GOnY7v9uI9g",
"oBbAoUm2I6d",
"3i8f9yJEX0N",
"SYvKOZtoewE",
"uRG8bjXlkhP",
"xFB4RlqwqAT",
"gg1Yp2wbM_f"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"In this paper, the authors study an important problem on how the choice of batch size (i.e., the number of sampled nodes) affects the training efficiency and accuracy of graph neural networks (GNN). Focusing on the layer-wise and graph-wise sampling for training, the authors theoretically characterize the impact o... | [
5,
-1,
-1,
-1,
-1,
4,
4,
4
] | [
3,
-1,
-1,
-1,
-1,
4,
4,
4
] | [
"iclr_2021_HeEzgm-f4g1",
"uRG8bjXlkhP",
"s7USH2rCCP",
"xFB4RlqwqAT",
"gg1Yp2wbM_f",
"iclr_2021_HeEzgm-f4g1",
"iclr_2021_HeEzgm-f4g1",
"iclr_2021_HeEzgm-f4g1"
] |
iclr_2021_rYt0p0Um9r | Do Deeper Convolutional Networks Perform Better? | Over-parameterization is a recent topic of much interest in the machine learning community. While over-parameterized neural networks are capable of perfectly fitting (interpolating) training data, these networks often perform well on test data, thereby contradicting classical learning theory. Recent work provided... | withdrawn-rejected-submissions | Reviewers appreciate the numerical results presented in this paper. However, the paper needs a more rigorous theoretical investigation of the empirical phenomenon, or a more comprehensive empirical exploration to pinpoint the key factors. I recommend the authors to incorporate the suggestions from the reviewers and sub... | test | [
"fbuiJYV5G3",
"9Ykn_NTOd2",
"FYjd_6wP0DC",
"6ZHPAwsby3Q",
"zr920ORltN1",
"zbT6Wq8ZwDu",
"V6IIMbB1-Rh",
"gZXwt4rKvq6",
"r8rDbxx9X_8",
"icAawgZ_z9p",
"TqPc_OdFuq_",
"dFmIFSBZR7V",
"rkcmeuRb_pi",
"sNx3TbZSzEv",
"c9wWorobwZF",
"bHMXxSGNQuw",
"NuwS-uvpxxf",
"b8i3yWfW9t4",
"YD5Oyi7AyCP... | [
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"... | [
"### Post-rebuttal update:\n\nI am raising my score from 4 to 5, to recognize extensive updates to the paper experiments and numerous clarifications during the discussion, as well as to acknowledge that some of my initial concerns were not justified, e.g. asking for similar studies for MLPs (which was effectively d... | [
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
6,
6
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
5,
4
] | [
"iclr_2021_rYt0p0Um9r",
"iclr_2021_rYt0p0Um9r",
"zr920ORltN1",
"gZXwt4rKvq6",
"zbT6Wq8ZwDu",
"6ZHPAwsby3Q",
"icAawgZ_z9p",
"r8rDbxx9X_8",
"V6IIMbB1-Rh",
"dFmIFSBZR7V",
"c9wWorobwZF",
"TqPc_OdFuq_",
"b8i3yWfW9t4",
"rkcmeuRb_pi",
"bHMXxSGNQuw",
"sNx3TbZSzEv",
"dnmeD8RSciS",
"YD5Oyi7A... |
iclr_2021_MPO4oML_JC | Coordinated Multi-Agent Exploration Using Shared Goals | Exploration is critical for good results of deep reinforcement learning algorithms and has drawn much attention. However, existing multi-agent deep reinforcement learning algorithms still use mostly noise-based techniques. It was recognized recently that noise-based exploration is suboptimal in multi-agent settings, an... | withdrawn-rejected-submissions | The paper presents an approach to multi-agent coordination using goal-driven exploration on subspaces of the observation space.
The results of the paper show that the authors' approach performs baselines on grid worlds and two tasks from the StartCraft Multi-agent Challenge. While the rebuttal clarified many points ra... | train | [
"4O1aLIyii_R",
"2aaWYWg1sLF",
"a0Ra2P0kp3",
"yQGX5U3GYMT",
"VnjYjbx61rS",
"riZodbOe1IQ",
"Ozd8isoj1gd",
"CptaCupegeq",
"Za0xr7LSuO3",
"AFPeknCI_oT"
] | [
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"---Post rebuttal---\n\nThank you for the detailed response. My main concern was regarding the scalability of the method to larger environments, e.g. w/ visual state space. I agree with the other reviewers regarding limited applicability of the method, and maintain my original score (Weak Reject).\n\n---\n\nThe pap... | [
5,
6,
-1,
-1,
-1,
-1,
-1,
-1,
4,
5
] | [
3,
4,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4
] | [
"iclr_2021_MPO4oML_JC",
"iclr_2021_MPO4oML_JC",
"yQGX5U3GYMT",
"Ozd8isoj1gd",
"Za0xr7LSuO3",
"2aaWYWg1sLF",
"AFPeknCI_oT",
"4O1aLIyii_R",
"iclr_2021_MPO4oML_JC",
"iclr_2021_MPO4oML_JC"
] |
iclr_2021_zCu1BZYCueE | Response Modeling of Hyper-Parameters for Deep Convolutional Neural Networks | Hyper-parameter optimization (HPO) is critical in training high performing Deep Neural Networks (DNN). Current methodologies fail to define an analytical response surface and remain a training bottleneck due to their use of additional internal hyper-parameters and lengthy evaluation cycles. We demonstrate that the low-... | withdrawn-rejected-submissions | This paper studies the important problem of efficiently identifying good hyperparameters for convolutional neural networks. The proposed approach is based on using an SVD of unfolded weight tensors to build a response surface that can be optimized with a dynamic tracking algorithm. Reviewers raised a number of concerns... | train | [
"5VX1yfhfr37",
"qJQCtR11NR_",
"gxYImLhINcq",
"AlikF7AjWv",
"nnOujBPpNMI",
"4ZKN0liGcB",
"Du_demeJSH",
"daFLVTMUGcq",
"xn7nM2Uuvwl",
"EAlnoDZWs64",
"ymYBNgM0-_T",
"ihV52pW-MuQ",
"yHNg9RpdCZo",
"UOw_Ik_lLeM"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"###############################################################\n\nSummary:\n\nThis paper used the knowledge gain to provide an algorithm to choose the initial learning rate. The paper explained the reason that choosing the specific response function and demonstrated the effectiveness of the algorithm by several e... | [
5,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4
] | [
3,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3
] | [
"iclr_2021_zCu1BZYCueE",
"iclr_2021_zCu1BZYCueE",
"iclr_2021_zCu1BZYCueE",
"yHNg9RpdCZo",
"iclr_2021_zCu1BZYCueE",
"Du_demeJSH",
"daFLVTMUGcq",
"yHNg9RpdCZo",
"iclr_2021_zCu1BZYCueE",
"5VX1yfhfr37",
"qJQCtR11NR_",
"UOw_Ik_lLeM",
"iclr_2021_zCu1BZYCueE",
"iclr_2021_zCu1BZYCueE"
] |
iclr_2021_Q1aiM7sCi1 | Fuzzy c-Means Clustering for Persistence Diagrams | Persistence diagrams concisely represent the topology of a point cloud whilst having strong theoretical guarantees. Most current approaches to integrating topological information into machine learning implicitly map persistence diagrams to a Hilbert space, resulting in deformation of the underlying metric structure whi... | withdrawn-rejected-submissions | This paper is overall clearly written, and the proposed approach of performing clustering on the space of persistence diagrams can be a significant contribution.
However, during the discussion, the reviewers share the concern about insufficient empirical evaluation. In particular, datasets are limited and (Lacombe et ... | val | [
"6iofO9Cr1e",
"CWomj3cJi_2",
"6wTpzgLuFL2",
"B3pEyKPTj-l",
"ir3-GLdJPFf",
"IBAYHfya-aY",
"dYMZpX1it7X",
"bzQiL98KJ0J",
"PoDoSU0onRA",
"zioJpjG8WIa",
"0DzH9WZsaW",
"tTjc_D7dTm",
"WEdl0dkAmP"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer"
] | [
"The authors propose a novel algorithm for the fuzzy clustering of persistence diagrams. To determine cluster centroids, the Wasserstein-2 distance is used to minimize the weighted Fr\\’echet mean between a potential cluster center and all PDs considered for clustering. The authors proof convergence of the clusteri... | [
6,
6,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3
] | [
4,
5,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5
] | [
"iclr_2021_Q1aiM7sCi1",
"iclr_2021_Q1aiM7sCi1",
"iclr_2021_Q1aiM7sCi1",
"0DzH9WZsaW",
"zioJpjG8WIa",
"bzQiL98KJ0J",
"PoDoSU0onRA",
"WEdl0dkAmP",
"6wTpzgLuFL2",
"tTjc_D7dTm",
"6iofO9Cr1e",
"CWomj3cJi_2",
"iclr_2021_Q1aiM7sCi1"
] |
iclr_2021_R43miizWtUN | Analysing the Update step in Graph Neural Networks via Sparsification | In recent years, Message-Passing Neural Networks (MPNNs), the most prominent Graph Neural Network (GNN) framework, have celebrated much success in the analysis of graph-structured data. In MPNNs the computations are split into three steps, Aggregation, Update and Readout. In this paper a series of models to successivel... | withdrawn-rejected-submissions | The paper focuses on the update step in Message-Passing Neural Networks, specifically for GNN. A series of sparse variants of the update step, say complete removal and expander graphs with varying density, are compared in empirical studies. The findings are quite useful for practice, and the paper is organized and writ... | train | [
"yn9_H4V90nq",
"5Oc5pZ8BljM",
"9ejokxJDlDU",
"KFJGy_biDPV",
"bkFuRu3PHs5",
"287aS9XMjl",
"q9eV4HPXk5e",
"JxgoG7blyH-",
"iREFEsHL7VJ",
"R27fSFj2GDg",
"lhz7GgpoGNI",
"olDx-NNqHMM"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"**Post response update:**\nI would like to thank the authors for their response and revision addressing my concerns, at least to some extent. However, after careful consideration as well as looking at the other reviews (and responses), I am still not convinced the case can be made for increasing the score beyond m... | [
6,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4
] | [
4,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3
] | [
"iclr_2021_R43miizWtUN",
"iclr_2021_R43miizWtUN",
"KFJGy_biDPV",
"lhz7GgpoGNI",
"287aS9XMjl",
"5Oc5pZ8BljM",
"JxgoG7blyH-",
"olDx-NNqHMM",
"yn9_H4V90nq",
"iREFEsHL7VJ",
"iclr_2021_R43miizWtUN",
"iclr_2021_R43miizWtUN"
] |
iclr_2021_snaT4xewUfX | Variational inference for diffusion modulated Cox processes | This paper proposes a stochastic variational inference (SVI) method for computing an approximate posterior path measure of a Cox process. These processes are widely used in natural and physical sciences, engineering and operations research, and represent a non-trivial model of a wide array of phenomena. In our work, we... | withdrawn-rejected-submissions | The paper presents a stochastic variational inference method for posterior estimation in a Cox process with intensity given by the solution to a diffusion stochastic differential equation. The reviewers highlight the novelty of the approach. Some of the concerns with regards to clarity have been addressed by the author... | test | [
"ErYNOvdSXK0",
"6um-uRxyGxd",
"gushUQKdu19",
"xZHOtFZSujK",
"ObdL0M4CSQy",
"3S9O1xujXTc",
"keGz9unhkJE",
"BpJCfvKaThx",
"BL6joWsS59"
] | [
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"This should be good enough. Thanks for taking my remarks into account.\n\nNote: would it really make the presentation much more complex to include the estimation of $h$ and $b$ in the paper? ",
"We agree that often, the functions $b(\\cdot), \\sigma(\\cdot)$ and $h(\\cdot)$ are known only up to an unknown parame... | [
-1,
-1,
-1,
-1,
-1,
-1,
7,
7,
6
] | [
-1,
-1,
-1,
-1,
-1,
-1,
3,
2,
3
] | [
"6um-uRxyGxd",
"gushUQKdu19",
"xZHOtFZSujK",
"keGz9unhkJE",
"BpJCfvKaThx",
"BL6joWsS59",
"iclr_2021_snaT4xewUfX",
"iclr_2021_snaT4xewUfX",
"iclr_2021_snaT4xewUfX"
] |
iclr_2021_bfTUfrqL6d | Aspect-based Sentiment Classification via Reinforcement Learning | Aspect-based sentiment classification aims to predict sentimental polarities of one or multiple aspects in texts. As texts always contain a large proportion of task-irrelevant words, accurate alignment between aspects and their sentimental descriptions is the most crucial and challenging step. State-of-the-art approach... | withdrawn-rejected-submissions | In this paper, the authors proposed a reinforcement-learning-based model for aspect-based sentiment analysis. As raised by the reviewers, 1) the writing needs to be improved: e.g., presenting the details of the proposed method clearly, citing the references properly, etc. 2) related methods need to be implemented for c... | train | [
"aI1jBqZ_8w0",
"rHZhRxfIh0",
"BKWaQCOnAV",
"zDXzKasOQhC",
"eqk8D-1TlhJ",
"OHRCUKnCFGA"
] | [
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Thank you very much for reviewing and providing such valuable and frank comments on our work. The suggestions help us a lot in improving our paper. Below are our detailed replies:\n\n$\\bullet$ This paper failed to compare against a number of publications\n\nThanks a lot for your suggestions. We will definitely di... | [
-1,
-1,
-1,
5,
5,
3
] | [
-1,
-1,
-1,
4,
3,
5
] | [
"zDXzKasOQhC",
"eqk8D-1TlhJ",
"OHRCUKnCFGA",
"iclr_2021_bfTUfrqL6d",
"iclr_2021_bfTUfrqL6d",
"iclr_2021_bfTUfrqL6d"
] |
iclr_2021_JdCUjf9xvlc | Fourier Representations for Black-Box Optimization over Categorical Variables | Optimization of real-world black-box functions defined over purely categorical variables is an active area of research. In particular, optimization and design of biological sequences with specific functional or structural properties have a profound impact in medicine, materials science, and biotechnology. Standalone ac... | withdrawn-rejected-submissions | This paper considers the problem of black-box optimization over categorical variables using expensive function evaluations.
- Fourier representation is proposed as surrogate model by treating the categorical input as the direct sum of cyclic groups. The parameters are learned using exponentially-weighted update algori... | train | [
"aKPyautVyYX",
"Um-hbeKlCXq",
"noGMzQsyFJ9",
"51onIB2AFyU",
"f5F59yaRs0",
"X5G1KtYlCX",
"Kp_ISY2764h",
"QKXO-GvrY3L",
"AuqT4pEgdWr",
"JQ1CeqY5Vw"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"The paper proposes two representations, namely one-hot encoded Boolean expansion and group-theoretical Fourier expansion, for the surrogate model used for the black-box evaluations on purely categorical variables. With the two surrogate models, the authors tackle both the black-box optimization problem and the des... | [
6,
5,
-1,
-1,
-1,
-1,
-1,
-1,
6,
6
] | [
3,
3,
-1,
-1,
-1,
-1,
-1,
-1,
3,
1
] | [
"iclr_2021_JdCUjf9xvlc",
"iclr_2021_JdCUjf9xvlc",
"JQ1CeqY5Vw",
"AuqT4pEgdWr",
"aKPyautVyYX",
"Um-hbeKlCXq",
"iclr_2021_JdCUjf9xvlc",
"Um-hbeKlCXq",
"iclr_2021_JdCUjf9xvlc",
"iclr_2021_JdCUjf9xvlc"
] |
iclr_2021_bIwkmDnSeu | Unbiased Learning with State-Conditioned Rewards in Adversarial Imitation Learning | Adversarial imitation learning has emerged as a general and scalable framework for automatic reward acquisition. However, we point out that previous methods commonly exploited occupancy-dependent reward learning formulation—which hinders the reconstruction of optimal decision as an energy-based model. Despite the theor... | withdrawn-rejected-submissions | The paper was evaluated by 3 knowledgeable reviewers. All reviewers raised concerns about the motivation of the contribution of the paper. It is unclear why the use of an additional discriminator should reduce the variance of the log density ratio estimate. Also, the derivations were found to be not convincing or intui... | train | [
"AA_Q88D-RJj",
"UYqITumYs-",
"xDkZH387wqx",
"wJbHoPt0ST",
"ppz88Bgj-M5",
"vMvOOCAl0WA",
"j5W3xy7p-En",
"CQxVFW7eko8",
"7FrHmdTFgs",
"4IWr-zl98X",
"MvjbbHCujnE",
"ERAiWU5FQC_"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"__Additional Remark on transfer guarantees__\nI think that I do understand your setting of training a policy in a different environment. However, the paper claims several times that the learned _reward function_ transfers to different environments. To test whether a learned reward function transfers to a different... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
5
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
5,
4
] | [
"xDkZH387wqx",
"wJbHoPt0ST",
"wJbHoPt0ST",
"vMvOOCAl0WA",
"ERAiWU5FQC_",
"CQxVFW7eko8",
"iclr_2021_bIwkmDnSeu",
"4IWr-zl98X",
"MvjbbHCujnE",
"iclr_2021_bIwkmDnSeu",
"iclr_2021_bIwkmDnSeu",
"iclr_2021_bIwkmDnSeu"
] |
iclr_2021_CGFN_nV1ql | Non-Attentive Tacotron: Robust and controllable neural TTS synthesis including unsupervised duration modeling | This paper presents Non-Attentive Tacotron based on the Tacotron 2 text-to-speech model, replacing the attention mechanism with an explicit duration predictor. This improves robustness significantly as measured by unaligned duration ratio and word deletion rate, two metrics introduced in this paper for large-scale robu... | withdrawn-rejected-submissions | This paper investigates a non-attentive architecture of Tacotron 2 for TTS where the attention mechanism is replaced by a duration predictor. The authors show that this change can significantly improve the robustness. In addition, the authors propose two evaluation metrics for TTS robustness, namely, unaligned duratio... | train | [
"6U-H4OGfeiN",
"M_2CzDIIAHe",
"2DZIozDcDl-",
"fXTNGFDJCwu",
"EeUOZL0pfh",
"cUjROPFRw5s",
"WsBfYtL-HjZ",
"bjnPODjg5el",
"Wt4_drsmMH",
"ldocLkL41_V",
"aXZ6D5Furiw",
"sJceBIR9_Y3"
] | [
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"public",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Thank you for the great feedback. Below are the itemized responses regarding each comment. We have incorporated them into the revised version.\n\n**Re: Which attention mechanism was used between the input tokens and target mel-spectrograms in your work? Sun et al. 2020 and Lee & Kim 2019 use different mechanisms.*... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
8,
5,
6
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
3,
2
] | [
"bjnPODjg5el",
"ldocLkL41_V",
"aXZ6D5Furiw",
"EeUOZL0pfh",
"Wt4_drsmMH",
"sJceBIR9_Y3",
"iclr_2021_CGFN_nV1ql",
"Wt4_drsmMH",
"iclr_2021_CGFN_nV1ql",
"iclr_2021_CGFN_nV1ql",
"iclr_2021_CGFN_nV1ql",
"iclr_2021_CGFN_nV1ql"
] |
iclr_2021_cvNYovr16SB | Unsupervised Active Pre-Training for Reinforcement Learning | We introduce a new unsupervised pre-training method for reinforcement learning called APT, which stands for ActivePre-Training. APT learns a representation and a policy initialization by actively searching for novel states in reward-free environments. We use the contrastive learning framework for learning the represent... | withdrawn-rejected-submissions | The authors propose a particle-based entropy estimate for intrinsic motivation for pre-training an RL agent to then perform in an environment with rewards. As the reviewers discussed, and also mentioned in their reviews, this paper bears stark similarity to work of 5 months ago, presented at the ICML 2020 Lifelong ML w... | val | [
"2KKrTYFYHIb",
"5REgx8PElUB",
"wL39wnkiiV",
"RjSxBb9nEqH",
"hElXib5LpQQ",
"3qnsZEQGFVm",
"5HmyVCn_zwC",
"2hLjEszBGVh",
"iVLp-mK7bXp",
"IKCjrk4jHC"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"*SUMMARY*\n\nThe paper proposes a method to simultaneously learn effective representations and efficient exploration in a reward-free context. The algorithm iterates between minimizing a contrastive loss and maximizing an intrinsic reward derived from a k-NN entropy estimation of the state distribution. Then, auth... | [
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
5
] | [
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3
] | [
"iclr_2021_cvNYovr16SB",
"3qnsZEQGFVm",
"RjSxBb9nEqH",
"2KKrTYFYHIb",
"iVLp-mK7bXp",
"hElXib5LpQQ",
"2hLjEszBGVh",
"IKCjrk4jHC",
"iclr_2021_cvNYovr16SB",
"iclr_2021_cvNYovr16SB"
] |
iclr_2021_4sCyjwaVtZ9 | Whitening and second order optimization both destroy information about the dataset, and can make generalization impossible | Machine learning is predicated on the concept of generalization: a model achieving low error on a sufficiently large training set should also perform well on novel samples from the same distribution. We show that both data whitening and second order optimization can harm or entirely prevent generalization. In general, ... | withdrawn-rejected-submissions | The paper suggests that whitening the data harms generalization and optimization performance when learning models of the form h(W x) i.e. those that are based on a linear projection of the inputs (which includes DNNs for instance). The main concern of the reviewers is that their theoretical developments were not that c... | train | [
"-dwA5KqjxVC",
"L_-J4cCvyDu",
"OOSltbPUxFm",
"k5Jin9TRqHp",
"Xxb4YHzTBI5",
"Qgr3l14vW4p",
"b8W9MNtGeFw",
"y7iscJNnM0v",
"TLBd67TVS0",
"kzhkon3fu7X",
"uspevoEF4ed",
"v-rnSrEmXMI",
"IuKeTm6epmv",
"xj45JwN-KrO",
"a9FkSnIQvwE",
"g3j1UGnvCVw",
"wrtOXBmh1t7",
"pAITO92RrT",
"FwgXvV30lBO... | [
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"Topic: whitening destroys generalization\n\n\nmain contribution\n\nThis work offers a mutual information perspective to explain the relations between whitening/second-order optimization and generalization.\n\nStrength\n\n- The perspective is interesting and a bit intriguing. But many discrepancies may need clarifi... | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
7,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
7,
4
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
5
] | [
"iclr_2021_4sCyjwaVtZ9",
"y7iscJNnM0v",
"Xxb4YHzTBI5",
"Qgr3l14vW4p",
"a9FkSnIQvwE",
"uspevoEF4ed",
"IuKeTm6epmv",
"TLBd67TVS0",
"xj45JwN-KrO",
"iclr_2021_4sCyjwaVtZ9",
"-dwA5KqjxVC",
"kzhkon3fu7X",
"kzhkon3fu7X",
"pAITO92RrT",
"-dwA5KqjxVC",
"FwgXvV30lBO",
"FwgXvV30lBO",
"iclr_202... |
iclr_2021_oQyb8NrFzu | Revisiting the Stability of Stochastic Gradient Descent: A Tightness Analysis | The technique of algorithmic stability has been used to capture the generalization power of several learning models, especially those trained with stochastic gradient descent (SGD). This paper investigates the tightness of the algorithmic stability bounds for SGD given by~\cite{hardt2016train}. We show that the analysi... | withdrawn-rejected-submissions | The AC and reviewers agree this is an important line of research. However, only one reviewer was initially positive, as the other reviewers raised some issues, and the rebuttal only partially addressed some of them (e.g., the reviewer is now OK with Lemma 1 being correct), but there are typos in the proofs, and there w... | train | [
"b0OPmXmY-Kj",
"JD59_QNr_kf",
"J1UFX7vCrRQ",
"0Gmj3iUBRW",
"_6Y5BxFm0v",
"2_HUVIAWmw",
"EC4f5rUgRuf",
"Vfeu8UlUelP",
"lE4liCZv-b-",
"44hRd56y2PF"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"This paper considers the stability of the stochastic gradient decent algorithm under different conditions. They show a lower bound for the stability of SGD in the smooth and convex case, and show that the bound can be tightened for linear models. They give a tight bound for the stability of SGD with decreasing ste... | [
4,
5,
-1,
-1,
-1,
-1,
-1,
-1,
7,
4
] | [
4,
4,
-1,
-1,
-1,
-1,
-1,
-1,
3,
3
] | [
"iclr_2021_oQyb8NrFzu",
"iclr_2021_oQyb8NrFzu",
"_6Y5BxFm0v",
"44hRd56y2PF",
"Vfeu8UlUelP",
"JD59_QNr_kf",
"lE4liCZv-b-",
"b0OPmXmY-Kj",
"iclr_2021_oQyb8NrFzu",
"iclr_2021_oQyb8NrFzu"
] |
iclr_2021_5i4vRgoZauw | Wiring Up Vision: Minimizing Supervised Synaptic Updates Needed to Produce a Primate Ventral Stream | After training on large datasets, certain deep neural networks are surprisingly good models of the neural mechanisms of adult primate visual object recognition. Nevertheless, these models are poor models of the development of the visual system because they posit millions of sequential, precisely coordinated synaptic up... | withdrawn-rejected-submissions | This paper received 2 borderline accepts, 1 accept, and 1 reject.
This paper was discussed on the forum and no consensus was reached. The two reviewers who rated the paper as borderline accept emphasized that the biological claims are overblown, that the intellectual contributions (the initialization scheme and partia... | train | [
"x69s-QVkUXe",
"ZLa-QoRZoaj",
"FR9Pg100AJ_",
"mLOqe2iMKBY",
"LDDTX0DUsft",
"dZsRkolWip",
"XPIhGBJPtmr",
"_8B3rNPcTVp",
"ukxUIgf6KH5",
"beqW60hA6wi",
"CyRW9Pv0e6",
"nqpB8mapTJD",
"pk3Rx0YjuPx",
"6edoeUPGvl"
] | [
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Summary\n-------\nThe paper is about ANN being best-known models of developed primate visual systems. However this fact does not yet mean that the way those systems are trained is also similar. This distinction and a step towards answering this question is the main motivation of this work. The authors demonstrate ... | [
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
8,
3
] | [
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3,
4
] | [
"iclr_2021_5i4vRgoZauw",
"FR9Pg100AJ_",
"iclr_2021_5i4vRgoZauw",
"LDDTX0DUsft",
"CyRW9Pv0e6",
"nqpB8mapTJD",
"nqpB8mapTJD",
"pk3Rx0YjuPx",
"6edoeUPGvl",
"6edoeUPGvl",
"x69s-QVkUXe",
"iclr_2021_5i4vRgoZauw",
"iclr_2021_5i4vRgoZauw",
"iclr_2021_5i4vRgoZauw"
] |
iclr_2021_3EM0a2wC-jo | Learning Online Data Association | When an agent interacts with a complex environment, it receives a stream of percepts in which it may detect entities, such as objects or people. To build up a coherent, low-variance estimate of the underlying state, it is necessary to fuse information from multiple detections over time. To do this fusion, the agent mus... | withdrawn-rejected-submissions | This paper was on the borderline. While there was some support for the ideas presented, concerns were raised about the experiments. The exposition would also need to better demonstrate the significance of the contribution. | train | [
"YiuZXrKS_Al",
"NEvn8hLfRCp",
"SJtroGmYQuC",
"Pq6-0gPRS65",
"PifoXUg43m",
"veCuEhjqv9r",
"tdFKqput3Vy",
"n9msQjDTWe",
"GFWfNP-as4U",
"Ir6aFOqvRBC",
"V1GG4vq0PKW"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"This work aims to solve the data association problem in a general setting, which is suitable for various scenarios such as online clustering and online tracking. The motivation of having this setting is that temporal dense observations are not always available. \n\nThe high level idea of this work is to learn a ne... | [
6,
7,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
6
] | [
3,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
2
] | [
"iclr_2021_3EM0a2wC-jo",
"iclr_2021_3EM0a2wC-jo",
"tdFKqput3Vy",
"PifoXUg43m",
"Ir6aFOqvRBC",
"YiuZXrKS_Al",
"V1GG4vq0PKW",
"GFWfNP-as4U",
"NEvn8hLfRCp",
"iclr_2021_3EM0a2wC-jo",
"iclr_2021_3EM0a2wC-jo"
] |
iclr_2021_T4gXBOXoIUr | Contrastive Learning of Medical Visual Representations from Paired Images and Text | Learning visual representations of medical images is core to medical image understanding but its progress has been held back by the small size of hand-labeled datasets. Existing work commonly relies on transferring weights from ImageNet pretraining, which is suboptimal due to drastically different image characteristics... | withdrawn-rejected-submissions | The proposed ConVIRT learns representations of medical data from paired image and text data.
While the paper addresses a relevant problem, the reviewers agree that the method has limited novelty. Two reviewers find and that the experiments are not convincing. One reviewer notes that the paper does not compare to the st... | val | [
"triDpDQjme7",
"x_7VKroEJ4g",
"z78CXoitGhF",
"rtcs2Ykpbrs",
"QcO02LON8RQ",
"X-PCVhoqwqb",
"NQlsOFVSZfR",
"2THIW-CiMnP"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"##################### Summary ####################\n\nThis paper presents the Contrastive VIsual Representation Learning from Text (ConVIRT) pretraining strategy to learn fine-grained medical visual representations of medical images by pretraining on large-scale image-report pairs. As a result, ConVIRT improve... | [
6,
-1,
-1,
-1,
-1,
-1,
4,
5
] | [
4,
-1,
-1,
-1,
-1,
-1,
4,
5
] | [
"iclr_2021_T4gXBOXoIUr",
"iclr_2021_T4gXBOXoIUr",
"triDpDQjme7",
"2THIW-CiMnP",
"iclr_2021_T4gXBOXoIUr",
"NQlsOFVSZfR",
"iclr_2021_T4gXBOXoIUr",
"iclr_2021_T4gXBOXoIUr"
] |
iclr_2021_PsdsEbzxZWr | Analyzing and Improving Generative Adversarial Training for Generative Modeling and Out-of-Distribution Detection | Generative adversarial training (GAT) is a recently introduced adversarial defense method. Previous works have focused on empirical evaluations of its application to training robust predictive models. In this paper we focus on theoretical understanding of the GAT method and extending its application to generative model... | withdrawn-rejected-submissions | This paper conducts a theoretical and empirical analysis of the Generative Adversarial Training method (GAT). Although many comments have been addressed in the rebuttal, the reviewers still have few (but important) concerns, including the memorization effects and the lack of comparisons. | train | [
"iAkiqxK2V2x",
"V8qV1CmkgiR",
"n8ux7o5Bea",
"tgB0sY20Uy",
"9T2bgqhc2a9",
"1TI1-8WhuUp",
"VYYX2wy0nT",
"x7gMykgYXC",
"qKlTMRZcu0",
"UXdKOsFrG0z",
"WHbaIksuqlQ",
"tSGmwiv2jGU",
"O62PxY7-hqV"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"In this paper the authors provide a theoretical and empirical analysis of the Generative Adversarial Training method (GAT) which is used to train models for OOD and adversarial example detection.\n\nThe GAT method is analyzed from a game theoretical prespective, focusing on the differences with GAN training, which... | [
7,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5
] | [
3,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5
] | [
"iclr_2021_PsdsEbzxZWr",
"iclr_2021_PsdsEbzxZWr",
"WHbaIksuqlQ",
"iAkiqxK2V2x",
"iclr_2021_PsdsEbzxZWr",
"iclr_2021_PsdsEbzxZWr",
"V8qV1CmkgiR",
"qKlTMRZcu0",
"tSGmwiv2jGU",
"iclr_2021_PsdsEbzxZWr",
"VYYX2wy0nT",
"O62PxY7-hqV",
"iclr_2021_PsdsEbzxZWr"
] |
iclr_2021_taQNxF9Sj6 | Adding Recurrence to Pretrained Transformers | Fine-tuning a pretrained transformer for a downstream task has become a standard method in NLP in the last few years. While the results from these models are impressive, applying them can be extremely computationally expensive, as is pretraining new models with the latest architectures. We present a novel method for ap... | withdrawn-rejected-submissions | In this paper, the authors propose to add recurrence to pre-trained language models such as GPT-2 or BERT. The idea is similar to the compressive transformer paper: a small module is added to the network, and used to compress the representations from the previous chunk of data from the sequence to a single vector. Then... | train | [
"uzImIRbM8n",
"aGPcOc-V6-m",
"_3liUUFUsoO",
"Se9mZDWotL0",
"qQNEkzJ2OvR",
"9bPwIXKKFYE",
"Nd8-ViHCEzW"
] | [
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Thank you to each of the reviewers for the helpful comments. We have revised the paper, making some points more clear, and also added two new experiments to address specific concerns which were raised.\n\nThe more substantial experiment is on the HotpotQA task, although due to time constraints we give preliminary ... | [
-1,
-1,
-1,
-1,
4,
7,
7
] | [
-1,
-1,
-1,
-1,
4,
5,
2
] | [
"iclr_2021_taQNxF9Sj6",
"qQNEkzJ2OvR",
"9bPwIXKKFYE",
"Nd8-ViHCEzW",
"iclr_2021_taQNxF9Sj6",
"iclr_2021_taQNxF9Sj6",
"iclr_2021_taQNxF9Sj6"
] |
iclr_2021_wqRvVvMbJAT | One Size Doesn't Fit All: Adaptive Label Smoothing | This paper concerns the use of objectness measures to improve the calibration performance of Convolutional Neural Networks (CNNs). CNNs have proven to be very good classifiers and generally localize objects well; however, the loss functions typically used to train classification CNNs do not penalize inability to locali... | withdrawn-rejected-submissions | The paper introduces an adaptive label smoothing technique, where the smoothing factor is computed based on the relative object size within an image, in order to address the problem of overconfident predictions. All reviewers recommend rejection based on limited technical contribution and unclear benefits of the propos... | train | [
"wSwnm2j-31w",
"0Yh3a-x0E3R",
"IuHv9tagLw_",
"pMeYFuI-Fkn",
"MgCt4-Jaf9",
"YueA2p3Op_-",
"3nmT5zeih22",
"5LX7I3egJsw",
"LWxMSfHJnH",
"KZoX8ReurAJ"
] | [
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"**Summary**\nMany modern classifiers are trained with class labels - it's natural because you train the model with precisely the target label you want the model to produce! But let's think out of the box and introduce stronger forms of supervision - e.g. bounding boxes for the object of interest. What benefits wil... | [
4,
4,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4
] | [
4,
3,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4
] | [
"iclr_2021_wqRvVvMbJAT",
"iclr_2021_wqRvVvMbJAT",
"pMeYFuI-Fkn",
"MgCt4-Jaf9",
"wSwnm2j-31w",
"LWxMSfHJnH",
"0Yh3a-x0E3R",
"KZoX8ReurAJ",
"iclr_2021_wqRvVvMbJAT",
"iclr_2021_wqRvVvMbJAT"
] |
iclr_2021_3ZeGLibhFo0 | Enabling counterfactual survival analysis with balanced representations | Balanced representation learning methods have been applied successfully to counterfactual
inference from observational data. However, approaches that account for
survival outcomes are relatively limited. Survival data are frequently encountered
across diverse medical applications, i.e., drug developme... | withdrawn-rejected-submissions | Summary: This paper provides an approach for causal inference in observational survival dataset in which the outcome is of time-to-event type with right-censored samples. To this end, the paper adapts the balanced representation learning approach proposed in (Shalit et al, 2017) to the context of survival analysis.
Th... | train | [
"NG7yV071o9",
"v3wYBvhfQHG",
"iExrYgLY_Vg",
"F79D2tt-yFq",
"Z42pvO5pWR7",
"T5vIQuizHHa",
"LmeeU1Xvdw",
"gSC5ARWW7QH",
"lJk5ROlM9I2",
"dAyVns7P2pW",
"ncuqnOmWwsy"
] | [
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"We want to thank the reviewer for engaging during the discussion period.\n\n**Counterfactual survival analysis**\n\nWe want to clarify that counterfactual inference in the context of survival analysis refers to **time-to-event** predictions of potential outcomes $(T_1, T_0)$, conditioned on the covariates $X$.\nAs... | [
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
7,
7,
5
] | [
-1,
5,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
4
] | [
"iExrYgLY_Vg",
"iclr_2021_3ZeGLibhFo0",
"F79D2tt-yFq",
"v3wYBvhfQHG",
"dAyVns7P2pW",
"lJk5ROlM9I2",
"ncuqnOmWwsy",
"iclr_2021_3ZeGLibhFo0",
"iclr_2021_3ZeGLibhFo0",
"iclr_2021_3ZeGLibhFo0",
"iclr_2021_3ZeGLibhFo0"
] |
iclr_2021_qbRv1k2AcH | Learning to Reason in Large Theories without Imitation | In this paper, we demonstrate how to do automated higher-order logic theorem proving in the presence of a large knowledge base of potential premises without learning from human proofs. We augment the exploration of premises based on a simple tf-idf (term frequency-inverse document frequency) based lookup in a deep rein... | withdrawn-rejected-submissions | *Overview* This paper applies RL to automated theorem proving to eliminate the need for human-written proofs as training data. The method uses TF-IDF for premise selections. The experiments compared with supervised baseline demonstrate some good performance.
*Pro* The paper provides a side-by-side comparison of the ef... | train | [
"D0lFP4taUoe",
"z_HPQCkHWv7",
"CP2WNOOrTCo",
"cUFUjHLKEA5",
"6p7s2_GwldV",
"WgzDxIFi5FE",
"xjtTj7f8H_1",
"X6vsIcV_eaS",
"l0K9BG265pr",
"MIAu6KfrWzT",
"meiy75PJbl6",
"iOgQXw0O0Oo",
"ECP8GklE_VH",
"o6lN4Y9YxTZ",
"NO0N7RgM7yC",
"8cGgVX7O_bx",
"PFRWfrLoYTD",
"n4WD6iv-6hr",
"F8LeEDxob... | [
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"public",
"author",
"author",
"author",
"public",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Summary:\n\nThe authors apply reinforcement learning to automated theorem proving to eliminate the need for human-written proofs as training data. During training, the prover is improved incrementally by using the current version to prove theorems and add proved theorems to the training data. The authors propose t... | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
6,
6,
6
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
4,
4,
3
] | [
"iclr_2021_qbRv1k2AcH",
"cUFUjHLKEA5",
"z_HPQCkHWv7",
"o6lN4Y9YxTZ",
"xjtTj7f8H_1",
"xjtTj7f8H_1",
"X6vsIcV_eaS",
"8cGgVX7O_bx",
"PFRWfrLoYTD",
"F8LeEDxob4_",
"ECP8GklE_VH",
"n4WD6iv-6hr",
"NO0N7RgM7yC",
"D0lFP4taUoe",
"o6lN4Y9YxTZ",
"iclr_2021_qbRv1k2AcH",
"iclr_2021_qbRv1k2AcH",
... |
iclr_2021_FoM-RnF6SNe | Evaluating Agents Without Rewards | Reinforcement learning has enabled agents to solve challenging control tasks from raw image inputs. However, manually crafting reward functions can be time consuming, expensive, and prone to human error. Competing objectives have been proposed for agents to learn without external supervision, such as artificial input e... | withdrawn-rejected-submissions | The reviewers agree that the paper, in its current form, is not strong enough to allow for publication. There are specific weaknesses that need to be tackled: a better correlation study; a clearer relationship to existing literature (and improvement on the novelty); clearer, more precise use of descriptions.
The auth... | train | [
"uOZc2Z57ry3",
"jnSyUINxpW7",
"eYE3jv7MVh5",
"_gbO5H-neYX",
"izsjrXHPwJ7",
"_NFCZ5LOBFI",
"fyrskBE7ku0",
"wv5SWmQ0kt5",
"__S18xWcfnX",
"38AT_4erR8c",
"JhvpBBRpD-m",
"wxH0-OWfJN2",
"EYQs2jNgT7a",
"dJeMKJEcAs7",
"F93qR5b0tE"
] | [
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Thank you for your detailed response.\n\n> **Choice of environment**: Thank you for adding the Minecraft environment to the list. My point was to add environments that do not express any task, just pure open-ended environments.\n\nWe find that correlations hold when only the four task-agnostic agents are incorpora... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
4,
3
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
4,
5,
4
] | [
"_NFCZ5LOBFI",
"_gbO5H-neYX",
"izsjrXHPwJ7",
"38AT_4erR8c",
"__S18xWcfnX",
"wv5SWmQ0kt5",
"F93qR5b0tE",
"EYQs2jNgT7a",
"wxH0-OWfJN2",
"dJeMKJEcAs7",
"iclr_2021_FoM-RnF6SNe",
"iclr_2021_FoM-RnF6SNe",
"iclr_2021_FoM-RnF6SNe",
"iclr_2021_FoM-RnF6SNe",
"iclr_2021_FoM-RnF6SNe"
] |
iclr_2021_Dw8vAUKYq8C | Near-Optimal Glimpse Sequences for Training Hard Attention Neural Networks | Hard visual attention is a promising approach to reduce the computational burden of modern computer vision methodologies. Hard attention mechanisms are typically non-differentiable. They can be trained with reinforcement learning but the high-variance training this entails hinders more widespread application. We show h... | withdrawn-rejected-submissions | This paper proposes a new and unusual way of training hard attention mechanisms in vision models. Instead of training with reinforcement learning (as is typical), the authors develop a procedure for generating "glimpse sequences" that can be effectively used as supervision. Models trained in this way produce qualitativ... | train | [
"nk-Oyzg1PUR",
"wy7b1UZE5Lq",
"-zbq_TrTIzI",
"1LQZFuN0XwM",
"NJkYEGluQcE",
"CxYVpWx5RI",
"SIFfkWrhQRz",
"c001hcpEa3v",
"YF99DaFL96N",
"V2KwevymYwi",
"ahMzACSERU-",
"B2JvGXVCa6R",
"cOB9tgKuVBr",
"gCjDe_ApEKg"
] | [
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"This paper introduces a way to annotate a glimpse sequence for an image. It uses BAYESIAN OPTIMAL EXPERIMENTAL DESIGN to achieve this. Using the obtained annotations, hard attention can be trained with a partially supervised way.\n\nMy concerns mainly focus on the motivation and evaluation. Specifically, the autho... | [
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
6,
7
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
2,
2
] | [
"iclr_2021_Dw8vAUKYq8C",
"-zbq_TrTIzI",
"YF99DaFL96N",
"c001hcpEa3v",
"SIFfkWrhQRz",
"iclr_2021_Dw8vAUKYq8C",
"1LQZFuN0XwM",
"B2JvGXVCa6R",
"nk-Oyzg1PUR",
"cOB9tgKuVBr",
"gCjDe_ApEKg",
"iclr_2021_Dw8vAUKYq8C",
"iclr_2021_Dw8vAUKYq8C",
"iclr_2021_Dw8vAUKYq8C"
] |
iclr_2021_wE-3ly4eT5G | FactoredRL: Leveraging Factored Graphs for Deep Reinforcement Learning | We propose a simple class of deep reinforcement learning (RL) methods, called FactoredRL, that can leverage factored environment structures to improve the sample efficiency of existing model-based and model-free RL algorithms. In tabular and linear approximation settings, the factored Markov decision process literature... | withdrawn-rejected-submissions | The paper introduces variants of RL algorithms that can consume factored state representations. Under the assumption that actions only affect a few factors, these factored RL algorithms can learn more efficiently than their vanilla counterparts. Learning a factored dynamics model (to be used in a model-based algorithm)... | train | [
"yEkHtl6UnVQ",
"RbOSxB7WnKx",
"js62h1-SEv5",
"WvHpEFIorze",
"_pJ7wsuWCn",
"9_KAvTvdkOu",
"JJOZcciQ7X5",
"pY1uEuu8CvH",
"2D376K7tnmu",
"MCBt-_TKDly"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"This paper looks at how to build deep RL agents that can perform more efficient learning by directly leveraging the factored structure of problems when that information is given. The approach proposed by the authors is to design a neural network architecture that explicitly honors the factored structure of the pro... | [
6,
5,
-1,
-1,
-1,
-1,
-1,
-1,
6,
6
] | [
3,
4,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3
] | [
"iclr_2021_wE-3ly4eT5G",
"iclr_2021_wE-3ly4eT5G",
"RbOSxB7WnKx",
"yEkHtl6UnVQ",
"iclr_2021_wE-3ly4eT5G",
"iclr_2021_wE-3ly4eT5G",
"MCBt-_TKDly",
"2D376K7tnmu",
"iclr_2021_wE-3ly4eT5G",
"iclr_2021_wE-3ly4eT5G"
] |
iclr_2021_B9nDuDeanHK | Weights Having Stable Signs Are Important: Finding Primary Subnetworks and Kernels to Compress Binary Weight Networks | Binary Weight Networks (BWNs) have significantly lower computational and memory costs compared to their full-precision counterparts. To address the non-differentiable issue of BWNs, existing methods usually use the Straight-Through-Estimator (STE). In the optimization, they learn optimal binary weight outputs represent... | withdrawn-rejected-submissions | ## Description
The paper discovers interesting phenomena in training neural networks with binary weights:
- Connection between latent weight magnitude and how important its binarized version for the network performance
-training dynamics, indicating that large latent weights are identified and stabilize early on
- Obse... | test | [
"d2BGrqHnphy",
"t4Gd-Z3BGzq",
"Hh3YhqXcfrW",
"U7Ow_rI-R5_",
"irze5KX48qn",
"agWICurksvF",
"l14Jc43Yd7Y",
"-fN9j3q6zk",
"f-r9ifrwtW3",
"xeN_H7z0Gc",
"F5V1E6gJH6V"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"Overview:\nThe Authors show that scaling factors with hand-crafted or learnable methods are not so important when training Binary Weight Networks (BWNs), while the change of weight signs is crucial. They make two observations: The weight signs of the primary binary sub-networks are determined and fixed at the earl... | [
6,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
5
] | [
4,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
2
] | [
"iclr_2021_B9nDuDeanHK",
"iclr_2021_B9nDuDeanHK",
"iclr_2021_B9nDuDeanHK",
"xeN_H7z0Gc",
"xeN_H7z0Gc",
"d2BGrqHnphy",
"F5V1E6gJH6V",
"t4Gd-Z3BGzq",
"iclr_2021_B9nDuDeanHK",
"iclr_2021_B9nDuDeanHK",
"iclr_2021_B9nDuDeanHK"
] |
iclr_2021_Xa3iM4C1nqd | Transferable Unsupervised Robust Representation Learning | Robustness is an important, and yet, under-explored aspect of unsupervised representation learning, which has seen a lot of recent developments. In this work, we address this gap by developing a novel framework: Unsupervised Robust Representation Learning (URRL), which combines unsupervised representation learning's pr... | withdrawn-rejected-submissions | The authors propose a framework which combines pretext tasks and data augmentation schemes with the goal of improving robustness of image representations. The authors show that this approach empirically can lead to increased accuracy on both corrupted and uncorrupted data simultaneously. Furthermore, the authors propos... | train | [
"cV6iNfZge1v",
"ysTPLuahTjB",
"-60iXO0XUXM",
"kKJ0ly0c7Jm",
"F9gh6QRlXI",
"_C2ThqeRjQh",
"n4OmXhvV9A",
"1Col7qP0h_m",
"AW3BWF4eEr",
"Ic1zysvllwF",
"aSBHV0_Thik"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"author",
"author"
] | [
"The authors improve the fine-tuning of a data representation allowing\nresistance and recognition of adversarial or corrupt inputs. Their method,\nURRL, uses an auxiliary unsupervised task and robust data augementation to\nimprove performance for both clean and altered inputs. A key addition is that,\nwhile fine... | [
7,
7,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
4,
5,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
"iclr_2021_Xa3iM4C1nqd",
"iclr_2021_Xa3iM4C1nqd",
"iclr_2021_Xa3iM4C1nqd",
"F9gh6QRlXI",
"Ic1zysvllwF",
"1Col7qP0h_m",
"iclr_2021_Xa3iM4C1nqd",
"AW3BWF4eEr",
"ysTPLuahTjB",
"-60iXO0XUXM",
"cV6iNfZge1v"
] |
iclr_2021_nlWgE3A-iS | ReaPER: Improving Sample Efficiency in Model-Based Latent Imagination | Deep Reinforcement Learning (DRL) can distill behavioural policies from sensory input that solve complex tasks, however, the policies tend to be task-specific and sample inefficient, requiring a large number of interactions with the environment that may be costly or impractical for many real world applications. Model-b... | withdrawn-rejected-submissions | This paper explores losses and other training details to produce a model-based agent for pixel-input continuous control problems. The authors present a rainbow-like approach that combines various separate innovations into a single system. They show an improvement over a previous baseline on this class of problem, and... | train | [
"aBwelCYryav",
"oNvR6cNkeDP",
"WBtwyppY0fI",
"d-0wWMZOnhO",
"SDX_gEuuOya",
"uU4h5wuHIHp",
"weSWHbrRV4z"
] | [
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"We appreciate the constructive feedback from the reviewer; we address their main concerns below as well as in the revised paper.\n\nOn Appendices 6.2 and 6.3: The methods were chosen for evaluation since they were promising ideas in the existing literature that warranted further study in the context of sample effi... | [
-1,
-1,
-1,
4,
5,
4,
6
] | [
-1,
-1,
-1,
3,
5,
3,
4
] | [
"d-0wWMZOnhO",
"SDX_gEuuOya",
"uU4h5wuHIHp",
"iclr_2021_nlWgE3A-iS",
"iclr_2021_nlWgE3A-iS",
"iclr_2021_nlWgE3A-iS",
"iclr_2021_nlWgE3A-iS"
] |
iclr_2021_axNDkxU9-6z | MDP Playground: Controlling Orthogonal Dimensions of Hardness in Toy Environments | We present MDP Playground, an efficient benchmark for Reinforcement Learning (RL) algorithms with various dimensions of hardness that can be controlled independently to challenge algorithms in different ways and to obtain varying degrees of hardness in generated environments. We consider and allow control over a wide v... | withdrawn-rejected-submissions | I thank the authors for their submission and very active participation in the author response period. The reviewers and I acknowledge the importance of designing toy environments that allow the community to systematically investigate strengths and weaknesses of RL approaches. That said, the reviewers have criticized th... | train | [
"40aRW9_FEe8",
"nhRR80pptcd",
"jk5mOioRFqD",
"-nP0iERFIOp",
"BC5iiPyHfW6",
"523OsiVNFdM",
"I2aPqoii-KX",
"GplNcqLLc6r",
"mgXe0VjsyDS",
"uAMKKHih3jV",
"OV-e4Ekw8j",
"217Cc4velS-",
"jJ4wf0Vcmhz",
"CsYglxpypdt",
"5JO1ZGOGTE4",
"6-n8dbEQG2x",
"5wII6BnPxq",
"zM_Vc2t-a1-",
"sa-YEL1_qup... | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"... | [
"--------------------------------------\n\nPOST-REBUTTAL COMMENTS\n\nAs a result of the discussion the paper has improved, so I'm increasing my score. However, the core issue, that the proposed benchmarks don't seem to capture the difficulty structure of either real problems or more complex benchmarks, remains.\n\n... | [
5,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4
] | [
4,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5
] | [
"iclr_2021_axNDkxU9-6z",
"iclr_2021_axNDkxU9-6z",
"iclr_2021_axNDkxU9-6z",
"I2aPqoii-KX",
"GplNcqLLc6r",
"mgXe0VjsyDS",
"_PberakT51R",
"217Cc4velS-",
"217Cc4velS-",
"CsYglxpypdt",
"6-n8dbEQG2x",
"sa-YEL1_qup",
"zM_Vc2t-a1-",
"sa-YEL1_qup",
"NgZMKd8QbR",
"EjbRZ18qOZ0",
"iclr_2021_axND... |
iclr_2021_L-88RyVtXGr | Learning Deeply Shared Filter Bases for Efficient ConvNets | Recently, inspired by repetitive block structure of modern ConvNets, such as ResNets, parameter-sharing among repetitive convolution layers has been proposed to reduce the size of parameters. However, naive sharing of convolution filters poses many challenges such as overfitting and vanishing/exploding gradients, res... | withdrawn-rejected-submissions | I agree with the concerns raised by the reviewers. In particular, the issues of novelty and experimental evaluation (mentioned in the revision summary) remain the major weak points of the paper. My impression is that the changes made in the revision represent a significant experimental addition to the paper, one which ... | test | [
"GYkhKpJaErQ",
"Ww6rPvj_zsp",
"JUPq1REID8G",
"HCNJ86FgmFY",
"BUX5eM9QvbB",
"SsOwnzl9tUi",
"xBx_zYJflq5",
"1EL4Vpq1Y0q",
"awBMMQLPaCE",
"Y8nBhfZ0FgP"
] | [
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"**Q:** \n(1) The experiments mainly compared the proposed method with non-shared net designs. However, the comparison results or methods/algorithms with other weight sharing paper are lacking. (e.g., Jastrzebski et al.,2018; Köpüklü et al., 2019 mentioned in the paper). Therefore, it's not easy to judge the novelt... | [
-1,
-1,
-1,
-1,
-1,
-1,
5,
5,
6,
4
] | [
-1,
-1,
-1,
-1,
-1,
-1,
3,
4,
3,
4
] | [
"1EL4Vpq1Y0q",
"xBx_zYJflq5",
"awBMMQLPaCE",
"Y8nBhfZ0FgP",
"Y8nBhfZ0FgP",
"iclr_2021_L-88RyVtXGr",
"iclr_2021_L-88RyVtXGr",
"iclr_2021_L-88RyVtXGr",
"iclr_2021_L-88RyVtXGr",
"iclr_2021_L-88RyVtXGr"
] |
iclr_2021_4emQEegFhSy | Adaptive Multi-model Fusion Learning for Sparse-Reward Reinforcement Learning | In this paper, we consider intrinsic reward generation for sparse-reward reinforcement learning based on model prediction errors. In typical model-prediction-error-based intrinsic reward generation, an agent has a learning model for the underlying environment. Then intrinsic reward is designed as the error between the ... | withdrawn-rejected-submissions | The paper extends previous work on intrinsic reward design based on curiosity or surprise toward multiple intrinsic rewards based multiple model predictions and fuse the reward using meta-gradient optimization. While most reviewers find the paper clearly written, several reviewers do bring up the concern on limited co... | train | [
"R4rxPzX52-V",
"0Kx-Oj7D8pP",
"Uw9GZTFgWnZ",
"pAQyileyOO0",
"4u5LC_bwYpk",
"X2kOPRLTxj6",
"VK15NzCGeXy",
"pUYU16Rr-p",
"yTyzf5Fs9a4",
"Us-HLzJkj6c",
"5DxdHbKnlHD",
"7Wjoxp1yOfU"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Summary\nThe papers looks at the problem of using intrinsic rewards to help agent explore in sparse reward settings. The paper proposes combining multiple intrinsic rewards and proposes a meta-gradient based method to learning the fusion of these intrinsic rewards.\n\n\nStrengths\n1. Learning in a sparse rewards s... | [
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
7,
-1,
5,
6
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
-1,
3,
3
] | [
"iclr_2021_4emQEegFhSy",
"5DxdHbKnlHD",
"R4rxPzX52-V",
"yTyzf5Fs9a4",
"iclr_2021_4emQEegFhSy",
"7Wjoxp1yOfU",
"iclr_2021_4emQEegFhSy",
"iclr_2021_4emQEegFhSy",
"iclr_2021_4emQEegFhSy",
"pAQyileyOO0",
"iclr_2021_4emQEegFhSy",
"iclr_2021_4emQEegFhSy"
] |
iclr_2021_w_BtePbtmx4 | Accelerating DNN Training through Selective Localized Learning | Training Deep Neural Networks (DNNs) places immense compute requirements on the underlying hardware platforms, expending large amounts of time and energy. We proposeLoCal+SGD, a new algorithmic approach to accelerate DNN train-ing by selectively combining localized or Hebbian learning within a StochasticGradient Descen... | withdrawn-rejected-submissions | In this paper, the authors proposed a new approach by the name of LoCal + SGD (Localized Updates) to replace the traditional Backpropagation method. The key idea is to selectively update some layers’ weights using localized learning rules, so as to reduce the computational complexity of training these layers so as to a... | train | [
"ZaXIt7ZDPhl",
"xd5iQX7dhT",
"RUqcSUX3_yk",
"UCL9azocUz2",
"vspWjV_NsDW",
"eFf6KAQstUq",
"SVflQWKVKiF",
"rjD_VHlsjoz",
"Qk5T8hfQNfX",
"6YizFfZAPun",
"NY4HU8Sy-tZ",
"xYbpldQCHMC"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"This paper try to leverage the benefit of Hebb learning to reduce CNN training time cost. In order to achieve this, a learning mode selection algorithm is proposed to progressively increase number of layers using Hebb learning. The writing of this paper is good and the idea is also interesting, however, the expe... | [
6,
7,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
4,
6
] | [
4,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
4
] | [
"iclr_2021_w_BtePbtmx4",
"iclr_2021_w_BtePbtmx4",
"Qk5T8hfQNfX",
"xd5iQX7dhT",
"6YizFfZAPun",
"xYbpldQCHMC",
"ZaXIt7ZDPhl",
"NY4HU8Sy-tZ",
"eFf6KAQstUq",
"iclr_2021_w_BtePbtmx4",
"iclr_2021_w_BtePbtmx4",
"iclr_2021_w_BtePbtmx4"
] |
iclr_2021_0Zxk3ynq7jE | An Empirical Exploration of Open-Set Recognition via Lightweight Statistical Pipelines | Machine-learned safety-critical systems need to be self-aware and reliably know their unknowns in the open-world. This is often explored through the lens of anomaly/outlier detection or out-of-distribution modeling. One popular formulation is that of open-set classification, where an image classifier trained for 1-of-K... | withdrawn-rejected-submissions | The addresses open-set recognition, namely, detecting anomalous samples that belong to classes not observed during training.
It has been shown that existing methods fail on open-world images. The current paper shows empirically that performance can be greatly improved if based on low-dimensional features.
Reviewers ... | val | [
"jVXQb4TxgaN",
"MzbifBfBwQK",
"Y9s6YaOVW_o",
"yioNUfRjuDg",
"Qk_dwHzosEv",
"VaGuRP-R0kR",
"hGo92z2m7QD",
"FtJVUjKewGi"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"This paper proposes an open-set recognition approach that uses simple statistical measures (such as GMM’s and KMeans) on top of post-processed intermediate features extracted from closed-set deep models. It finds that i) this “lightweight” pipeline outperforms prior methods on open-set image recognition across mul... | [
7,
-1,
-1,
-1,
-1,
3,
3,
4
] | [
4,
-1,
-1,
-1,
-1,
5,
3,
4
] | [
"iclr_2021_0Zxk3ynq7jE",
"jVXQb4TxgaN",
"VaGuRP-R0kR",
"hGo92z2m7QD",
"FtJVUjKewGi",
"iclr_2021_0Zxk3ynq7jE",
"iclr_2021_0Zxk3ynq7jE",
"iclr_2021_0Zxk3ynq7jE"
] |
iclr_2021_AMoDLAx6GCC | Uncertainty Prediction for Deep Sequential Regression Using Meta Models | Generating high quality uncertainty estimates for sequential regression, particularly deep recurrent networks, remains a challenging and open problem.
Existing approaches often make restrictive assumptions (such as stationarity) yet still perform poorly in practice, particularly in presence of real world non-stat... | withdrawn-rejected-submissions | This is a borderline paper. Initially, it received weaker reviews (than the current ones) and after rebuttal, one reviewer slightly increased the review rating. Among 4 reviews, there is one clearly positive one.
Among negative reviews, there are repeated concerns about evaluation. While recognizing the difficulty o... | train | [
"nTy-ImkUITk",
"uIeQv8A4dHO",
"rm1bVAYFuPs",
"x07MLtZyDbg"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"In this paper, the authors propose a technique for uncertainty estimation in regression with neural networks. The basic idea is to use an auxiliary \"meta model\" that (in the authors' best performing setting) has access to the base model and is trained jointly with it. The purpose of the meta model is to predict ... | [
6,
7,
5,
5
] | [
4,
3,
3,
4
] | [
"iclr_2021_AMoDLAx6GCC",
"iclr_2021_AMoDLAx6GCC",
"iclr_2021_AMoDLAx6GCC",
"iclr_2021_AMoDLAx6GCC"
] |
iclr_2021_IfEkus1dpU | Cut-and-Paste Neural Rendering | Cut-and-paste methods take an object from one image and insert it into another. Doing so often results in unrealistic looking images because the inserted object's shading is inconsistent with the target scene's shading. Existing reshading methods require a geometric and physical model of the inserted object, which is... | withdrawn-rejected-submissions | Overall the review is borderline: R2 and R4 are slightly positive and R3 is slightly negative. All the reviewers like the novel shading consistency loss proposed in the paper and, improved DIP that produces consistent image decomposition inferences, and good experimental results. However, reviewers also shared concerns... | train | [
"uYEUBUknKRY",
"paXCVRKxuex",
"ntbO5jRYCmM",
"Aq3rLC7lXI",
"MwYNvtLyOJB",
"AjTiVcsEck"
] | [
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"Summary\nThis paper introduces a reshading method for cut-and-paste image composition. It uses a modified deep image prior as the rendering networking, and trains with a novel shading consistency loss. The results are plausible. Inserted fragments have proper shading with the source image and unchanged albedo.\n\n... | [
6,
-1,
-1,
-1,
5,
6
] | [
2,
-1,
-1,
-1,
4,
4
] | [
"iclr_2021_IfEkus1dpU",
"AjTiVcsEck",
"uYEUBUknKRY",
"MwYNvtLyOJB",
"iclr_2021_IfEkus1dpU",
"iclr_2021_IfEkus1dpU"
] |
iclr_2021_FKotzp6PZJw | On the Estimation Bias in Double Q-Learning | Double Q-learning is a classical method for reducing overestimation bias, which is caused by taking maximum estimated values in the Bellman operator. Its variants in the deep Q-learning paradigm have shown great promise in producing reliable value prediction and improving learning performance. However, as shown by prio... | withdrawn-rejected-submissions | This paper is rejected.
I and the reviewers appreciate the changes made by the authors. The paper presents:
* An analysis (based on techniques from previous work) of double Q-learning which shows that in an analytic model, double Q-learning can have multiple sub-optimal "approximated" fixed points.
* Propose a modific... | train | [
"emBgDSCMqia",
"xuMhIYSNhYX",
"L3orggSMMlb",
"Nz_FwY7Fne9",
"NclccXkh3mO",
"Hgs85k0lk5x",
"DzgsIW-waV",
"KRK02TduV7a",
"eLEoM-JUrI3",
"R13h2nrv5Y-",
"qnEpfqqvRcb",
"07u6jF0r1bB"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"public",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"This paper analysed the underestimation bias induced by approximation error, by formalizing the underlying approximation, they theoretically proved the existence of multiple approximated fixed points which causes the converging to non-optimal solution. Besides, they proposed the lower bound double q-learning to ov... | [
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
3,
6
] | [
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
3
] | [
"iclr_2021_FKotzp6PZJw",
"R13h2nrv5Y-",
"eLEoM-JUrI3",
"NclccXkh3mO",
"emBgDSCMqia",
"qnEpfqqvRcb",
"Hgs85k0lk5x",
"07u6jF0r1bB",
"iclr_2021_FKotzp6PZJw",
"iclr_2021_FKotzp6PZJw",
"iclr_2021_FKotzp6PZJw",
"iclr_2021_FKotzp6PZJw"
] |
iclr_2021_0qbEq5UBfGD | Latent Space Semi-Supervised Time Series Data Clustering | Time series data is abundantly available in the real world, but there is a distinct lack of large, labeled datasets available for many types of learning tasks. Semi-supervised models, which can leverage small amounts of expert-labeled data along with a larger unlabeled dataset, have been shown to improve performance ov... | withdrawn-rejected-submissions | This paper addresses an important problem of semi-supervised learning of time-series data. Their approach is based on a convolution autoencoder for learning a time-series latent space. To guide learning an appropriate embedding, they explore three alternative internal clustering metrics (prototype loss, Silhouette lo... | train | [
"-jtD34iXp9r",
"dLrJYLAaIrc",
"j_N840cf9bn",
"2gaoJnkOW73",
"C9XFegpubmq",
"nfKqtmgALF",
"dm5jdr3tlOU",
"PMEYgqXAIEK",
"PW6vdbzUuiC"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Disclaimer: I am not an expert in the time-series domain, although I did some literature review while performing this review.\n\n#####################################\nSummary:\nThe work investigates semi-supervised (SS) clustering of time-series data (i.e., clustering with few labelled points, which can also be s... | [
4,
-1,
-1,
-1,
-1,
-1,
6,
5,
4
] | [
3,
-1,
-1,
-1,
-1,
-1,
3,
2,
5
] | [
"iclr_2021_0qbEq5UBfGD",
"j_N840cf9bn",
"-jtD34iXp9r",
"dm5jdr3tlOU",
"PMEYgqXAIEK",
"PW6vdbzUuiC",
"iclr_2021_0qbEq5UBfGD",
"iclr_2021_0qbEq5UBfGD",
"iclr_2021_0qbEq5UBfGD"
] |
iclr_2021_6VPl9khIMz | Adaptive Stacked Graph Filter | We study Graph Convolutional Networks (GCN) from the graph signal processing viewpoint by addressing a difference between learning graph filters with fully-connected weights versus trainable polynomial coefficients. We find that by stacking graph filters with learnable polynomial parameters, we can build a highly adapt... | withdrawn-rejected-submissions | The topic covered by the paper is timely, and the way the authors have addressed the problem seems correct. The provided empirical evidence seems to be sufficient to support the main claim of the paper. Presentation is well structured and clear.
Notwithstanding the above merits, the proposed approach seems to confirm o... | train | [
"ffPTbQW8Yyk",
"k6-ag9JXOPO",
"IMxMpGPCVVD",
"LGQDKIrJMd",
"04kSepVOBw9",
"uAoPqUpOWZd",
"t3lqGZQML8V",
"_eOSuichm5i",
"CTtKJBPBgk",
"F8oAnsp4A1d",
"Y5Oh8AwuGAp",
"C7aiOcbl1aE",
"nwpZ5x1N70Z",
"SLyGfHR8vJ",
"SMtn1JswFjv",
"VXgr7O_yV5p",
"V8RPke0cqDi"
] | [
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Adaptive stacked graph filter\nThe paper proposes a relatively simple formulation for a graph convolutional filter, that has the advantage of providing useful insights on the characteristic of the considered datasets. Many points of the paper are however not convincing in the present form, mainly regarding the nov... | [
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
5,
5
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
5,
5
] | [
"iclr_2021_6VPl9khIMz",
"IMxMpGPCVVD",
"LGQDKIrJMd",
"Y5Oh8AwuGAp",
"_eOSuichm5i",
"F8oAnsp4A1d",
"CTtKJBPBgk",
"C7aiOcbl1aE",
"SLyGfHR8vJ",
"nwpZ5x1N70Z",
"SMtn1JswFjv",
"VXgr7O_yV5p",
"ffPTbQW8Yyk",
"V8RPke0cqDi",
"iclr_2021_6VPl9khIMz",
"iclr_2021_6VPl9khIMz",
"iclr_2021_6VPl9khIM... |
iclr_2021_TCAmP8zKZ6k | Towards a Reliable and Robust Dialogue System for Medical Automatic Diagnosis | Dialogue system for medical automatic diagnosis (DSMAD) aims to learn an agent that mimics the behavior of a human doctor, i.e. inquiring symptoms and informing diseases. Since DSMAD has been formulated as a Markov decision-making process, many studies apply reinforcement learning methods to solve it. Unfortunately, ex... | withdrawn-rejected-submissions | The authors propose a novel approach to a dialog-based automated medical diagnosis, and present promising empirical results. The focus of this work is on robustness and reliability besides just the accuracy of diagnosis, which appears to be an important aspect in medical applications. The paper is clearly written and ... | train | [
"lT9iMi39ERw",
"mLbOYJdTIHm",
"GSzoAb3l8yk",
"NrxpIzEAAKX",
"TlfIBbReia1",
"Np0-NxD_I4d",
"SCxQNdf67Rw",
"WTtDpzZzO9O",
"734BN_BtTwW"
] | [
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Thank you for your constructive comments. We have adjusted our manuscript accordingly.\n\nAccording to the comments from all reviewers, we concluded the three major concerns/interests from the reviewers.\n\n1) __Adding more experiment details__\n\n2) __Elaborating on the analysis of the experimental results__\n\n3... | [
-1,
-1,
-1,
-1,
-1,
6,
4,
6,
6
] | [
-1,
-1,
-1,
-1,
-1,
4,
4,
4,
4
] | [
"iclr_2021_TCAmP8zKZ6k",
"WTtDpzZzO9O",
"SCxQNdf67Rw",
"734BN_BtTwW",
"Np0-NxD_I4d",
"iclr_2021_TCAmP8zKZ6k",
"iclr_2021_TCAmP8zKZ6k",
"iclr_2021_TCAmP8zKZ6k",
"iclr_2021_TCAmP8zKZ6k"
] |
iclr_2021_J5LS3YJH7Zi | CaLFADS: latent factor analysis of dynamical systems in calcium imaging data | Dynamic latent variable modelling has been a hugely powerful tool in understanding how spiking activity in populations of neurons can perform computations necessary for adaptive behaviour. The success of such approaches has been enabled by the ability to construct models derived with the characterization of spiking act... | withdrawn-rejected-submissions | Inferring latent trajectory from noisy Ca time series is an important and timely problem and the current study shows some progress in the inference problem. Although the proposed model has some originality, there are remaining issues rendering the manuscript not ready for publication yet. Reviewers raised issues on rea... | test | [
"QczkeV0Tu0M",
"pCfJ7xISc8",
"1c2x6RaAdo6",
"XoSscj2joHl",
"_nTWpzx6YOM",
"_s4yJ0QaaC",
"4IkfsC8-Vln",
"vTUo2vYk7-j",
"aUDuOQ0Demv",
"RaPgQP9YLX-",
"a-C6Q-w2f0y",
"aiXnD6t5TjI",
"Ck7S0urUIRi",
"wgXoFsbXXPB",
"QMAz1mINAOK",
"Mu2fey6BaIi",
"1-VRJIbzFHQ",
"y-MCRg_sn7O",
"xbOXz9XER4"... | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author"
] | [
"This submission extends a previous nonlinear data analysis method ( Latent Factor Analysis of Dynamical Systems, LFADS) to deal with the calcium imaging data (CaLFADS). \n\nQuality: Developing appropriate methods to handle the calcium imaging data is an important question as calcium imaging becomes popular. This s... | [
5,
4,
7,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
4,
4,
4,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
"iclr_2021_J5LS3YJH7Zi",
"iclr_2021_J5LS3YJH7Zi",
"iclr_2021_J5LS3YJH7Zi",
"iclr_2021_J5LS3YJH7Zi",
"4IkfsC8-Vln",
"Mu2fey6BaIi",
"RaPgQP9YLX-",
"iclr_2021_J5LS3YJH7Zi",
"a-C6Q-w2f0y",
"aiXnD6t5TjI",
"QMAz1mINAOK",
"wgXoFsbXXPB",
"iclr_2021_J5LS3YJH7Zi",
"XoSscj2joHl",
"pCfJ7xISc8",
"Q... |
iclr_2021_qFQTP00Q0kp | Self-Supervised Time Series Representation Learning by Inter-Intra Relational Reasoning | Self-supervised learning achieves superior performance in many domains by extracting useful representations from the unlabeled data. However, most of traditional self-supervised methods mainly focus on exploring the inter-sample structure while less efforts have been concentrated on the underlying intra-temporal struct... | withdrawn-rejected-submissions | This paper presents a general self-supervised time series representation learning framework. The organization is good, and the architecture is well motivated. However, the paper has limited novelty, and is a straightforward application of ideas in self-supervised learning literature.
Experimental results are not entir... | val | [
"T1_UQN6H_c",
"U4IqNAxpJPQ",
"J7mJRTAHZVZ",
"JlKCqIxMFT",
"7FgvfBN3FCe",
"fyqTqWYZan7",
"27fhjSytv5Z",
"jtkoXsiIWSw",
"poKzl5rgbVa",
"yq0U2WHh4xg",
"PBQWxbacltX",
"vkfzkieLSx",
"nvHEN0Jh_NS",
"uh7sToYENZ4",
"1h5WsmO-7m4"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"public"
] | [
"===========Update after rebuttal==================\n\nI will remain my score of weak rejection.\n\n\n=============================================================\n\nThis paper presents to model both inter-sample and intra-temporal relations. The idea is naive but is easy to follow and reasonable. Experimental res... | [
5,
6,
-1,
-1,
-1,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
4,
3,
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
"iclr_2021_qFQTP00Q0kp",
"iclr_2021_qFQTP00Q0kp",
"27fhjSytv5Z",
"T1_UQN6H_c",
"T1_UQN6H_c",
"iclr_2021_qFQTP00Q0kp",
"vkfzkieLSx",
"iclr_2021_qFQTP00Q0kp",
"1h5WsmO-7m4",
"U4IqNAxpJPQ",
"fyqTqWYZan7",
"nvHEN0Jh_NS",
"PBQWxbacltX",
"yq0U2WHh4xg",
"iclr_2021_qFQTP00Q0kp"
] |
iclr_2021_D4A-v0kltaX | Neural Partial Differential Equations with Functional Convolution | We present a lightweighted neural PDE representation to discover the hidden structure and predict the solution of different nonlinear PDEs. Our key idea is to leverage the prior of ``"translational similarity" of numerical PDE differential operators to drastically reduce the scale of learning model and training data. W... | withdrawn-rejected-submissions | The objective of the paper is to develop a framework for solving PDES with reduced model size and for scarce observation settings. It proposes to use functional input dependent convolutions for learning spatio-temporal differential operators together with a non linear numerical scheme (Picard solver). Training makes us... | train | [
"FhVjYOZu-bp",
"tYxwypDbtUc",
"dYuAViybQ5t",
"8YhAYkXu_s2",
"Pg4YKtydUma",
"rHen4nc4x6"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author"
] | [
"Summary:\nThe paper proposes a neural network based solver for PDEs based on the Picard Iteration. In the numerical experiments section the paper applies the method to solve 1d or 2d PDEs.\n\nDisclaimer:\nI am not an expert in PDE solvers but I am researching topics related to your frequently cited works of Kipf e... | [
4,
5,
4,
4,
-1,
-1
] | [
3,
4,
2,
4,
-1,
-1
] | [
"iclr_2021_D4A-v0kltaX",
"iclr_2021_D4A-v0kltaX",
"iclr_2021_D4A-v0kltaX",
"iclr_2021_D4A-v0kltaX",
"iclr_2021_D4A-v0kltaX",
"iclr_2021_D4A-v0kltaX"
] |
iclr_2021_GSTrduvZSjT | Adaptive Gradient Methods Converge Faster with Over-Parameterization (and you can do a line-search) | Adaptive gradient methods are typically used for training over-parameterized models capable of exactly fitting the data; we thus study their convergence in this interpolation setting. Under an interpolation assumption, we prove that AMSGrad with a constant step-size and momentum can converge to the minimizer at the fas... | withdrawn-rejected-submissions | Dear authors,
the paper contains many interesting and novel ideas. Indeed, tuning step-size is very time and energy-consuming, and deriving and analyzing new adaptive algorithms has not only theoretical benefits but, more importantly, is a key when training more complicated ML models.
The paper contains many weakness... | val | [
"UIi7TeuUY6T",
"5Xns0rKrA8h",
"MlZRHZHQre",
"ka-KZZOhvcq",
"QDBQ3DMCo1h",
"f4zD2Ybmfo",
"OMyFCyULIM7",
"_XnSpF03qvC",
"k_jJ76AaV_B",
"bawcqX8l03"
] | [
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"##########################################################################\n\n\nSummary:\n\nThis paper studies the convergence of adaptive gradient methods under an interpolation assumption, showing for example these methods can converge at an O(1/t), instead of O(1/\\sqrt{t}) rate when perfect interpolation is sa... | [
7,
5,
-1,
5,
-1,
-1,
-1,
-1,
-1,
6
] | [
4,
4,
-1,
4,
-1,
-1,
-1,
-1,
-1,
4
] | [
"iclr_2021_GSTrduvZSjT",
"iclr_2021_GSTrduvZSjT",
"iclr_2021_GSTrduvZSjT",
"iclr_2021_GSTrduvZSjT",
"ka-KZZOhvcq",
"bawcqX8l03",
"5Xns0rKrA8h",
"ka-KZZOhvcq",
"UIi7TeuUY6T",
"iclr_2021_GSTrduvZSjT"
] |
iclr_2021_L4v_5Qtshj7 | Goal-Driven Imitation Learning from Observation by Inferring Goal Proximity | Humans can effectively learn to estimate how close they are to completing a desired task simply by watching others fulfill the task. To solve the task, they can then take actions towards states with higher estimated proximity to the goal. From this intuition, we propose a simple yet effective method for imitation learn... | withdrawn-rejected-submissions | The paper proposes a method for learning intristic reward from demonstrations. The inartistic reward is computed as time-to-reach and generalizes to unseen states. The reviewers agree that the method is novel useful, and of interest to ICLR community.
Although the authors' significantly improved the manuscript durin... | train | [
"IsW3Rh015v9",
"nK7--QR2Maj",
"AQ86I-AzmFO",
"YNMdQd1Eity",
"Gy0x_pmkawx",
"gG27xMG2f7u",
"uTkFxkF3Zr",
"Z1F0P8_Zd4H",
"-wk_IU522kO",
"OQq3rQrpmgG",
"LmJVjPG4AVj",
"Nsdv-v3VN5Q",
"9tpAL1JEtf-",
"XHZI_tWxB2",
"Bhvm85BAR6A",
"-6ouwWmatVB",
"tfLxTWvMxR",
"GR8HlwemOmZ",
"r-RN0RFl5_9"... | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_re... | [
"Summary\n--------\nThe paper proposes a method for imitation learning for goal-directed tasks that uses a learned proximity function for computing rewards.\nAn ensemble of proximity functions is trained in supervised way to predict the time step (rescaled to the range $[0,1]$) for the expert's states. The ensemble... | [
5,
7,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5
] | [
5,
4,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3
] | [
"iclr_2021_L4v_5Qtshj7",
"iclr_2021_L4v_5Qtshj7",
"iclr_2021_L4v_5Qtshj7",
"iclr_2021_L4v_5Qtshj7",
"GR8HlwemOmZ",
"uTkFxkF3Zr",
"Z1F0P8_Zd4H",
"-wk_IU522kO",
"XHZI_tWxB2",
"Nsdv-v3VN5Q",
"iclr_2021_L4v_5Qtshj7",
"9tpAL1JEtf-",
"Bhvm85BAR6A",
"nK7--QR2Maj",
"LmJVjPG4AVj",
"r-RN0RFl5_9"... |
iclr_2021_lH2ukHnGDdq | Data augmentation for deep learning based accelerated MRI reconstruction | Deep neural networks have emerged as very successful tools for image restoration and reconstruction tasks. These networks are often trained end-to-end to directly reconstruct an image from a noisy or corrupted measurement of that image. To achieve state-of-the-art performance, training on large and diverse sets of imag... | withdrawn-rejected-submissions | Reviewers could not reach consensus here and legitimate concerns are raised on novelty and on empirical results, although this can be attributed to the important computation times required to run experiments on 3D MRI volumes. The authors have provided a comprehensive response to the reviews, the general feedback is th... | test | [
"ZXeur66yH-",
"9xXRRKkwzFs",
"FItQA-lZQ33",
"1LUEzn13ACB",
"J8CoccKolVd",
"h2R-_MGtfHe",
"c0mhJRaJrB",
"y4tgjlrFUb",
"lWMUt4y5nsf",
"y28TzRN9x3Q",
"Vvkqncil3K_",
"9gVTGNqc5yS",
"zkY1023eHHO",
"ubdjsbBBq0u"
] | [
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Example 2: \n \n\nReview: This paper proposes data augmentation methods for medical imaging(especially for accelerated MRI) based on the MR physics. The augmentation includes both pixel preserving augmentations/general affine augmentations on both real and imaginary values in the image domain. Then, the augmented ... | [
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
5,
6,
6
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
5,
4,
4
] | [
"iclr_2021_lH2ukHnGDdq",
"FItQA-lZQ33",
"lWMUt4y5nsf",
"J8CoccKolVd",
"Vvkqncil3K_",
"9gVTGNqc5yS",
"zkY1023eHHO",
"lWMUt4y5nsf",
"ZXeur66yH-",
"ubdjsbBBq0u",
"iclr_2021_lH2ukHnGDdq",
"iclr_2021_lH2ukHnGDdq",
"iclr_2021_lH2ukHnGDdq",
"iclr_2021_lH2ukHnGDdq"
] |
iclr_2021_cQzf26aA3vM | Design-Bench: Benchmarks for Data-Driven Offline Model-Based Optimization | Black-box model-based optimization (MBO) problems, where the goal is to find a design input that maximizes an unknown objective function, are ubiquitous in a wide range of domains, such as the design of drugs, aircraft, and robot morphology. Typically, such problems are solved by actively querying the black-box objecti... | withdrawn-rejected-submissions | This paper proposes a benchmark suite of offline model-based optimization problems. This benchmark includes diverse and realistic tasks derived from real-world problems in biology, material science, and robotics contains a wide variety of domains, and it covers both continuous and discrete, low and high dimensional des... | train | [
"7rrdJAu4Bdn",
"Zl1UlEAtvcC",
"UZBtph1m20J",
"TFn_HGbBbgo",
"KKjdu9sTZcm",
"4c0IUzJZ4zg",
"_Cs3nvyGs5J",
"pDotXIQLccs",
"OMqbMVsHdm2",
"vJaILsyDdNT",
"L8OXvLcnIp0",
"lhI1Z4vS6O-",
"gYZowACjvct",
"oda1uz3iBx",
"WY8shfJ0i24",
"fpMFldCMmRn",
"6AMi-FWucRf",
"LB07yCpLJfa",
"todhQSW7DU... | [
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"Summary: \nThis paper focuses on model-based black box optimization problems in the offline setting. These are settings where access to ground truth is expensive, and instead the optimizer has access only to a trained model of the ground truth based on limited data. While optimizing on this surrogate space, a good... | [
5,
-1,
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
7,
5
] | [
4,
-1,
-1,
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
2
] | [
"iclr_2021_cQzf26aA3vM",
"TFn_HGbBbgo",
"pDotXIQLccs",
"WY8shfJ0i24",
"PWRnhaLgG4C",
"7rrdJAu4Bdn",
"iclr_2021_cQzf26aA3vM",
"OMqbMVsHdm2",
"vJaILsyDdNT",
"oda1uz3iBx",
"_Cs3nvyGs5J",
"7rrdJAu4Bdn",
"PWRnhaLgG4C",
"_Cs3nvyGs5J",
"fpMFldCMmRn",
"7rrdJAu4Bdn",
"todhQSW7DU1",
"PWRnhaL... |
iclr_2021_1Kxxduqpd3E | Rotograd: Dynamic Gradient Homogenization for Multitask Learning | GradNorm (Chen et al., 2018) is a broadly used gradient-based approach for training multitask networks, where different tasks share, and thus compete during learning, for the network parameters. GradNorm eases the fitting of all individual tasks by dynamically equalizing the contribution of each task to the overall gra... | withdrawn-rejected-submissions | The paper is proposing a novel representation of the GradNorm. GradNorm is presented as a Stackelberg game and its theory is used to understand and improve the convergence of the GradNorm. Moreover, in addition to the magnitude normalization, a direction normalization objective is added to the leader and a rotation mat... | train | [
"ZqxUamCQxpR",
"Q8wnabSju2z",
"9hbSopefdZt",
"ICZTcYck4nV",
"FLsccGCB-1N",
"IYvAkj55pwL",
"l6DdqWZLzLf",
"UGj7g5UdbCr",
"8xR_2doyjng"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"In the paper, Rotograd is proposed as a new gradient-based approach for training multi-task deep neural networks based on GradNorm. GradNorm is first formulated as a Stackelberg game, where the leader aims at normalizing the gradient of different tasks and the follower aims at optimizing the collective weighted lo... | [
4,
4,
-1,
-1,
-1,
-1,
-1,
-1,
4
] | [
4,
4,
-1,
-1,
-1,
-1,
-1,
-1,
4
] | [
"iclr_2021_1Kxxduqpd3E",
"iclr_2021_1Kxxduqpd3E",
"l6DdqWZLzLf",
"Q8wnabSju2z",
"8xR_2doyjng",
"ZqxUamCQxpR",
"UGj7g5UdbCr",
"iclr_2021_1Kxxduqpd3E",
"iclr_2021_1Kxxduqpd3E"
] |
iclr_2021_CHTHamtufWN | Continual Invariant Risk Minimization | Empirical risk minimization can lead to poor generalization behaviour on unseen environments if the learned model does not capture invariant feature represen- tations. Invariant risk minimization (IRM) is a recent proposal for discovering environment-invariant representations. It was introduced by Arjovsky et al. (2019... | withdrawn-rejected-submissions | The authors address the problem of learning environment-invariant representations in the case where environments are observed sequentially.
This is done by using a variational Bayesian and bilevel framework.
The paper is borderline, with two reviewers (R2 and R3) favoring slightly acceptance and two reviewrs (R4 and R... | test | [
"r5tbGSrYuSw",
"kqaX38Tdp39",
"Nq7iopGhwdZ",
"WB5nCBAlvS7",
"q2FLHNN4IMn",
"XsYDPQ7Mxm",
"Q3q4GlAJKQ-",
"WykHjqi_9D",
"RbuyYKZAzq",
"HkkguB4sWuO",
"LbI_9J3oxOI",
"SxP-kymqRVQ",
"i0udmEtSTB",
"ZeHHueYDMK",
"OhARvWKdaw",
"EfNvGph3hyX",
"Ef8gE5-xPiW"
] | [
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"# Post-discussion update \n\nThe authors have significantly updated the paper during the discussion period. I have seen the changes, but they unfortunately do not substantiate the claim that the proposed methods are learning anything meaningful. \n\nIn the new results in table 1, the train accuracy now is better t... | [
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
6,
6
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
4,
2
] | [
"iclr_2021_CHTHamtufWN",
"Nq7iopGhwdZ",
"Q3q4GlAJKQ-",
"iclr_2021_CHTHamtufWN",
"Ef8gE5-xPiW",
"EfNvGph3hyX",
"RbuyYKZAzq",
"HkkguB4sWuO",
"LbI_9J3oxOI",
"ZeHHueYDMK",
"OhARvWKdaw",
"EfNvGph3hyX",
"Ef8gE5-xPiW",
"r5tbGSrYuSw",
"iclr_2021_CHTHamtufWN",
"iclr_2021_CHTHamtufWN",
"iclr_2... |
iclr_2021_JyDnXkeJpjU | Task-similarity Aware Meta-learning through Nonparametric Kernel Regression | This paper investigates the use of nonparametric kernel-regression to obtain a task- similarity aware meta-learning algorithm. Our hypothesis is that the use of task- similarity helps meta-learning when the available tasks are limited and may contain outlier/ dissimilar tasks. While existing meta-learning approaches im... | withdrawn-rejected-submissions | This paper addresses a method that incorporates the task-similarity (via task gradients) into the meta-learning. The inner loop update is done by kernel regression with the similarity between gradients of tasks considered, and the outer loop is the gradient update with a particular regularization. Without any doubt, it... | train | [
"mxnqg70QLyV",
"r_4zK9mOoV7",
"uykEiVTo0nV",
"AMTtrhGEeIl",
"-TGlYXb7AK3",
"s9Y6MxazSDG",
"dkQ2zIaS00h",
"D_I0_i7WbcD",
"EJv4um4rlZv"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"---- Update ----\n\nI thank the authors for clarifications. I trust that the suggestions of all reviewers, taken together, provide substantial avenues for improving the work. However, at this point I must keep my score and encourage the authors to continue the work with the valuable honest feedback provided here.\... | [
3,
-1,
-1,
-1,
-1,
-1,
4,
4,
4
] | [
5,
-1,
-1,
-1,
-1,
-1,
4,
4,
3
] | [
"iclr_2021_JyDnXkeJpjU",
"AMTtrhGEeIl",
"mxnqg70QLyV",
"dkQ2zIaS00h",
"D_I0_i7WbcD",
"EJv4um4rlZv",
"iclr_2021_JyDnXkeJpjU",
"iclr_2021_JyDnXkeJpjU",
"iclr_2021_JyDnXkeJpjU"
] |
iclr_2021_bd66LuDPPFh | Towards Understanding Label Smoothing | Label smoothing regularization (LSR) has a great success in training deep neural networks by stochastic algorithms such as stochastic gradient descent and its variants. However, the theoretical understanding of its power from the view of optimization is still rare. This study opens the door to a deep understanding of ... | withdrawn-rejected-submissions | The paper considers ways to understand label smoothing methods, which are widely used in many applications. There is some theory on the performance of SGD with and without the methods of the paper, but there is s significant gap in terms of how the theory offers insight into label smoothing. There are some empirical ... | train | [
"wPPet3n6Q4U",
"Hkf5MNMqNg",
"W_w8QfwOjDD",
"CAymXgoBYRy",
"X5TrPx_aqpG",
"8CHeQWm9m-x",
"mUsk2pg36LR",
"PFwyjTTPMUO",
"_PYStEv-jOD",
"l-8ZjStGwe",
"LeRAfjx3S-2"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author"
] | [
"[strong points]\n1. This paper is smooth and well-motivated. Label smoothing is a well-known trick to improve the multi-class neural network. This is the first work to theoretically understand the effect of label smoothing by analyzing the convergence speed.\n\n2. Based on some reasonable assumption, this paper pr... | [
4,
1,
6,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
3,
5,
3,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
"iclr_2021_bd66LuDPPFh",
"iclr_2021_bd66LuDPPFh",
"iclr_2021_bd66LuDPPFh",
"iclr_2021_bd66LuDPPFh",
"8CHeQWm9m-x",
"PFwyjTTPMUO",
"iclr_2021_bd66LuDPPFh",
"CAymXgoBYRy",
"Hkf5MNMqNg",
"W_w8QfwOjDD",
"wPPet3n6Q4U"
] |
iclr_2021_RrSuwzJfMQN | TOWARDS NATURAL ROBUSTNESS AGAINST ADVERSARIAL EXAMPLES | Recent studies have shown that deep neural networks are vulnerable to adversarial examples, but most of the methods proposed to defense adversarial examples cannot solve this problem fundamentally. In this paper, we theoretically prove that there is an upper bound for neural networks with identity mappings to constrain... | withdrawn-rejected-submissions | I thank the authors and reviewers for the discussions. Reviewers raised major concerns regarding the significance of the results and experiments. Given all, I think the paper needs more work before being accepted. I encourage authors to address comments raised by the reviewers to improve their paper.
- AC | val | [
"6odP2itSmB5",
"YjZ2xnF9BbR",
"jNFkq5yet27",
"XL2Q3y0j6Cq",
"3ErT9EHcuaR",
"ZANkEmjvJpf",
"YccuMVHxZp_",
"3eWYZAEljs",
"EyDkVrjxCaV",
"yHBcPTwNtGO",
"6j7unwbVUK-"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"**Update**: Thank you to the authors for addressing the comments and updating the paper. I decrease my rating from 4 to 3 as the original claims of the paper were disproved by the experiments with black-box attacks on CIFAR10 which showed that Neural ODEs offer little advantage over Resnets. I believe that more ex... | [
3,
5,
-1,
-1,
-1,
-1,
-1,
-1,
5,
3,
3
] | [
5,
4,
-1,
-1,
-1,
-1,
-1,
-1,
4,
5,
4
] | [
"iclr_2021_RrSuwzJfMQN",
"iclr_2021_RrSuwzJfMQN",
"YjZ2xnF9BbR",
"6odP2itSmB5",
"yHBcPTwNtGO",
"iclr_2021_RrSuwzJfMQN",
"EyDkVrjxCaV",
"6j7unwbVUK-",
"iclr_2021_RrSuwzJfMQN",
"iclr_2021_RrSuwzJfMQN",
"iclr_2021_RrSuwzJfMQN"
] |
iclr_2021_OodqmQT3fir | XLVIN: eXecuted Latent Value Iteration Nets | Value Iteration Networks (VINs) have emerged as a popular method to perform implicit planning within deep reinforcement learning, enabling performance improvements on tasks requiring long-range reasoning and understanding of environment dynamics. This came with several limitations, however: the model is not explicitly ... | withdrawn-rejected-submissions | The work extends the line of work based on value iteration networks. The main goal is to extend VINS to continuous and partially observable state spaces. The approach combines self-supervised contrastive learning and graph representation learning with VINs to address these issues. Reviewers liked the premise of the pap... | train | [
"_JgiyPWZ5m7",
"ue4y_qT8SzN",
"hftQ5Vl1Y7Q",
"7BeMlTQml4-",
"AwiUgcIplR",
"bU21SI0_42",
"iEfpwdrRchz",
"hNaKuPG_o7i",
"ZKp4IO1NBPW",
"Pgf3gOXkvrt",
"thH8oVlcywI",
"0aL7gHULQe",
"2wDmBS9DoUZ",
"dZt_PovnE4Q",
"x7u-Ylgw9V7",
"ALFd5exH0Tt",
"2LRGzVQTi_x",
"ACNtmeSDiea"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"**Summary**\n\nThe authors propose a generalization of Value Iteration Networks to unknown, potentially continuous state spaces. They describe a framework for leveraging a learned graph embedding model (TransE) in combination with a deep RL model and an execution model based on graphical message passing to perform... | [
6,
-1,
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
7,
6
] | [
3,
-1,
-1,
-1,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
2
] | [
"iclr_2021_OodqmQT3fir",
"iclr_2021_OodqmQT3fir",
"hNaKuPG_o7i",
"iEfpwdrRchz",
"bU21SI0_42",
"dZt_PovnE4Q",
"iclr_2021_OodqmQT3fir",
"7BeMlTQml4-",
"2LRGzVQTi_x",
"ACNtmeSDiea",
"0aL7gHULQe",
"2wDmBS9DoUZ",
"_JgiyPWZ5m7",
"2LRGzVQTi_x",
"iEfpwdrRchz",
"ACNtmeSDiea",
"iclr_2021_Oodqm... |
iclr_2021_Hw2Za4N5hy0 | Federated Learning with Decoupled Probabilistic-Weighted Gradient Aggregation | In the federated learning paradigm, multiple mobile clients train local models independently based on datasets generated by edge devices, and the server aggregates parameters/gradients from local models to form a global model. However, existing model aggregation approaches suffer from high bias on both data distribu... | withdrawn-rejected-submissions | The authors’ feedback has not fully addressed the reviewers’ concerns and the reviewers think that the paper is not ready for the publication. The authors should consider the following issues for the future submission:
1) The concern from Reviewer 1: if a local device receives very little data but its data come from a... | train | [
"EkQwXEMWLpQ",
"TMJDYUOoP7U",
"sPEOWJ789pe",
"ZQpLbw5Qlp",
"5xLOhNWVev1",
"1Sqn3S000yu",
"-fD37iZrWTX",
"gtgd22v3j1",
"YsCPK2SAvIW",
"5UndtHcTAac",
"TgHqjnG2DX"
] | [
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"We thank all reviewers for their time and detailed comments about this paper. Most of the comments are helpful to improve this paper. Based on the problems we focus, we first address shared concerns and then respond to specific comments for every reviewer. And we have updated the paper with edits highlighted in re... | [
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
6,
3,
4
] | [
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
4,
3,
4
] | [
"iclr_2021_Hw2Za4N5hy0",
"ZQpLbw5Qlp",
"iclr_2021_Hw2Za4N5hy0",
"-fD37iZrWTX",
"5UndtHcTAac",
"TgHqjnG2DX",
"sPEOWJ789pe",
"YsCPK2SAvIW",
"iclr_2021_Hw2Za4N5hy0",
"iclr_2021_Hw2Za4N5hy0",
"iclr_2021_Hw2Za4N5hy0"
] |
iclr_2021_67q9f8gChCF | Learning Efficient Planning-based Rewards for Imitation Learning | Imitation learning from limited demonstrations is challenging. Most inverse reinforcement learning (IRL) methods are unable to perform as good as the demonstrator, especially in a high-dimensional environment, e.g, the Atari domain. To address this challenge, we propose a novel reward learning method, which streamlines... | withdrawn-rejected-submissions | The authors introduce vPERL, a model that generates an intrinsic reward for imitation learning. vPERL is trained on demonstrations to minimise a variational objective that matches a posterior formed by "action backtracking" and a forward model, with the intrinsic reward coming from the reward map. The authors might be ... | train | [
"cdHV8ibZuy",
"OXKjo6y8msG",
"g5EvgN_N2-B",
"CQXzp_B63sT",
"gGfFVtBcRma",
"oI6k0XJ_KNn",
"3wiiXIFqiM",
"2s3C1RloX2t",
"RI3bA8rYNik",
"lXOBFeMkZQZ",
"wdlLyoaGROH"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"This paper proposes a method for inverse reinforcement learning that incorporates a differential planning module. Explicit transition dynamics modeling with inverse value iteration is added to promote meaningful reward learning. Empirical evaluations on several high-dimensional Atari environments and 2 continuous ... | [
6,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
6
] | [
3,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
2
] | [
"iclr_2021_67q9f8gChCF",
"iclr_2021_67q9f8gChCF",
"OXKjo6y8msG",
"RI3bA8rYNik",
"wdlLyoaGROH",
"cdHV8ibZuy",
"2s3C1RloX2t",
"lXOBFeMkZQZ",
"g5EvgN_N2-B",
"iclr_2021_67q9f8gChCF",
"iclr_2021_67q9f8gChCF"
] |
iclr_2021_paE8yL0aKHo | CAT-SAC: Soft Actor-Critic with Curiosity-Aware Entropy Temperature | The trade-off between exploration and exploitation has long been a crucial issue in reinforcement learning~(RL). Most of the existing RL methods handle this problem by adding action noise to the policies, such as the Soft Actor-Critic (SAC) that introduces an entropy temperature for maximizing both the external value a... | withdrawn-rejected-submissions | The concept of increasing entropy in novel states to promote exploration appears to be quite interesting. I do appreciate this idea, and I would encourage the authors to study it further. I think the reviewers also agree with this. Unfortunately, the paper as written has a number of issues: (1) the theoretical motivati... | train | [
"khvXYitj567",
"PCPEKf6huE",
"GzrO0Idq1k1",
"u-CZOydNzVe",
"pdhRtCvkcgN",
"PmPFuUJdX79",
"GsXgdH1lGXI",
"-VtAhx5F6vk",
"r4wNW3w6VD7",
"OfFjowdEWnx"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Summary: \n\nThis work proposes to incorporate curiosity into the entropy temperature of Soft Actor-Critic, and applies a modified version of Random Network Distillation as their curiosity model. Its key insight is that the entropy temperature in Soft Actor-Critic should encourage the agent to explore unfamiliar s... | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
6,
4,
4
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
3,
3,
5
] | [
"iclr_2021_paE8yL0aKHo",
"OfFjowdEWnx",
"khvXYitj567",
"-VtAhx5F6vk",
"iclr_2021_paE8yL0aKHo",
"khvXYitj567",
"r4wNW3w6VD7",
"iclr_2021_paE8yL0aKHo",
"iclr_2021_paE8yL0aKHo",
"iclr_2021_paE8yL0aKHo"
] |
iclr_2021_nLYMajjctMh | Federated Learning of a Mixture of Global and Local Models | We propose a new optimization formulation for training federated learning models. The standard formulation has the form of an empirical risk minimization problem constructed to find a single global model trained from the private data stored across all participating devices. In contrast, our formulation seeks an explici... | withdrawn-rejected-submissions | The paper studies an elegant formulation of personalized federated learning, which balances between a global model and locally trained models. It then analyzes algorithm variants inspired by local update SGD in this setting. The problem formulation using the explicit trade-off between model differences and global objec... | train | [
"R4cRgYjXq3",
"LLbSRB6-AIo",
"n56zYRqKsT2",
"hVJ5fKkCWXe",
"faPNn6EWFJ6",
"MaQc5wy8Mg7",
"yKGgV2fqiaH",
"1nt_anxUt1X",
"FZHqZK3gKSu",
"4zne66bu82V",
"3evcG0lmFR",
"52bnQCEZ-ZN",
"miLKDykR_ik",
"ZAXPwDBqTy",
"NnacZ4JB1g3",
"1zGD6ZhbZJ8",
"LXAasl0rmvq",
"sDg6GZHRktx",
"KPYcFDG0ti",... | [
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_r... | [
"## Summary\n\nThis paper proposed a new formulation of federated learning, which balances between traditional global model and purely local models. The authors discuss the advantages of the new formulation and propose a new algorithm L2GD to solve the problem. They theoretically analyzed the communication complexi... | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
4,
4
] | [
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
4
] | [
"iclr_2021_nLYMajjctMh",
"hVJ5fKkCWXe",
"FZHqZK3gKSu",
"NnacZ4JB1g3",
"FZHqZK3gKSu",
"1zGD6ZhbZJ8",
"1nt_anxUt1X",
"ZAXPwDBqTy",
"4zne66bu82V",
"sDg6GZHRktx",
"R4cRgYjXq3",
"R4cRgYjXq3",
"R4cRgYjXq3",
"R4cRgYjXq3",
"KPYcFDG0ti",
"LXAasl0rmvq",
"zGzn9-i-wtp",
"iclr_2021_nLYMajjctMh"... |
iclr_2021_3YQAVD9_Dz3 | NOSE Augment: Fast and Effective Data Augmentation Without Searching | Data augmentation has been widely used for enhancing the diversity of training data and model generalization. Different from traditional handcrafted methods, recent research introduced automated search for optimal data augmentation policies and achieved state-of-the-art results on image classification tasks. However, t... | withdrawn-rejected-submissions | All reviewers generally admit that the motivation of realizing search-free autoaugment is reasonable and important. However, they also raised many concerns regarding the experimental evaluation to validate the practical effectiveness of the method. In particular, unclear discussion with respect to ablation studies, and... | train | [
"YzxUMHmuhDE",
"olBHhAmBT6",
"rSn2Ci3Mr_t",
"x2A7t8oQfJ9",
"SSwBso19fy",
"WfG5Rq_1QrC",
"9hh2rri44c",
"0674N7nNN9x",
"9yKsVrnfCO7",
"MDDfeoi4NYB",
"4EJhbV75RVI",
"7CqobXrjGE"
] | [
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Thank you for your reply with inspiring words and kind understanding. Following your kind suggestion, we are collecting our no-augmentation (standard) baseline results. Based on the results we obtained so far, we can see that most of our baseline results are close to what AutoAugment reported in their paper, which... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
4,
5
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
3,
4
] | [
"SSwBso19fy",
"MDDfeoi4NYB",
"7CqobXrjGE",
"7CqobXrjGE",
"WfG5Rq_1QrC",
"MDDfeoi4NYB",
"4EJhbV75RVI",
"4EJhbV75RVI",
"7CqobXrjGE",
"iclr_2021_3YQAVD9_Dz3",
"iclr_2021_3YQAVD9_Dz3",
"iclr_2021_3YQAVD9_Dz3"
] |
iclr_2021_U6Xpa5R-E1 | Neural Potts Model | We propose the Neural Potts Model objective as an amortized optimization problem. The objective enables training a single model with shared parameters to explicitly model energy landscapes across multiple protein families. Given a protein sequence as input, the model is trained to predict a pairwise coupling matrix for... | withdrawn-rejected-submissions | This is a creative piece of work wherein learning of what is normally family-specific Potts models is turned into an amortized optimization problem across different families of proteins. The Potts models are learned with a pseudolikelihood approach, and the evaluation of the model against baselines is performed only on... | train | [
"jGK-5l3MqJ0",
"RR-bm_lhDG-",
"o4uDVmiPlb-",
"5R5b0pIDXSY",
"MkIFKNGsSU",
"HEnQk7dHAYr",
"7GbrtFxsjO",
"R0uqyDPyLbB",
"RaKBBHsAfI",
"2sUOV0hn9Gm",
"9u-kxZKIe6k"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"The paper proposes a new object called Neural Potts Model (NPM) to train a Transformer to learn the local energy landscape of protein sequences. The problem of modeling energy landscapes using the power of techniques in natural language processing (NLP) is a timely and interesting problem. However, there are some ... | [
6,
6,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
7
] | [
4,
3,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4
] | [
"iclr_2021_U6Xpa5R-E1",
"iclr_2021_U6Xpa5R-E1",
"iclr_2021_U6Xpa5R-E1",
"iclr_2021_U6Xpa5R-E1",
"HEnQk7dHAYr",
"iclr_2021_U6Xpa5R-E1",
"o4uDVmiPlb-",
"jGK-5l3MqJ0",
"RR-bm_lhDG-",
"9u-kxZKIe6k",
"iclr_2021_U6Xpa5R-E1"
] |
iclr_2021_4VixXVZJkoY | TRIP: Refining Image-to-Image Translation via Rival Preferences | We propose a new model to refine image-to-image translation via an adversarial ranking process. In particular, we simultaneously train two modules: a generator that translates an input image to the desired image with smooth subtle changes with respect to some specific attributes; and a ranker that ranks rival preferenc... | withdrawn-rejected-submissions | All reviewers gave either borderline or negative scores; unfortunately, discussion was not lively, so scores remained the same. No reviewers voice strong support for acceptance, but acknowledge several merits of the work. | train | [
"kQeAomb7ebx",
"bSAEV4ziSv",
"NlA9T75d3uP",
"MnZxA3IqE4M",
"o_gaRz3-WAj",
"FcLuM4yVS9n",
"_2TE9JUnBYl",
"bfHngJtXf3",
"xDYAIPavoPv",
"7GjmtVTe0Nc",
"8P46HAYl1zO"
] | [
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Thank you very much for your comments. Below is a report on how we address your concern. \n\nWeakness 1: Concerns about the contribution and the target scope of this work. \n - R1: We follow the community [1,2] to denote the problem we addressed with the “image-to-image translation” term. This is the first work ... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
6,
5
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
4,
4
] | [
"bfHngJtXf3",
"7GjmtVTe0Nc",
"iclr_2021_4VixXVZJkoY",
"iclr_2021_4VixXVZJkoY",
"FcLuM4yVS9n",
"xDYAIPavoPv",
"8P46HAYl1zO",
"iclr_2021_4VixXVZJkoY",
"iclr_2021_4VixXVZJkoY",
"iclr_2021_4VixXVZJkoY",
"iclr_2021_4VixXVZJkoY"
] |
iclr_2021_HP-tcf48fT | Learning to Search for Fast Maximum Common Subgraph Detection | Detecting the Maximum Common Subgraph (MCS) between two input graphs is fundamental for applications in biomedical analysis, malware detection, cloud computing, etc. This is especially important in the task of drug design, where the successful extraction of common substructures in compounds can reduce the number of exp... | withdrawn-rejected-submissions | The paper present a new learning-based approach` to solve the Maximum Common Subgraph problem. All the reviewers find the idea of using GCN and RL to guide the branch and bound interesting although, even after reading the rebuttal, there are some important concerns about the paper.
The main issue raised by many review... | train | [
"SBqsJfffHwL",
"-t191eSd9-c",
"oYIZU1hlgtI",
"JeV31kiGSDV",
"xNVAm85DkTb",
"LUc9TnjQBB-",
"uUWp_4bI-46",
"jatjDT21zsj",
"CRLWd04cSpe",
"ct7YNvubwA",
"eFUPHn8zGJ",
"SOpxPbN2Zch",
"JORRQ_FT8A2",
"SfO1khIjQys",
"IrsjyoIypTt",
"9BXM6zs1gV"
] | [
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"In general, GLSearch is a less interpretable but more powerful method compared to baselines. That said, we did find trends from GLSearch that may be useful to producing hand-crafted heuristics.\n \nGLSearch identifies “smart” nodes which can lead to larger common subgraphs faster. For example, in the road networks... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
5,
7
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
4,
4
] | [
"SfO1khIjQys",
"iclr_2021_HP-tcf48fT",
"SfO1khIjQys",
"SfO1khIjQys",
"SfO1khIjQys",
"IrsjyoIypTt",
"IrsjyoIypTt",
"9BXM6zs1gV",
"9BXM6zs1gV",
"iclr_2021_HP-tcf48fT",
"SfO1khIjQys",
"SfO1khIjQys",
"IrsjyoIypTt",
"iclr_2021_HP-tcf48fT",
"iclr_2021_HP-tcf48fT",
"iclr_2021_HP-tcf48fT"
] |
iclr_2021_arD29HCZG6O | Linking average- and worst-case perturbation robustness via class selectivity and dimensionality | Representational sparsity is known to affect robustness to input perturbations in deep neural networks (DNNs), but less is known about how the semantic content of representations affects robustness. Class selectivity—the variability of a unit’s responses across data classes or dimensions—is one way of quantifying the s... | withdrawn-rejected-submissions | The reviewers had raised a number of concerns which were mostly addressed during the discussion phase thanks to the additional experiments/explanations that the authors provided. However, some of the reviewers are not yet convinced about the main claims of the paper. While the paper provides a number of interesting/imp... | train | [
"tdR-DJgJAnz",
"EKpqhBPFcBP",
"ZRbly7aZt9N",
"Z1wpLGlrzR9",
"EXyXADjiJId",
"O98GCekT4RB",
"7AaWE5USG0A",
"0GnF9xxOag",
"FnPBYDGeJMi",
"qLJeAjizrpA",
"cH9YIAbG3l4",
"_jQ0kNo7X5",
"hrMs7Idnqf"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author"
] | [
"This work investigates two classes of perturbation robustness, average-case perturbations which are considered to be naturally occurring in image data, and worst-case perturbations that are perturbations generated by an adversary. Neural network susceptibility to these perturbations is evaluated with respect to a ... | [
6,
7,
4,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
3,
4,
3,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
"iclr_2021_arD29HCZG6O",
"iclr_2021_arD29HCZG6O",
"iclr_2021_arD29HCZG6O",
"iclr_2021_arD29HCZG6O",
"ZRbly7aZt9N",
"7AaWE5USG0A",
"Z1wpLGlrzR9",
"FnPBYDGeJMi",
"EXyXADjiJId",
"EKpqhBPFcBP",
"_jQ0kNo7X5",
"tdR-DJgJAnz",
"iclr_2021_arD29HCZG6O"
] |
iclr_2021_u15gHPQViL | Zero-Shot Recognition through Image-Guided Semantic Classification | We present a new visual-semantic embedding method for generalized zero-shot learning. Existing embedding-based methods aim to learn the correspondence between an image classifier (visual representation) and its class prototype (semantic representation) for each class. Inspired by the binary relevance method for multi-l... | withdrawn-rejected-submissions | This paper presents work on zero-shot learning. The reviewers appreciated the simplicity of the method and its clear exposition. However, concerns were raised over novelty, motivation, and empirical validation. After reading the authors' response, the reviewers remained of the opinion that these concerns have not ye... | train | [
"RUpNdrCWu4i",
"keueJPowQXP",
"thCRuYL2mn",
"MttQK45Spgf",
"7G_H_hmOst0",
"29LfKikK6i",
"KS16KcuKPMi",
"uQi_d3WESi",
"lZW_eBwJ7NH"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"This paper proposes a visual-semantic embedding model useful for generalized zero-shot learning. The proposed model transforms an image into a label classifier, which is then used to predict the correct label in the semantic space. \n\nThe paper is well constructed and easy to read. It provides a good presentation... | [
4,
3,
3,
-1,
-1,
-1,
-1,
-1,
4
] | [
4,
4,
5,
-1,
-1,
-1,
-1,
-1,
5
] | [
"iclr_2021_u15gHPQViL",
"iclr_2021_u15gHPQViL",
"iclr_2021_u15gHPQViL",
"iclr_2021_u15gHPQViL",
"keueJPowQXP",
"thCRuYL2mn",
"RUpNdrCWu4i",
"lZW_eBwJ7NH",
"iclr_2021_u15gHPQViL"
] |
iclr_2021_yKYiyoHG4N3 | Optimal Neural Program Synthesis from Multimodal Specifications | Multimodal program synthesis, which leverages different types of user input to synthesize a desired program, is an attractive way to scale program synthesis to challenging settings; however, it requires integrating noisy signals from the user (like natural language) with hard constraints on the program's behavior. This... | withdrawn-rejected-submissions | The paper proposes a new multimodal neuro-symbolic technique for synthesizing programs. The specification is given in natural language (soft constraints) and input-output examples (hard constraints). The multimodal program synthesis is formulated as a constrained maximization problem where the goal is to find a program... | train | [
"-X-ne7Z9l_",
"KYz3TYqXG9c",
"4DZYyTAXC9z",
"LMHqIoXwpI4",
"3X2jFvLU9vj",
"n1K5e_asDjr",
"R7NON_X0QKa",
"sa13c2I9Bx9",
"YrvQG5sIqMp",
"15LtDozPUVL",
"Jg3PTr_KmU8"
] | [
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Thanks for the detailed reply.\n\n> I tried to came up with a simple example and a prefix that could be pruned by your proposed method, but not by the prefix-matching trick. \n\nOur approach can prune those partial regexes that do not have deterministic prefixes, e.g., contain(concat(<let>,?)), and(endwith(?),?), ... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
5,
7,
4
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
4,
4,
4
] | [
"KYz3TYqXG9c",
"LMHqIoXwpI4",
"iclr_2021_yKYiyoHG4N3",
"sa13c2I9Bx9",
"YrvQG5sIqMp",
"Jg3PTr_KmU8",
"15LtDozPUVL",
"iclr_2021_yKYiyoHG4N3",
"iclr_2021_yKYiyoHG4N3",
"iclr_2021_yKYiyoHG4N3",
"iclr_2021_yKYiyoHG4N3"
] |
iclr_2021_1yXhko8GZEE | Precondition Layer and Its Use for GANs | One of the major challenges when training generative adversarial nets (GANs) is instability. To address this instability spectral normalization (SN) is remarkably successful. However, SN-GAN still suffers from training instabilities, especially when working with higher-dimensional data. We find that those instabilit... | withdrawn-rejected-submissions | This paper presents a new method for training GAN by adding a precondition Layer. All reviewers are positive about the empirical results. However, some concerns were raised about the justification: (1) Only linear networks are considered, which is a bit impractical; (2) Existing work has shown the importance of control... | train | [
"EdiKydx5iVM",
"7wV9IH_mWEa",
"Zsgu4PbUiI",
"gNTt87L3kw8",
"QTFkliunEBU",
"7PmaEJMDnZg",
"QioMTbcUNsB",
"Ys3bgScCUFo",
"-o9swdIngKP",
"nemx5xWqYwm",
"R5YnsewX608"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Update: \nThank the authors for the detailed feedback. I decide to keep the score.\n---\n---\nThe paper shows that the failure mode of spectral normalization (SN) is often accompanied by large condition numbers in the discriminator layers. Motivated from this observation, the paper proposes to control the conditi... | [
7,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
5,
6
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
3,
3
] | [
"iclr_2021_1yXhko8GZEE",
"nemx5xWqYwm",
"nemx5xWqYwm",
"-o9swdIngKP",
"-o9swdIngKP",
"R5YnsewX608",
"EdiKydx5iVM",
"R5YnsewX608",
"iclr_2021_1yXhko8GZEE",
"iclr_2021_1yXhko8GZEE",
"iclr_2021_1yXhko8GZEE"
] |
iclr_2021_Sm_4MDxPWXf | StructFormer: Joint Unsupervised Induction of Dependency and Constituency Structure from Masked Language Modeling | There are two major classes of natural language grammars --- the dependency grammar that models one-to-one correspondences between words and the constituency grammar that models the assembly of one or several corresponded words. While previous unsupervised parsing methods mostly focus on only inducing one class of gram... | withdrawn-rejected-submissions | This paper presents a novel approach to grammar induction. Like older work by Klein and Manning, the paper finds benefit in jointly inducing both constituency and dependency structure. However, unlike most approaches to grammar induction, the model is not generative -- rather, it is a transformer-based architecture tha... | train | [
"Z9WCzHGdU2O",
"bcP2cwJP_uT",
"VK4oHunUfi",
"GFfMiUlj8o7",
"9s_6eVvBvla"
] | [
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Updates after discussion/revision period:\n\nIt appears that the paper has improved. However, the changes appear to be so substantial that the paper is now essentially a different paper which would require a new review process. \n\n---------\n\nThis paper describes a neural architecture that resembles the transfor... | [
4,
-1,
4,
6,
5
] | [
4,
-1,
3,
4,
4
] | [
"iclr_2021_Sm_4MDxPWXf",
"iclr_2021_Sm_4MDxPWXf",
"iclr_2021_Sm_4MDxPWXf",
"iclr_2021_Sm_4MDxPWXf",
"iclr_2021_Sm_4MDxPWXf"
] |
iclr_2021_AJTAcS7SZzf | AUTOSAMPLING: SEARCH FOR EFFECTIVE DATA SAMPLING SCHEDULES | Data sampling acts as a pivotal role in training deep learning models. However, an effective sampling schedule is difficult to learn due to its inherent high-dimension as a hyper-parameter. In this paper, we propose the AutoSampling method to automatically learn sampling schedules for model training, which consists of ... | withdrawn-rejected-submissions | The work focuses on a new method for sampling hyper-parameter based on an "Population-Based Training" schedule that tend to sample more often configurations that gave good performances in the past. The authors have conducted experiments to verify the superior of their method, especially for the effectiveness and genera... | train | [
"y-sy9RvBOvR",
"QklijinItPa",
"y8STq3vpLos",
"-uekHB7HwvC",
"HcUj9JCXDX",
"SHzPcO5KVp"
] | [
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"We thank the reviewer for the valuable feedback and would like to address some of the reviewer’s concerns.\n\n* \"The chosen baselines are essentially standard sampling scheme or variants of the proposed method. Authors should compare with a few state-of-the-art data-sampling or data-reweighting methods, such as F... | [
-1,
-1,
-1,
3,
6,
5
] | [
-1,
-1,
-1,
5,
4,
4
] | [
"SHzPcO5KVp",
"-uekHB7HwvC",
"HcUj9JCXDX",
"iclr_2021_AJTAcS7SZzf",
"iclr_2021_AJTAcS7SZzf",
"iclr_2021_AJTAcS7SZzf"
] |
iclr_2021_34KAZ9HbJco | Adapt-and-Adjust: Overcoming the Long-tail Problem of Multilingual Speech Recognition | One crucial challenge of real-world multilingual speech recognition is the long-tailed distribution problem, where some resource-rich languages like English have abundant training data, but a long tail of low-resource languages have varying amounts of limited training data. To overcome the long-tail problem, in this pa... | withdrawn-rejected-submissions | As one of the reviewers' comment, the paper presents "a mixed of tricks" for the multilingual speech recognition, which includes 1) the use of a pretrained mBERT, 2) dual-adapter and 3) prior adjusting.
First, the relative gains of the pretrained mBERT is marginal (Section 3.3.1). Secondly, using 1) on top of 2) is un... | train | [
"Jo24Y4cz6eX",
"3MzOA4plY2L",
"_nuX9OJ12er",
"zbvalG-CEKQ",
"J4Yr0-qZrQ",
"kjX4RNWIlA",
"bMX4W3Hlc6D",
"ajWdOOp-c1_",
"p8V0rBcaGou",
"_AwUao5Z5PH",
"Abs6kJEPOIS",
"v8VKCulOBKd",
"ToY19jJfIo",
"LX2AQdUe6io",
"A50BpS-ipA0",
"um3uIS-vK0c",
"WYGcuNmQyJv"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"This paper addresses multi-lingual speech synthesis, where one ASR model is responsible for recognizing speech in multiple languages. In this example the authors look at 11 languages with between 80 and 4 hours of training data. The \"long-tail problem\" (which isn't clearly stated) that this work is addressing ... | [
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
4,
5,
6
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
4,
3
] | [
"iclr_2021_34KAZ9HbJco",
"um3uIS-vK0c",
"um3uIS-vK0c",
"LX2AQdUe6io",
"A50BpS-ipA0",
"WYGcuNmQyJv",
"iclr_2021_34KAZ9HbJco",
"_AwUao5Z5PH",
"ajWdOOp-c1_",
"Abs6kJEPOIS",
"ToY19jJfIo",
"A50BpS-ipA0",
"Jo24Y4cz6eX",
"iclr_2021_34KAZ9HbJco",
"iclr_2021_34KAZ9HbJco",
"iclr_2021_34KAZ9HbJco... |
iclr_2021_sbyjwhxxT8K | Near-Black-Box Adversarial Attacks on Graph Neural Networks as An Influence Maximization Problem | Graph neural networks (GNNs) have attracted increasing interests. With broad deployments of GNNs in real-world applications, there is an urgent need for understanding the robustness of GNNs under adversarial attacks, especially in realistic setups. In this work, we study the problem of attacking GNNs in a restricted ne... | withdrawn-rejected-submissions | This paper relates the problem of influence maximization and adversarial attacks on GCNs.
The paper, and its formulation and assumptions stirred up quite a discussion among the reviewers and the authors. I do appreciate the thorough rebuttal that the authors provided, and the reviewers did take it into account (and re... | train | [
"SQ2GEyyPzmN",
"IE1LpHqMwzT",
"XVQQbSInyEh",
"6DEHktX1-kd",
"PYmwdf6fbnz",
"npAglBvXep3",
"pz_h8e1ztHR",
"aeTKN5DI2OP",
"vvm2PTl7n51",
"xsrpHNpp33w",
"Iiv8NjsjPp0",
"guAkHqxA-DH",
"egHLd4L1Gj",
"v4tlgRbGopE",
"NtW2X-j04-T",
"1wI2ra_ph_",
"mABo_4IVos8",
"n0lJ2CJKKVa",
"nUUJy9ylFl3... | [
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
... | [
"Thanks for updating the evaluation. That is encouraging!\n\n---\n\nBeyond the theoretical justifications, this assumption also helps us derive practically effective attack strategies. We note that the derivation of Eq. (3) relies on this assumption, which is a critical foundation for the development of the followi... | [
-1,
5,
-1,
-1,
-1,
-1,
-1,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
6
] | [
-1,
4,
-1,
-1,
-1,
-1,
-1,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
4
] | [
"XVQQbSInyEh",
"iclr_2021_sbyjwhxxT8K",
"6DEHktX1-kd",
"pz_h8e1ztHR",
"vvm2PTl7n51",
"xsrpHNpp33w",
"NtW2X-j04-T",
"iclr_2021_sbyjwhxxT8K",
"mABo_4IVos8",
"Iiv8NjsjPp0",
"guAkHqxA-DH",
"egHLd4L1Gj",
"v4tlgRbGopE",
"nUUJy9ylFl3",
"IE1LpHqMwzT",
"aeTKN5DI2OP",
"1wI2ra_ph_",
"s3W8yQqf... |
iclr_2021_e60-SyRXtRt | GANMEX: Class-Targeted One-vs-One Attributions using GAN-based Model Explainability | Attribution methods have been shown as promising approaches for identifying key features that led to learned model predictions. While most existing attribution methods rely on a baseline input for performing feature perturbations, limited research has been conducted to address the baseline selection issues. Poor choice... | withdrawn-rejected-submissions | This work investigates the choice of a 'baseline' for attribution methods. Such a choice is important and can heavily influence the outcome of any analysis that involves attribution methods. The work proposes doing (1) one-vs-one attribution in a sort of contrastive fashion (2) generating baselines using StarGAN.
The ... | train | [
"RQF2_GVImA0",
"aLrtlV_FiWs",
"H4JiLQb7Ba9",
"xNSOgLJ0WrV",
"XVrQodO0FrY",
"Bo2Z-43O5-R",
"ROvE1AF-g9_",
"V6wirPRcQFw",
"M56_-tkKsZ",
"02S6oHO7VfN"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Paper Summary:\n\nThis paper considers the less-explored baseline selection issue in attribution methods for one-vs-one explanations of multi-class classifiers. The key insight is to construct the closest and realistic target class baseline. To this end, an existing image-to-image translation GAN model, namely Sta... | [
5,
-1,
-1,
-1,
-1,
-1,
-1,
5,
5,
4
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
3,
4,
2
] | [
"iclr_2021_e60-SyRXtRt",
"RQF2_GVImA0",
"02S6oHO7VfN",
"iclr_2021_e60-SyRXtRt",
"V6wirPRcQFw",
"ROvE1AF-g9_",
"M56_-tkKsZ",
"iclr_2021_e60-SyRXtRt",
"iclr_2021_e60-SyRXtRt",
"iclr_2021_e60-SyRXtRt"
] |
iclr_2021_sAzh_FTFDxz | Understanding the Effect of Bias in Deep Anomaly Detection | Anomaly detection presents a unique challenge in machine learning, due to the scarcity of labeled anomaly data. Recent work attempts to mitigate such problems by augmenting training of deep anomaly detection models with additional labeled anomaly samples. However, the labeled data often does not align with the target d... | withdrawn-rejected-submissions | This paper studies the effect of anomaly detection using supervised learning with non-representative abnormal examples on the TPR of the anomaly detection model. Experiments demonstrate that when the abnormal examples presented in the training set are not representative of the abnormal examples in the target distributi... | test | [
"5Ltv7Oj3GSa",
"B6ZuZ5vQd5N",
"hHGl3AqsHTA",
"e6ZKkHPOKHV",
"hB9ddk2zcXB",
"6NG3E9jJm_r",
"zuAhJyUY0SC",
"ZHhUB6rOTv",
"maJBTUgsEfm",
"ngz6YJdocnL",
"ni7pePhVC48",
"kYt--aVKFK",
"4qeVoQH5GUv"
] | [
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"\n**UPDATE**\n\nI acknowledge that I have read the author responses as well as the other reviews. I appreciate the clarifications and improvements made during the rebuttal phase, which I think have further strengthened this work.\n\nI find the key contributions of this work to be (i) demonstrating that recent meth... | [
7,
-1,
-1,
7,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4
] | [
4,
-1,
-1,
4,
2,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5
] | [
"iclr_2021_sAzh_FTFDxz",
"iclr_2021_sAzh_FTFDxz",
"ngz6YJdocnL",
"iclr_2021_sAzh_FTFDxz",
"iclr_2021_sAzh_FTFDxz",
"ZHhUB6rOTv",
"4qeVoQH5GUv",
"hB9ddk2zcXB",
"5Ltv7Oj3GSa",
"e6ZKkHPOKHV",
"zuAhJyUY0SC",
"ni7pePhVC48",
"iclr_2021_sAzh_FTFDxz"
] |
iclr_2021_ni_nys-C9D6 | Differentiate Everything with a Reversible Domain-Specific Language | Reverse-mode automatic differentiation (AD) suffers from the issue of having too much space overhead to trace back intermediate computational states for backpropagation.
The traditional method to trace back states is called checkpointing that stores intermediate states into a global stack and restore state throug... | withdrawn-rejected-submissions | After reading the paper, reviews and authors’ feedback. The meta-reviewer agrees with the reviewers that the paper touches an interesting topic (reversible computing) but could be improved in the area of presentation and evaluation. Therefore this paper is rejected.
Thank you for submitting the paper to ICLR.
| train | [
"-lBKDkNq67",
"JS9dfthHV2c",
"7FWdsGl9KPW",
"2L8JhFf9AAZ",
"KWSF6SVMC1G",
"P9t5qeUfiO",
"_AT54vZ7gVL",
"8mrrAHx7Ke",
"kD_XvDn7ReI",
"S9SDYgzXMgY",
"bp9eYLlCnWQ"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"# Changes after rebuttal\nThanks to the authors for their answers to the questions and their revisions to improve the manuscript. It is useful to have further descriptions of reversible computing for an audience that may be unfamiliar with the topic. I would encourage the authors to make further revisions to more ... | [
5,
6,
-1,
-1,
-1,
5,
-1,
-1,
-1,
6,
4
] | [
3,
4,
-1,
-1,
-1,
3,
-1,
-1,
-1,
2,
4
] | [
"iclr_2021_ni_nys-C9D6",
"iclr_2021_ni_nys-C9D6",
"iclr_2021_ni_nys-C9D6",
"JS9dfthHV2c",
"P9t5qeUfiO",
"iclr_2021_ni_nys-C9D6",
"-lBKDkNq67",
"S9SDYgzXMgY",
"bp9eYLlCnWQ",
"iclr_2021_ni_nys-C9D6",
"iclr_2021_ni_nys-C9D6"
] |
iclr_2021_zleOqnAUZzl | Are all outliers alike? On Understanding the Diversity of Outliers for Detecting OODs | Deep neural networks (DNNs) are known to produce incorrect predictions with very high confidence on out-of-distribution (OOD) inputs. This limitation is one of the key challenges in the adoption of deep learning models in high-assurance systems such as autonomous driving, air traffic management, and medical diagnosis. ... | withdrawn-rejected-submissions | **Problem Significance** This paper introduces an interesting taxonomy of OODs and proposed an integrated approach to detect different types of OODs. The AC agrees on the importance of a fine-grained characterization of outliers given the large OOD uncertainty space.
**Technical contribution** The key idea of the pa... | train | [
"cPhM2rIxXx",
"uS20jXEK3ox",
"shp5RiKLfWA",
"IEs6TrqM8HW",
"gs0OdJYScQL",
"19aq6t2odFB",
"16uTwLwSJgm",
"IPXB1gm5o5Z",
"LlbaGqTmlN",
"I3YZsU7tnj",
"IDg86c7wWde",
"X9Z8Oi8sbeN",
"zBBb4vQ2gpN",
"2ATb383ejfU",
"U2eexTw6S4",
"-Na6WT_XFbR",
"GZSJSM9JcK4",
"HEAezjYd_HF",
"8Xz_3Oeais",
... | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"author",
... | [
"This paper introduces a taxonomy of OODs and proposed an integrated approach to detect different types of OODs. Their taxonomy classifies OOD on the nature of their uncertainty and they show that no single state-of-the-art approach detects all these OOD types. Motivated by this observation, they combine multiple e... | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
5,
5
] | [
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
2,
5,
4
] | [
"iclr_2021_zleOqnAUZzl",
"iclr_2021_zleOqnAUZzl",
"iclr_2021_zleOqnAUZzl",
"iclr_2021_zleOqnAUZzl",
"iclr_2021_zleOqnAUZzl",
"iclr_2021_zleOqnAUZzl",
"qurbtGXSRj-",
"LlbaGqTmlN",
"I3YZsU7tnj",
"IDg86c7wWde",
"X9Z8Oi8sbeN",
"8Xz_3Oeais",
"2ATb383ejfU",
"U2eexTw6S4",
"-Na6WT_XFbR",
"GZSJ... |
iclr_2021_Z_TwEk_sP34 | Does Adversarial Transferability Indicate Knowledge Transferability? | Despite the immense success that deep neural networks (DNNs) have achieved, \emph{adversarial examples}, which are perturbed inputs that aim to mislead DNNs to make mistakes, have recently led to great concerns. On the other hand, adversarial examples exhibit interesting phenomena, such as \emph{adversarial transferabi... | withdrawn-rejected-submissions | This paper studies the relationship between adversarial transferability and knowledge transferability. It develops two metrics to measure adversarial transferability and a theoretical framework to justify the positive correlation between adversarial transferability and knowledge transferability. Synthetic experiments s... | train | [
"-4gPGWQhLnE",
"23uRuvqjUYx",
"lHQCdcrUfrg",
"ez-cjPebZHs",
"e-RbIL3bac",
"gCwkKYAc9hl",
"GaUkYV8YaL",
"Gm0TSMtcevn",
"4LWHPqj-iYT"
] | [
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"We thank the reviewers for the constructive comments. We have revised the manuscript to incorporate the reviewers’ valuable feedback, and added a set of synthetic experiments to further verify our theoretical findings. Please find the summary of our revision below, and major revisions are highlighted in blue in th... | [
-1,
-1,
-1,
-1,
-1,
5,
5,
5,
5
] | [
-1,
-1,
-1,
-1,
-1,
4,
4,
3,
3
] | [
"iclr_2021_Z_TwEk_sP34",
"Gm0TSMtcevn",
"GaUkYV8YaL",
"gCwkKYAc9hl",
"4LWHPqj-iYT",
"iclr_2021_Z_TwEk_sP34",
"iclr_2021_Z_TwEk_sP34",
"iclr_2021_Z_TwEk_sP34",
"iclr_2021_Z_TwEk_sP34"
] |
iclr_2021_Q2iaAc-4I1v | Causal Curiosity: RL Agents Discovering Self-supervised Experiments for Causal Representation Learning | Humans show an innate ability to learn the regularities of the world through interaction. By performing experiments in our environment, we are able to discern the causal factors of variation and infer how they affect the dynamics of our world. Analogously, here we attempt to equip reinforcement learning agents with the... | withdrawn-rejected-submissions | This paper discusses how one can equip reinforcement learning agents with an intrinsic reward function that helps identifying factors of variation within a family of MDPs, effectively allowing agents to do experiments in the environment. This is interpreted as causal factors that control important aspects of the enviro... | train | [
"WaAZgwcL8tO",
"V2yUJWES51_",
"RCd532TUIsw",
"IIFDQ0rqbRX",
"uYXE_Pw4Zo0",
"wISLvGVw05E",
"_DtrZtf9n-i",
"-HPoPtDESU8",
"UP7BMMX5ii",
"4PEvJg3LGWP",
"8wXKMJFGX2p",
"KXkgxcY-ffJ",
"URovzkDMCYG",
"MQSRODyOwEE",
"pRkOjNjrGAu",
"7rn-Gwzf_w",
"tG0gtaN1jk_",
"ER4_Qe8TI4g",
"8zElVBNSv8z... | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"#### **Summary**\n\nThis paper develops an intrinsic reward to help identify factors of variation within a family of MDPs. This intrinsic reward takes a form of curiosity and is used to develop initial behaviors to identify the causes of the latent variation in the environment dynamics. The experiments are used to... | [
5,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
5
] | [
4,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
4
] | [
"iclr_2021_Q2iaAc-4I1v",
"iclr_2021_Q2iaAc-4I1v",
"wISLvGVw05E",
"_DtrZtf9n-i",
"-HPoPtDESU8",
"8wXKMJFGX2p",
"KXkgxcY-ffJ",
"URovzkDMCYG",
"iclr_2021_Q2iaAc-4I1v",
"V2yUJWES51_",
"KXkgxcY-ffJ",
"URovzkDMCYG",
"8zElVBNSv8z",
"pRkOjNjrGAu",
"ER4_Qe8TI4g",
"tG0gtaN1jk_",
"WaAZgwcL8tO",... |
iclr_2021_QpT9Q_NNfQL | NeurWIN: Neural Whittle Index Network for Restless Bandits via Deep RL | Whittle index policy is a powerful tool to obtain asymptotically optimal solutions for the notoriously intractable problem of restless bandits. However, finding the Whittle indices remains a difficult problem for many practical restless bandits with convoluted transition kernels. This paper proposes NeurWIN, a neural W... | withdrawn-rejected-submissions | This paper approximates the Whittle index in restless bandits using a neural network. Finding the Whittle index is a difficult problem and all reviewers agreed on this. Nevertheless, the scores of this paper are split between 2x 4 and 2x 7, essentially along the line of whether this paper is too preliminary to be accep... | train | [
"8YclC9wKR5w",
"QZw48Xg9emH",
"REKYYeGiN0f",
"H1I_dsDTmec",
"AKJvLddDQ4H",
"3CCs68pmHQc",
"lx0V6i2MYtb",
"13c5ZHTEvxg",
"E-gFZgQQT85",
"z6TpuZv9m0b",
"FIlvERPoMUr",
"BLmR15lP1_",
"2vFX2xzXg3j",
"dp1CGDuBceC",
"680TsOaXtQZ",
"6aXXIIa5YEW"
] | [
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"This paper proposes the use of Deep Learning, namely a multi-layer perceptron, for approximating the Whittle index in restless bandits.\n\nThe introduction and the related works are well written. The problem and the background on restless bandits are clearly exposed.\n\nHowever, there are a lot of issues in the pr... | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
7,
7,
4
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3,
4
] | [
"iclr_2021_QpT9Q_NNfQL",
"REKYYeGiN0f",
"AKJvLddDQ4H",
"z6TpuZv9m0b",
"lx0V6i2MYtb",
"13c5ZHTEvxg",
"2vFX2xzXg3j",
"BLmR15lP1_",
"iclr_2021_QpT9Q_NNfQL",
"dp1CGDuBceC",
"680TsOaXtQZ",
"6aXXIIa5YEW",
"8YclC9wKR5w",
"iclr_2021_QpT9Q_NNfQL",
"iclr_2021_QpT9Q_NNfQL",
"iclr_2021_QpT9Q_NNfQL... |
iclr_2021_BUPIRa1D2J | Trans-Caps: Transformer Capsule Networks with Self-attention Routing | Capsule Networks (CapsNets) have shown to be a promising alternative to Convolutional Neural Networks (CNNs) in many computer vision tasks, due to their ability to encode object viewpoint variations. The high computational complexity and numerical instability of iterative routing mechanisms stem from the challenging na... | withdrawn-rejected-submissions | The paper proposes a new variant of capsule networks, where iterative routing is replaced by an attention-based procedure inspired by Induced Set Attention from Set Transformers. The method is competitive on several classification benchmarks and improves generalization to unseen views on SmallNORB.
The reviewers note ... | train | [
"t8vwhjndEG",
"TGnULJ4afQ9",
"A3pgMGE_pUF",
"TMljBKkK-0t",
"gWV5Af6R1Lm",
"Vi6N4bC1ZFk",
"t1cWqe11SCf",
"hbLKyiFZw8j",
"y8Q8tQalXOE",
"BtqWCcjI6lj",
"mZsu44K5d8V",
"-u9kLUvvl3",
"bhuTtX8Womt",
"DiOYP2On1Nt",
"wq0UsEjdJ_G",
"npNzkFhVL6b",
"--_BAEFDq-",
"x4qiZz21Q4Y",
"wIM8uB9fim3"... | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"... | [
"The paper proposes to use the self-attention to find the agreement among the capsules of consecutive layers of a capsule network instead of iterative routing procedure. To reduce the computational and memory complexity of self-transformer capsules of each layer are considered as a set and follow set-transformer an... | [
6,
7,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
6
] | [
4,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3
] | [
"iclr_2021_BUPIRa1D2J",
"iclr_2021_BUPIRa1D2J",
"y8Q8tQalXOE",
"BtqWCcjI6lj",
"hbLKyiFZw8j",
"iclr_2021_BUPIRa1D2J",
"x4qiZz21Q4Y",
"wq0UsEjdJ_G",
"--_BAEFDq-",
"npNzkFhVL6b",
"wIM8uB9fim3",
"TGnULJ4afQ9",
"t8vwhjndEG",
"Kb2GvJRoWqF",
"mZsu44K5d8V",
"mZsu44K5d8V",
"mZsu44K5d8V",
"m... |
iclr_2021_MMXhHXbNsa- | Blind Pareto Fairness and Subgroup Robustness | With the wide adoption of machine learning algorithms across various application domains, there is a growing interest in the fairness properties of such algorithms. The vast majority of the activity in the field of group fairness addresses disparities between predefined groups based on protected features such as gender,... | withdrawn-rejected-submissions | This paper studies the problem of Pareto fairness without having pre-defined protected groups. The reviewers agree that the problem studied here is interesting and relevant. During the initial review period, reviewers identified a major correctness issue. The authors have then substantially changed the algorithm and ex... | train | [
"jl9JKpJoh-m",
"VXWWlGm45rD",
"Rm5T2RH2Ovy",
"qcI4aFev1MD",
"hSs5Z87Oi1N",
"LVCTxL-Ollf",
"PmUtZQI4EnM"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"## Summary ##\n\nThis paper considers the relevant problem of group fairness in ML when there are no predefined groups. They aim to provide an algorithm that outputs a classifier that minimizes the risk of any group (of sufficient size) while at the same time being Pareto efficient. They do this by first showing t... | [
6,
6,
-1,
-1,
-1,
-1,
6
] | [
4,
4,
-1,
-1,
-1,
-1,
4
] | [
"iclr_2021_MMXhHXbNsa-",
"iclr_2021_MMXhHXbNsa-",
"VXWWlGm45rD",
"jl9JKpJoh-m",
"PmUtZQI4EnM",
"iclr_2021_MMXhHXbNsa-",
"iclr_2021_MMXhHXbNsa-"
] |
iclr_2021_Oj2hGyJwhwX | SelfNorm and CrossNorm for Out-of-Distribution Robustness | Normalization techniques are crucial in stabilizing and accelerating the training of deep neural networks. However, they are mainly designed for the independent and identically distributed (IID) data, not satisfying many real-world out-of-distribution (OOD) situations. Unlike most previous works, this paper presents tw... | withdrawn-rejected-submissions |
This paper proposes two mechanisms, SelfNorm (used during training and inference) leveraging an attention-based recalibration of mean and standard deviation for instance normalization, and CrossNorm which performs cross-channel swapping of mean/stdev. Is is shown that the combination (often combined with AugMix) perf... | train | [
"uqa9HExYji7",
"1-ZG0uBV5b",
"cUsb2hWuonQ",
"Ncpl20ootwP",
"Zm4zOAsKacQ",
"uXXJ9rN3i5B",
"AutVKoKBfhA",
"vQfsAjvR48F",
"DaNjwapGtn",
"8_MpViQfCVF",
"XZdvHHyWizO"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"1. Summary\n\nThe paper presents two new methods to improve corruption robustness and domain generalization: SelfNorm, a way to adapt style information during inference, and CrossNorm, a simple data augmentation technique diversifying image style in feature space. Both methods are tested on Cifar10/100-C, ImageNet... | [
7,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
7,
6
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
3
] | [
"iclr_2021_Oj2hGyJwhwX",
"Ncpl20ootwP",
"iclr_2021_Oj2hGyJwhwX",
"DaNjwapGtn",
"8_MpViQfCVF",
"uqa9HExYji7",
"uXXJ9rN3i5B",
"XZdvHHyWizO",
"iclr_2021_Oj2hGyJwhwX",
"iclr_2021_Oj2hGyJwhwX",
"iclr_2021_Oj2hGyJwhwX"
] |
iclr_2021_xCy9thPPTb_ | The Compact Support Neural Network | Neural networks are popular and useful in many fields, but they have the problem of giving high confidence responses for examples that are away from the training data. This results in the neural networks being very confident while making gross mistakes, thus limiting their reliability for safety critical applications ... | withdrawn-rejected-submissions | The paper presents an approach that supports better performance when out of distribution cases occur, by letting neurons be of only compact support and thus if the input is out of distribution (OOD).
Pros:
- The proposed strategy is interesting and may be useful.
Cons:
- The choice of the parameter alpha, whose valu... | train | [
"-2rBJXXeoEV",
"-Pzyt40XgEk",
"XZZyoUGU9e9",
"tHCMym2QqMZ",
"lctsUOy9Xat",
"vAmH94p6Oox",
"ToOnNDpVEgt",
"1CKd0aNFLhw",
"CfzclQz_ZG6",
"2KNYc_M5nW",
"q-9cVwJFZVC"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"The paper presents an approach that supports better performance when out of distribution cases occur. It does so by letting neurons be of only compact support and thus if the input is out of distribution (OOD) it is expected to be outside that support and therefore the output will be zero. This is used to detect O... | [
6,
-1,
5,
-1,
-1,
-1,
-1,
-1,
-1,
5,
6
] | [
3,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
3,
4
] | [
"iclr_2021_xCy9thPPTb_",
"lctsUOy9Xat",
"iclr_2021_xCy9thPPTb_",
"2KNYc_M5nW",
"CfzclQz_ZG6",
"XZZyoUGU9e9",
"-2rBJXXeoEV",
"q-9cVwJFZVC",
"vAmH94p6Oox",
"iclr_2021_xCy9thPPTb_",
"iclr_2021_xCy9thPPTb_"
] |
iclr_2021_PoP96DrBHnl | Gradient descent temporal difference-difference learning | Off-policy algorithms, in which a behavior policy differs from the target policy and is used to gain experience for learning, have proven to be of great practical value in reinforcement learning. However, even for simple convex problems such as linear value function approximation, these algorithms are not guaranteed to... | withdrawn-rejected-submissions | This paper introduces an simple but potentially effective off-policy TD algorithm.
Overall, the reviewers felt the work was incomplete and not yet ready for publication. The all recognized the authors made significant updates to the paper, but serious issues remain with the empirical work: studying the impact of the ... | train | [
"CmnOLsrs213",
"vT77JFVg6iw",
"lv6Rx7TUwy-",
"23cMPTzhR6f",
"OPohSeaTNLp",
"MEKtyJlMwRE",
"Fl_uRaaWied",
"lKmDvQjMqWM",
"Ak-RXadRr1d",
"pm1pMISp1Yw",
"k1SSLw_I7tf"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author"
] | [
"### Summary of Contributions\n\nThe paper proposes the gradient descent TD difference learning (GDD) algorithm which adds a term to the MSPBE objective to constrain how quickly a value function can change. They argue that their approach has a quicker convergence rate, and empirically demonstrate in several example... | [
5,
5,
5,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
4,
4,
4,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
"iclr_2021_PoP96DrBHnl",
"iclr_2021_PoP96DrBHnl",
"iclr_2021_PoP96DrBHnl",
"iclr_2021_PoP96DrBHnl",
"Fl_uRaaWied",
"lKmDvQjMqWM",
"Ak-RXadRr1d",
"lv6Rx7TUwy-",
"23cMPTzhR6f",
"vT77JFVg6iw",
"CmnOLsrs213"
] |
iclr_2021_bJLHjvYV1Cu | Optimizing Loss Functions Through Multivariate Taylor Polynomial Parameterization | Metalearning of deep neural network (DNN) architectures and hyperparameters has become an increasingly important area of research. Loss functions are a type of metaknowledge that is crucial to effective training of DNNs, however, their potential role in metalearning has not yet been fully explored. Whereas early work f... | withdrawn-rejected-submissions | Pros: Reviewers generally agreed the paper was well written and is easy to follow. The goal of learning loss functions also seems quite promising.
Cons: There were concerns about whether credit for experimental performance was attributable to the core algorithm+functional form presented in the paper. There was also so... | train | [
"QTtNGg2CeQI",
"2tZODRBOUC6",
"FycrBI3EX2i",
"9Ib7z82B0Kx",
"IAGZYwJU5aw",
"cLgkTS0IIf-",
"XtAUky9cDNe",
"NRs3f0TXoNx",
"-i14pOGorf"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"Thank you for the revisions. Appendix F contains some really striking results! Order 3 results in 5% accuracy drop, and even order 5 has 3% accuracy loss. That is really surprising to me but useful to know.\n\nI maintain my rating as I would have liked to see a more rigorous analysis of the learning rate effects (... | [
-1,
-1,
-1,
-1,
-1,
5,
5,
6,
6
] | [
-1,
-1,
-1,
-1,
-1,
4,
4,
2,
4
] | [
"9Ib7z82B0Kx",
"cLgkTS0IIf-",
"XtAUky9cDNe",
"NRs3f0TXoNx",
"-i14pOGorf",
"iclr_2021_bJLHjvYV1Cu",
"iclr_2021_bJLHjvYV1Cu",
"iclr_2021_bJLHjvYV1Cu",
"iclr_2021_bJLHjvYV1Cu"
] |
iclr_2021_yoVo1fThmS1 | Novelty Detection via Robust Variational Autoencoding | We propose a new method for novelty detection that can tolerate high corruption of the training points, whereas previous works assumed either no or very low corruption. Our method trains a robust variational autoencoder (VAE), which aims to generate a model for the uncorrupted training points. To gain robustness to hig... | withdrawn-rejected-submissions | The paper proposes a novelty detection method when training data is itself noisy. A VAE-based approach is developed that promotes robustness of the VAE. The paper assumes that the encoder a two-component Gaussian mixture distribution, individual components denoting inliers and outliers.
The paper hopes that the poster... | train | [
"rPJaUq2FdB1",
"CIHJlzgEIyp",
"fyIV3ZifVxz",
"wcv4c3wkw0r",
"deGuK_mtRt8",
"nQ6dJNIv5V",
"unuicR9lA94",
"e1MGH43ypvg",
"4NROKIwjU_P",
"3sWWZEWsuAH",
"LDBG5kJX32i",
"iQsvxb6Ugbp",
"Y0CC9uT3vL2",
"n1qRsuWsQ-e"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"This paper proposes a robust novelty detection method (\"MAW\") to model the distribution of the training data in the presence of high fraction (corruption ratios up to 30\\%) of outliers. The method add new features to the variational autoencoder (VAE), to detect and isolate the outlier so that the learned distri... | [
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
6,
5
] | [
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3,
4
] | [
"iclr_2021_yoVo1fThmS1",
"iclr_2021_yoVo1fThmS1",
"rPJaUq2FdB1",
"n1qRsuWsQ-e",
"n1qRsuWsQ-e",
"n1qRsuWsQ-e",
"Y0CC9uT3vL2",
"Y0CC9uT3vL2",
"iQsvxb6Ugbp",
"iQsvxb6Ugbp",
"iQsvxb6Ugbp",
"iclr_2021_yoVo1fThmS1",
"iclr_2021_yoVo1fThmS1",
"iclr_2021_yoVo1fThmS1"
] |
iclr_2021_8znruLfUZnT | Frequency Regularized Deep Convolutional Dictionary Learning and Application to Blind Denoising | Sparse representation via a learned dictionary is a powerful prior for natural images. In recent years, unrolled sparse coding algorithms (e.g. LISTA) have proven to be useful for constructing interpretable deep-learning networks that perform on par with state-of-the-art models on image-restoration tasks. In this study... | withdrawn-rejected-submissions | The paper proposes a deep learning approach to blind image denoising based on deep unrolling. In particular, the proposed network is derived from convolutional sparse coding algorithms, which are unrolled, untied across layers and learned from data. The paper proposes a frequency domain regularization scheme in which t... | val | [
"9kBDXCbX8g8",
"zwdRg7Ifv-R",
"jpkaLvGGfKg",
"-VtbbWc50C",
"YOiJgsusmrf",
"y8ty_4C6Md"
] | [
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"The authors would like to thank the reviewer for taking the time and careful reading of the manuscript and providing insightful comments that help improving the manuscript. Please see as follows how we have addressed the comments.\n\n1. We have improved the introduction to better present the motivation of the pape... | [
-1,
-1,
-1,
4,
3,
4
] | [
-1,
-1,
-1,
5,
3,
4
] | [
"-VtbbWc50C",
"YOiJgsusmrf",
"y8ty_4C6Md",
"iclr_2021_8znruLfUZnT",
"iclr_2021_8znruLfUZnT",
"iclr_2021_8znruLfUZnT"
] |
iclr_2021_dqyK5RKMaW4 | LEARNED HARDWARE/SOFTWARE CO-DESIGN OF NEURAL ACCELERATORS | The use of deep learning has grown at an exponential rate, giving rise to numerous specialized hardware and software systems for deep learning. Because the design space of deep learning software stacks and hardware accelerators is diverse and vast, prior work considers software optimizations separately from hardware ar... | withdrawn-rejected-submissions | This paper considers the problem of hardware and software co-design for neural accelerators. Specifically, it looks at hardware and the software compiler that maps DNN to hardware. It employs Bayesian Optimization (BO) to perform joint search over hardware and software design parameters in an alternating manner. To han... | train | [
"QlXOpKGkR-S",
"n1LnGAEBfUr",
"Vbo7Stxk40",
"EB2SsYdlTwb",
"vv5dClHmOSM",
"Fol3vCUqrR0",
"-X7_uOHJKq0",
"sTsaZVvg_2",
"_45302Z7jV",
"Dyc86-IiDs",
"l6Ayv8YqW07",
"vNVUZ2Ya-Td",
"e1BILumLSro",
"GJ1y-WjYemv",
"DEPnOZ2IuAZ",
"BzmjL_12UtP",
"OXk69LIQ5lA"
] | [
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"We find that our approach improves EDP by 24% over Diannao on DQN. More results (other workloads and search algorithms) have not finished in time for the discussion period, but we're working on them and will include results in the next version of the paper.",
"During the rebuttal period we find that our approach... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
4,
7,
6
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
2,
5,
4,
3
] | [
"vNVUZ2Ya-Td",
"Dyc86-IiDs",
"EB2SsYdlTwb",
"l6Ayv8YqW07",
"BzmjL_12UtP",
"OXk69LIQ5lA",
"DEPnOZ2IuAZ",
"GJ1y-WjYemv",
"OXk69LIQ5lA",
"DEPnOZ2IuAZ",
"GJ1y-WjYemv",
"BzmjL_12UtP",
"DEPnOZ2IuAZ",
"iclr_2021_dqyK5RKMaW4",
"iclr_2021_dqyK5RKMaW4",
"iclr_2021_dqyK5RKMaW4",
"iclr_2021_dqyK... |
iclr_2021_ysti0DEWTSo | Is deeper better? It depends on locality of relevant features | It has been recognized that a heavily overparameterized artificial neural network exhibits surprisingly good generalization performance in various machine-learning tasks. Recent theoretical studies have made attempts to unveil the mystery of the overparameterization. In most of those previous works, the overparameteriz... | withdrawn-rejected-submissions | Reviewers found the construction is very clever and the empirical results are interesting. However, a more thorough theoretical explanation is needed for acceptance. | train | [
"BfV-ovvpCR",
"C-NriC51j2",
"FauptPX7Al",
"9EjVujeDgVT",
"1oh9MWU2Tf_",
"bP87TqruHwL",
"BXxP3TcpqY7",
"t_ioJsQf1dS",
"wTNh0pyt-cG",
"iYBWx-4Xeyi",
"5otdIN6fMW"
] | [
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"In the revised paper, we added experimental results for the mean-square loss and commented on how our experimental results reflect the results in the references mentioned by the reviewer. \n\nOur main result does not change, but we found quantitative difference in test errors. Overall, the cross-entropy loss yield... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
7,
6,
4,
4
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3,
2,
4
] | [
"FauptPX7Al",
"BXxP3TcpqY7",
"5otdIN6fMW",
"wTNh0pyt-cG",
"t_ioJsQf1dS",
"iYBWx-4Xeyi",
"iclr_2021_ysti0DEWTSo",
"iclr_2021_ysti0DEWTSo",
"iclr_2021_ysti0DEWTSo",
"iclr_2021_ysti0DEWTSo",
"iclr_2021_ysti0DEWTSo"
] |
iclr_2021_uFBBOJ7xnu | Learning representations from temporally smooth data | Events in the real world are correlated across nearby points in time, and we must learn from this temporally “smooth” data. However, when neural networks are trained to categorize or reconstruct single items, the common practice is to randomize the order of training items. What are the effects of temporally smooth trai... | withdrawn-rejected-submissions | It appears that this paper can benefit from additional detail and work before it becomes a stronger publication that is more convincing. The authors have done an impressive job responding to the reviewers and updating their paper, and multiple reviewers raised their score consequently. However, while multiple reviewers... | train | [
"Trj9MAe6tHJ",
"Fp9yGugJi2",
"J-SRLE1Wwj",
"Mzs5RWW8Z16",
"D88TTNg6TZi",
"cG6cq-v6UxV",
"3LVyIcPAnsf",
"uV-bNYRcJ89",
"3CWdoTiRiCw",
"fuNN1JUpnwD",
"ashnGPCeyi",
"4nHeM-AjTAT",
"Udc0ZeeAWuO",
"o9q21zjVoE",
"h9Q_oBhXH5",
"AG0YRD1pMXs",
"fOGtZ09shis",
"G_Ahn9VNEdg",
"M0gB0LjcY40",
... | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"Temporal smoothness is a recurring feature of real-world data that has been unaccounted for when training neural networks. Much of the random sampling in training neural networks is done to remove the temporal correlations originally present when the data is collected. This work aims to propose a method to train o... | [
6,
4,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6
] | [
4,
4,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3
] | [
"iclr_2021_uFBBOJ7xnu",
"iclr_2021_uFBBOJ7xnu",
"iclr_2021_uFBBOJ7xnu",
"D88TTNg6TZi",
"OzrpY6_Nzr8",
"3LVyIcPAnsf",
"uV-bNYRcJ89",
"3CWdoTiRiCw",
"fuNN1JUpnwD",
"ashnGPCeyi",
"Fp9yGugJi2",
"Udc0ZeeAWuO",
"o9q21zjVoE",
"Trj9MAe6tHJ",
"AG0YRD1pMXs",
"fOGtZ09shis",
"G_Ahn9VNEdg",
"J-... |
iclr_2021_gdtGg1hCK2 | On Low Rank Directed Acyclic Graphs and Causal Structure Learning | Despite several important advances in recent years, learning causal structures represented by directed acyclic graphs (DAGs) remains a challenging task in high dimensional settings when the graphs to be learned are not sparse. In this paper, we propose to exploit a low rank assumption regarding the (weighted) adjacenc... | withdrawn-rejected-submissions | This paper studies the low-rank properties of DAG models, and illustrates through proof-of-concept how low-rank-ness can be exploited in structure learning of DAGs. After a lengthy discussion amongst the reviewers, it became clear that although there are some interesting ideas here, there is not enough enthusiasm for t... | train | [
"2G4gJTC1fmm",
"qwe5_DgHSFJ",
"vS5KA6HJqRN",
"l0jRnhN39Ty",
"_-4lnBW690j",
"YWMOGTCw4D",
"czFhBuEA0B0",
"Bi76cG8tY6",
"NmhdhiYA-j",
"9Tyfee7MSq",
"UdL0z8zFj8u",
"BnTxkTse2qh",
"kVruyfkWK1I",
"52hC-S8MdUx"
] | [
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"This paper attempts to exploit the low-rankness of the adjacency matrix of the DAG in Bayesian network structure learning. The overall framework is similar to NOTEARS, except that the adjacency matrix W is decomposed into low rank components W = UV'. To justify the approach, the paper also includes lower and upper... | [
6,
5,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5
] | [
3,
4,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4
] | [
"iclr_2021_gdtGg1hCK2",
"iclr_2021_gdtGg1hCK2",
"l0jRnhN39Ty",
"iclr_2021_gdtGg1hCK2",
"52hC-S8MdUx",
"l0jRnhN39Ty",
"52hC-S8MdUx",
"_-4lnBW690j",
"9Tyfee7MSq",
"52hC-S8MdUx",
"2G4gJTC1fmm",
"qwe5_DgHSFJ",
"iclr_2021_gdtGg1hCK2",
"iclr_2021_gdtGg1hCK2"
] |
iclr_2021_cuDFRRANJ-5 | Formalizing Generalization and Robustness of Neural Networks to Weight Perturbations | Studying the sensitivity of weight perturbation in neural networks and its impacts on model performance, including generalization and robustness, is an active research topic due to its implications on a wide range of machine learning tasks such as model compression, generalization gap assessment, and adversarial attack... | withdrawn-rejected-submissions | The authors study empirically and theoretically the behavior of neural networks under $l_\infty$-perturbations on the weight matrix.
For this purpose they first derive bounds on the logit-layer of the neural networks under perturbuations of a single or all layers. Then they propose to merge this bound (which depends on... | test | [
"stYefE1yobu",
"A0TH4PmP4G_",
"TKFkJ82Stad",
"BdzGfNAWjq8",
"BaI8sX76ZG",
"AEKL9nrfXzL",
"3w_p5ueOKg2",
"6y-JgbSuetq",
"Jh1P9C4GQKF",
"gMRvMnZTZpC",
"4kp5XK9anV",
"ObrtE1T1Hhz",
"Tnn0-2Vp-i-",
"jsiVL-VyfcW",
"pqvO8VO7lg0",
"nxyvMvdFcZ5",
"pDcTZEQ8qGG",
"XMxhzfNRKne",
"kaULOPuBuYP... | [
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
... | [
"The paper investigates the effects of weight perturbations on the output margin for multiclass classifcation problems. The paper shows that robustness to weight perturbations can be bounded using the (1,\\infty)-norm of the weight matrices. The paper then suggests that a low (1,\\infty)-norm of the weight matrices... | [
7,
-1,
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
7
] | [
4,
-1,
-1,
-1,
-1,
-1,
3,
-1,
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4
] | [
"iclr_2021_cuDFRRANJ-5",
"BdzGfNAWjq8",
"BdzGfNAWjq8",
"BaI8sX76ZG",
"nxyvMvdFcZ5",
"6y-JgbSuetq",
"iclr_2021_cuDFRRANJ-5",
"Tnn0-2Vp-i-",
"iclr_2021_cuDFRRANJ-5",
"4kp5XK9anV",
"pqvO8VO7lg0",
"iclr_2021_cuDFRRANJ-5",
"3w_p5ueOKg2",
"stYefE1yobu",
"ObrtE1T1Hhz",
"0uQPAbVCYZk",
"iclr_... |
iclr_2021_jnMjOctlfbZ | ATOM3D: Tasks On Molecules in Three Dimensions | While a variety of methods have been developed for predicting molecular properties, deep learning networks that operate directly on three-dimensional molecular structure have recently demonstrated particular promise. In this work we present ATOM3D, a collection of both novel and existing datasets spanning several key c... | withdrawn-rejected-submissions | For many problems such as ligand-protein binding, quantitative structure activity prediction (QSAR), predicting protein function from structure, etc., the 3D geometry of the molecules is of great importance. One way to represent this is simply to assign locations to all atoms in 3-dimensional space. If using graph co... | train | [
"1Ik1rh-svhX",
"MD8zWd6pgIs",
"Z26YvaHPKr3",
"Zmr1MYW0Hv",
"zBFZgi3GVM",
"J7X5b2qutD",
"45Bs7SCwosF",
"BqgWH0KG3z_",
"9tx-1rgzM4",
"gJQffi9YHPF",
"j8S72pGxZY"
] | [
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"On our LBA dataset, DeepAffinity [3] achieves the following performance:\n- RMSE: 1.893 +/- 0.119\n- Pearson R: 0.415 +/- 0.088\n- Spearman R: 0.426 +/- 0.097\n\nThis is worse than our current non-3D baseline. Compared to 3D methods, it is only competitive with the GNN on the RMSD (though not on the Pearson ... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
6,
5
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3,
3
] | [
"Zmr1MYW0Hv",
"Zmr1MYW0Hv",
"iclr_2021_jnMjOctlfbZ",
"j8S72pGxZY",
"J7X5b2qutD",
"gJQffi9YHPF",
"BqgWH0KG3z_",
"9tx-1rgzM4",
"iclr_2021_jnMjOctlfbZ",
"iclr_2021_jnMjOctlfbZ",
"iclr_2021_jnMjOctlfbZ"
] |
iclr_2021_IOqr2ZyXHz1 | Continual Lifelong Causal Effect Inference with Real World Evidence | The era of real world evidence has witnessed an increasing availability of observational data, which much facilitates the development of causal effect inference. Although significant advances have been made to overcome the challenges in causal effect estimation, such as missing counterfactual outcomes and selection bia... | withdrawn-rejected-submissions | The authors consider the problem of causal inference from multiple conditionally ignorable models that yield different observed data distributions. This problem is distinct from transportability (which assumes some types of causal invariance across domains, and aims to move causal conclusioned learned in one context t... | test | [
"F_TxBmjSRPM",
"q_ohX69xN8",
"JQaH7XjqCK",
"MwSJSPFOzNr",
"Lf_YrMRoGvy",
"TKYXsiO21WK",
"NLc9QOkmA51",
"DKp97tznGi",
"zPjJ9NFzj7",
"_youfDlCUE0",
"IyHObRVFzC",
"YeMiIpcp4Wo",
"8F-ugub_lDS",
"1j9Tuxyd_cf",
"NZZBckgfEwe"
] | [
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"I really appreciate your prompt reply. Due to the rebuttal time limit, we cannot provide solid results about the complexity or cost for each experiment. We will make constant efforts to provide more details about experiments maybe in the later arXiv version. Thanks for your suggestions as a practitioner. In the in... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
2,
3,
4,
4
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
4,
5
] | [
"JQaH7XjqCK",
"MwSJSPFOzNr",
"zPjJ9NFzj7",
"Lf_YrMRoGvy",
"YeMiIpcp4Wo",
"8F-ugub_lDS",
"DKp97tznGi",
"1j9Tuxyd_cf",
"_youfDlCUE0",
"NZZBckgfEwe",
"iclr_2021_IOqr2ZyXHz1",
"iclr_2021_IOqr2ZyXHz1",
"iclr_2021_IOqr2ZyXHz1",
"iclr_2021_IOqr2ZyXHz1",
"iclr_2021_IOqr2ZyXHz1"
] |
iclr_2021_xOBMyvoMQw8 | Improving Sampling Accuracy of Stochastic Gradient MCMC Methods via Non-uniform Subsampling of Gradients | Common Stochastic Gradient MCMC methods approximate gradients by stochastic ones via uniformly subsampled data points. A non-uniform subsampling scheme, however, can reduce the variance introduced by the stochastic approximation and make the sampling of a target distribution more accurate. For this purpose, an exponent... | withdrawn-rejected-submissions | This paper describes a non-uniformly weighted version of SGMCMC, combining aspects of SG methods and importance sampling. The idea is interesting and novel, but unfortunately the authors have not made a compelling case for the resulting algorithm being a practical addition to the literature. The experimental analysis i... | train | [
"OsiRjzDvdmP",
"SzlC0bFqXuy",
"Xm_f00GC2vx",
"OdAEuS77JlV",
"SJ4_LiZHsXh",
"q0sqwzv3bdo"
] | [
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"##########################################################################\nSummary:\n\nThe paper proposes an alternative to the uniform sampling scheme used for constructing mini-batches in stochastic gradient sampling algorithms. The proposed scheme, called Exponentially Weighted Stochastic Gradient (EWSG), is d... | [
6,
-1,
-1,
-1,
5,
4
] | [
3,
-1,
-1,
-1,
4,
3
] | [
"iclr_2021_xOBMyvoMQw8",
"OsiRjzDvdmP",
"q0sqwzv3bdo",
"SJ4_LiZHsXh",
"iclr_2021_xOBMyvoMQw8",
"iclr_2021_xOBMyvoMQw8"
] |
iclr_2021_Mh1Abj33qI | Data-driven Learning of Geometric Scattering Networks | Many popular graph neural network (GNN) architectures, which are often considered as the current state of the art, rely on encoding graph structure via smoothness or similarity between neighbors. While this approach performs well on a surprising number of standard benchmarks, the efficacy of such models does not transl... | withdrawn-rejected-submissions | This paper proposes a simple yet powerful generalisation of graph scattering transforms that allows a flexible scale dilation structure, retaining the stability guarantees of dyadic transforms. Experiments with strong empirical performance are reported on a variety of biochemical tasks.
Reviewers acknowledged the soun... | train | [
"U0hn8n_iCnA",
"JHpZO3uubL3",
"KBnUCJgnkXX",
"tppZ1rXDuoG",
"z5C-FW1Ozla",
"ITTqO9Txcf",
"RGl2tHxP9n",
"zB7kUdSN89",
"n8lVTGWqaXp",
"iAUGk30CM-I",
"-eDLx_Pi-M",
"B1yj1W7Kogw",
"Pwoy85H4Q4J",
"2zyR5NHSBEX",
"KjHZ8ijc6w0"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"Summary:\n\nThis paper proposes a parameterization of the scatter transform on graphs and builds graph neural networks based on this parameterization. Authors also demonstrate that this scatter transform could theoretically lead to stable hidden representations of GNNs. Experimental results on biochemical datasets... | [
6,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
6
] | [
4,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
3
] | [
"iclr_2021_Mh1Abj33qI",
"iclr_2021_Mh1Abj33qI",
"JHpZO3uubL3",
"2zyR5NHSBEX",
"KjHZ8ijc6w0",
"U0hn8n_iCnA",
"iclr_2021_Mh1Abj33qI",
"n8lVTGWqaXp",
"U0hn8n_iCnA",
"-eDLx_Pi-M",
"KjHZ8ijc6w0",
"2zyR5NHSBEX",
"JHpZO3uubL3",
"iclr_2021_Mh1Abj33qI",
"iclr_2021_Mh1Abj33qI"
] |
iclr_2021_P63SQE0fVa | ScheduleNet: Learn to Solve MinMax mTSP Using Reinforcement Learning with Delayed Reward | Combinatorial Optimization (CO) problems are theoretically challenging yet crucial in practice. Numerous works used Reinforcement Learning (RL) to tackle these CO problems. As current approaches mainly focus on single-worker CO problems such as the famous Travelling Salesman Problem (TSP), we focus on more practical ex... | withdrawn-rejected-submissions | This paper proposes a deep reinforcement learning approach for solving minimax multiple TSP problem. Their main algorithmic contribution is to propose a specialized graph neural network to parameterize the policy and used a clipped idea to stabilize the training. Unfortunately, the reviewers remain to be unconvinced by... | val | [
"gOu69LaeVD-",
"wHDCNRZBDyA",
"v1peR8EJf-J",
"VQm326aeCwO",
"llAkcugWL4U",
"4FYRO_fjFKE",
"fPpcno89ZZX",
"juwszY8jV9Y",
"d0mcg8kxVVa",
"7tweN75oHy7",
"r0LDSStHGC"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"The authors propose an RL framework, called ScheduleNet trained by clipped REINFORCE, for minmax multiple traveling salesman problem (minimax mTSP), which uses a clipping idea to stabilize learning process as PPO does. The authors empirically show the feasibility of the proposed framework.\n\n- Unfortunately, the ... | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
4,
5
] | [
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
5,
5
] | [
"iclr_2021_P63SQE0fVa",
"v1peR8EJf-J",
"7tweN75oHy7",
"4FYRO_fjFKE",
"gOu69LaeVD-",
"d0mcg8kxVVa",
"r0LDSStHGC",
"llAkcugWL4U",
"iclr_2021_P63SQE0fVa",
"iclr_2021_P63SQE0fVa",
"iclr_2021_P63SQE0fVa"
] |
iclr_2021_SzjyTIc5qMP | Information Lattice Learning | Information Lattice Learning (ILL) is a general framework to learn decomposed representations, called rules, of a signal such as an image or a probability distribution. Each rule is a coarsened signal used to gain some human-interpretable insight into what might govern the nature of the original signal. To summarize th... | withdrawn-rejected-submissions | This paper has been evaluated by four expert reviewers resulting in two rejections one marginal score and one acceptance recommendation. The authors provided rebuttals to the critiques, but they did not sway the reviewers' assessments. The prevailing impression is that the work is interesting but perhaps not yet mature... | test | [
"FWmOYiJgGlq",
"XDw3UdboOjF",
"3anVv2KBFQ",
"42TGPAsN-xB",
"FiNchLEfNWW",
"_yORo1ZesN",
"T34fFJy_uv",
"9gCOdmZ0GA",
"fbmbP-XaE21",
"afIQKoSffg",
"me5zCO-oyK8"
] | [
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"We are grateful that all the central ideas from our proposed framework regarding goals and motivations as well as algorithmic developments were clearly understood.\n\n**Regarding interpretability.** We apologize that interpretability could have been more clearly presented, and have revised the paper accordingly. W... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
6,
7,
4
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3,
3,
4
] | [
"fbmbP-XaE21",
"fbmbP-XaE21",
"afIQKoSffg",
"me5zCO-oyK8",
"9gCOdmZ0GA",
"9gCOdmZ0GA",
"iclr_2021_SzjyTIc5qMP",
"iclr_2021_SzjyTIc5qMP",
"iclr_2021_SzjyTIc5qMP",
"iclr_2021_SzjyTIc5qMP",
"iclr_2021_SzjyTIc5qMP"
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.