Split (1)

train · 15k rows

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_2022_Ee2ugKwgvyy	Graph Information Matters: Understanding Graph Filters from Interaction Probability	Graph Neural Networks (GNNs) have received extensive affirmation for their promising performance in graph learning problems. Despite their various neural architectures, most are intrinsically graph filters that provide theoretical foundations for model explanations. In particular, low-pass filters show superiority in ...	Reject	In this paper, in order to theoretically investigate the relationship between graph structure and labels in GNNs, interaction probabity and frequency indicators are introduced and analyzed, and a new family of GNNs with multiple filters is proposed based on the insights from the theoretical analysis, In the discussion,...	train	[ "zXk-1X1ZEJm", "VO2NNu2pRxR", "ge1uvR93kZZ", "lXBAq_LgjFx", "GNfoc53n6dP", "T2xI3RkuDqG", "MSLTBUAbFbF", "nkajjRvp_b2", "eTY4sU-h-c3", "PlynZk9-Xx", "BLs-k2pH87", "u9IVPY7BHD", "cp9Bw4SOBK4", "65ZhM4X-ig-", "VxIT1al8Cwn", "9CzHlfLQb5", "7cXxhmAHPBU" ]	[ "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer" ]	[ " Thank you for the reply and raising the score, but most of your concerns have been clarified in the rebuttal, and we feel that the two major reasons for your low scores do not have enough evidence. \n\n1. Presentation: First, we do not understand why the concerns are clarified from \"reading other responses\", be...	[ -1, 3, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5 ]	[ -1, 4, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4 ]	[ "ge1uvR93kZZ", "iclr_2022_Ee2ugKwgvyy", "9CzHlfLQb5", "iclr_2022_Ee2ugKwgvyy", "nkajjRvp_b2", "MSLTBUAbFbF", "eTY4sU-h-c3", "PlynZk9-Xx", "u9IVPY7BHD", "cp9Bw4SOBK4", "iclr_2022_Ee2ugKwgvyy", "lXBAq_LgjFx", "lXBAq_LgjFx", "7cXxhmAHPBU", "7cXxhmAHPBU", "VO2NNu2pRxR", "iclr_2022_Ee2ugK...
iclr_2022_rS9t6WH34p	Decomposing 3D Scenes into Objects via Unsupervised Volume Segmentation	We present ObSuRF, a method which turns a single image of a scene into a 3D model represented as a set of Neural Radiance Fields (NeRFs), with each NeRF corresponding to a different object. A single forward pass of an encoder network outputs a set of latent vectors describing the objects in the scene. These vectors are...	Reject	The paper develops a method for decomposing 3D scenes into objects by coupling NeRF decoders to representations produced by a slot-based encoder. After the discussion phase, reviewer ratings are mixed with three on either side of above/below threshold, and one higher (but low confidence) accept score. Drawbacks inclu...	val	[ "2i8dtyHN6t", "hO41BflTTMc", "R4zYkPB6spN", "pp-X3GiiQ6K", "FIoB3wSaDuw", "x-SnFTN_syD", "fcIT2jYJ8MB", "CSksmytokGW", "eqTeJxmgHEw", "xO6FtnVm9p-", "o0oaROTRLV0", "2ntJssUwByE" ]	[ "official_reviewer", "author", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ "The authors presented a method to infer a neural radiance field per object in a scene from a single input view. \nuses Object-centric learning with slot attention\n\nThe proposed architecture takes the shape of an autoencoder. The encoder side takes a single image and pose as inputs and generates N latent codes f...	[ 6, -1, -1, -1, 8, -1, -1, -1, -1, -1, 6, 5 ]	[ 3, -1, -1, -1, 2, -1, -1, -1, -1, -1, 3, 3 ]	[ "iclr_2022_rS9t6WH34p", "o0oaROTRLV0", "2ntJssUwByE", "2i8dtyHN6t", "iclr_2022_rS9t6WH34p", "iclr_2022_rS9t6WH34p", "2ntJssUwByE", "FIoB3wSaDuw", "2i8dtyHN6t", "o0oaROTRLV0", "iclr_2022_rS9t6WH34p", "iclr_2022_rS9t6WH34p" ]
iclr_2022_86sEVRfeGYS	Continual Backprop: Stochastic Gradient Descent with Persistent Randomness	The Backprop algorithm for learning in neural networks utilizes two mechanisms: first, stochastic gradient descent and second, initialization with small random weights, where the latter is essential to the effectiveness of the former. We show that in continual learning setups, Backprop performs well initially, but over...	Reject	The paper studies an important newly identified problem in continual learning of rapid adaptation, and proposes the use of a generate-and-test method to continually inject random features alongside SGD, enabling better learning on non-stationary data streams. Unfortunately the paper remained borderline in the discussio...	test	[ "3FVxxURkAZ5", "07iB3rwQjgE", "rCb0o_Fzaa4", "BG9j8Tl9EMC", "Ve0Omw55q9d", "IYS6DqvgpT", "_XluXbKxK6g", "cN9LKrkEHpA", "2TLrIVKjY2w", "WphkvDp4M7B", "j0JbuJwn8Al", "mlMvODZ7x0s", "5n7CigYHgi0" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ "This paper demonstrates and proposes a solution for a new problem in continual learning which is the inverse of catastrophic forgetting. Compared to prior work, they study problems where the data distribution changes much more rapidly. They demonstrate that backpropagation based optimization loses its ability to a...	[ 6, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, 5, 3 ]	[ 4, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, 5, 4 ]	[ "iclr_2022_86sEVRfeGYS", "Ve0Omw55q9d", "iclr_2022_86sEVRfeGYS", "_XluXbKxK6g", "IYS6DqvgpT", "WphkvDp4M7B", "cN9LKrkEHpA", "5n7CigYHgi0", "mlMvODZ7x0s", "3FVxxURkAZ5", "rCb0o_Fzaa4", "iclr_2022_86sEVRfeGYS", "iclr_2022_86sEVRfeGYS" ]
iclr_2022_3M3t3tUbA2Y	DreamerPro: Reconstruction-Free Model-Based Reinforcement Learning with Prototypical Representations	In model-based reinforcement learning (MBRL) such as Dreamer, the approaches based on observation reconstruction often fail to discard task-irrelevant details, thus struggling to handle visual distractions or generalize to unseen distractions. To address this issue, previous work has proposed to contrastively learn the...	Reject	The authors propose an alteration to Dreamer that incorporates a swav-like objective. The reviewers raised a number of issues with the paper, overall arguing for rejection. In particular, the reviewers felt that the work was not well motivated, weak performance, that a number of baselines were missing, and a lack of an...	train	[ "3df2--VrMUS", "79pMk0jFc0Q", "tXRW6R683c", "ZP9PcI9KaF", "q2ukeq0f147", "Zv3-2aSU1Oe", "4J99bNmzkOZ", "hslcrqXozR", "OW3U8v49iBp", "4jXTAIkXgfr", "rSNrwSW4Wz-", "1fArBQv6hOD", "KT_zI1_r_yM", "z_zxf1Kv50g" ]	[ "official_reviewer", "official_reviewer", "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ " I thank the authors for the additional results on Atari and the clarification. While the proposed method is reasonably motivated, it requires a strong experimental results to be justified. DreamerPro is able to be comparable with Dreamer only on a subset of the Atari environments tested, and it remains unclear to...	[ -1, 5, -1, -1, -1, 6, -1, -1, -1, -1, -1, -1, 3, 5 ]	[ -1, 3, -1, -1, -1, 4, -1, -1, -1, -1, -1, -1, 4, 4 ]	[ "hslcrqXozR", "iclr_2022_3M3t3tUbA2Y", "ZP9PcI9KaF", "rSNrwSW4Wz-", "4jXTAIkXgfr", "iclr_2022_3M3t3tUbA2Y", "iclr_2022_3M3t3tUbA2Y", "z_zxf1Kv50g", "Zv3-2aSU1Oe", "79pMk0jFc0Q", "KT_zI1_r_yM", "iclr_2022_3M3t3tUbA2Y", "iclr_2022_3M3t3tUbA2Y", "iclr_2022_3M3t3tUbA2Y" ]
iclr_2022_miA4AkGK00R	EF21 with Bells & Whistles: Practical Algorithmic Extensions of Modern Error Feedback	First proposed by Seide et al (2014) as a heuristic, error feedback (EF) is a very popular mechanism for enforcing convergence of distributed gradient-based optimization methods enhanced with communication compression strategies based on the application of contractive compression operators. However, existing theory of ...	Reject	This paper presents several variants and extensions (including stochastic and proximal) of the error-feedback method EF21 and provides convergence rates for each of them and shows that they improve upon previous state of the arts. Despite the much broadened application scenarios and SOTA in convergence rates/complexit...	train	[ "44-hlRhKA3", "fGyS17yu6Fg", "x6vHpA5gvlK", "oUBodTQv4JK", "W_Qlwy6Kjcz", "8WDD6P6J7bK", "-1ubB-ZNfxK", "64-C10Gu1eD", "gDsSJZnmlez", "aiFr5S3S19v", "bDVphbUzHaM", "Cfi2UxDu7hA" ]	[ "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Dear reviewers, \n\nThanks for your valuable comments. **We feel our paper and our results were not understood and hence not appreciated. We propose several new and practical enhancements of the newly proposed EF21 error feedback method, obtaining state-of-the-art theoretical results for the error feedback mechan...	[ -1, -1, -1, -1, -1, -1, -1, -1, 5, 5, 5, 6 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, 4, 3, 4, 4 ]	[ "iclr_2022_miA4AkGK00R", "iclr_2022_miA4AkGK00R", "iclr_2022_miA4AkGK00R", "Cfi2UxDu7hA", "bDVphbUzHaM", "aiFr5S3S19v", "gDsSJZnmlez", "gDsSJZnmlez", "iclr_2022_miA4AkGK00R", "iclr_2022_miA4AkGK00R", "iclr_2022_miA4AkGK00R", "iclr_2022_miA4AkGK00R" ]
iclr_2022_w7Nb5dSMM-	Evolutionary perspective on model fine-tuning	Be it in natural language generation or in the image generation, massive performances gains have been achieved in the last years. While a substantial part of these advances can be attributed to improvement in machine learning architectures, an important role has also been played by the ever-increasing parameter number ...	Reject	The reviewers agree that this is an interesting treatise on some relationships between SGD fine tuning and evolutionary algorithms. All reviewers have requested some experimental validation or demonstration of the theory developed in this paper, which is not currently included. Whilst the computational requirements (an...	train	[ "RFsWr0lOBFW", "MoAtuq2p-qm", "ZG8KyjG1DUC", "i-pfkEd67NH", "C5qKt1g2Y08", "7Gv_QVesqxS", "0fRjU2WAi9Z", "cpXpKt4xzpi", "x8I1Uq8FmiE", "Vy8p9fFyBf1", "r858839Dze6", "Y_pztwVgd0i", "PvZIrtvEvqR" ]	[ "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " We thank the reviewers for their feedback, that contributed to improving the paper.\n\nAt this stage, the reviewers seem to have reached consensus that while theoretical approach is novel and significant to the field, experimental validation is still essential to support the main claims and provide empirical supp...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 5, 3, 3 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4, 4, 5 ]	[ "iclr_2022_w7Nb5dSMM-", "7Gv_QVesqxS", "cpXpKt4xzpi", "x8I1Uq8FmiE", "0fRjU2WAi9Z", "PvZIrtvEvqR", "Vy8p9fFyBf1", "Y_pztwVgd0i", "r858839Dze6", "iclr_2022_w7Nb5dSMM-", "iclr_2022_w7Nb5dSMM-", "iclr_2022_w7Nb5dSMM-", "iclr_2022_w7Nb5dSMM-" ]
iclr_2022_WQVouCWioh	Design in the Dark: Learning Deep Generative Models for De Novo Protein Design	The design of novel protein sequences is providing paths towards the development of novel therapeutics and materials. Generative modelling approaches to design are emerging and to date have required conditioning on 3D protein structure-derived information, and unconditional models of protein sequences have so far perf...	Reject	Despite a lively discussion and author explanation and revision, this paper remains below the bar for publication at ICLR. The technical exposition and goals remain poorly explained. The technical contribution is not sufficient. And the utility of the empirical results remain in question. The strong consensus among the...	train	[ "dmRX19EhIP8", "mAOtsik51rV", "PI2fI7Lj5EW", "0eU1E4HRrz", "ptRNbBexn8n", "DesRdNKAqfJ", "hr8y16sMNxa", "TYBcOnZQ8m2", "BFI_348DAAC", "qTl_Ek9vsih", "Qx4uONjt1mf", "e6Ydhb1GLKl", "aYHx7JQvoIn", "zr7Jo27_-MP", "ypzqAD-swVP", "lujemb4SqGr", "mYVB2JLf2hC", "u8oo-GLUv8j", "Bfwh8eQQjz...	[ "author", "official_reviewer", "official_reviewer", "author", "author", "official_reviewer", "official_reviewer", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_...	[ " We would like to thank all the reviewers for their time and effort evaluating our manuscript. We have tried to address all concerns raised. As a result we feel the paper has considerably improved, both from the addition of significant new methods, new result, and better communication of the work. Overall, we have...	[ -1, -1, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 3, 5, 3 ]	[ -1, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 4, 4, 5 ]	[ "iclr_2022_WQVouCWioh", "PI2fI7Lj5EW", "iclr_2022_WQVouCWioh", "ptRNbBexn8n", "DesRdNKAqfJ", "aYHx7JQvoIn", "qTl_Ek9vsih", "BFI_348DAAC", "mYVB2JLf2hC", "Qx4uONjt1mf", "e6Ydhb1GLKl", "JqpCv6xg3Q4", "zr7Jo27_-MP", "ypzqAD-swVP", "lujemb4SqGr", "WtQojQ-vKv", "u8oo-GLUv8j", "C1ryAbZL1...
iclr_2022_xxyTjJFzy3C	Contrastive Learning of 3D Shape Descriptor with Dynamic Adversarial Views	View-based deep learning models have shown the capability to learn 3D shape descriptors with superior performance on 3D shape recognition, classification, and retrieval. Most popular techniques often leverage the class label to train deep neural networks under supervision to learn to extract 3D deep representation by a...	Reject	This submission received a diverging set of the final ratings: 6, 3, 6, 5. On the positive side, reviewers appreciated practicality of the approach and supporting empirical results. At the same time, all of them expressed concerns with the presentation (typos, unfinished sentences, inconsistent notations). Additional r...	train	[ "tlTLaP0Jxaw", "M6LmLjTl_YW", "_sS0f_TjQYD", "ZnqpYFCPt9", "bVKRPq9KMzJ", "TlRZuziuo5S", "-dp8G93TimT", "6tdXVqM9J_S", "laF1azT6WeE", "a14fam0sDZ", "T2rcaJbO2jY", "EQC8dSVoWPC", "2f1Y4kU0uxW", "-0hYFvVCOGM" ]	[ "author", "official_reviewer", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Thank you so much for your encouraging comments.", " I would like to thank the authors for their response. They have addressed most of my questions and comments, particularly when it comes to comparing with other baseline augmentation schemes, and the overhead of online view generation. I am satisfied with the ...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, 3, 6, 5 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4, 3, 4 ]	[ "M6LmLjTl_YW", "laF1azT6WeE", "ZnqpYFCPt9", "-dp8G93TimT", "-0hYFvVCOGM", "-0hYFvVCOGM", "2f1Y4kU0uxW", "EQC8dSVoWPC", "T2rcaJbO2jY", "iclr_2022_xxyTjJFzy3C", "iclr_2022_xxyTjJFzy3C", "iclr_2022_xxyTjJFzy3C", "iclr_2022_xxyTjJFzy3C", "iclr_2022_xxyTjJFzy3C" ]
iclr_2022_WKWAkkXGpWN	Efficient Training and Inference of Hypergraph Reasoning Networks	We study the problem of hypergraph reasoning in large domains, e.g., predicting the relationship between several entities based on the input facts. We observe that in logical reasoning, logical rules (e.g., my parent's parent is my grandparent) usually apply locally (e.g., only three people are involved in a grandparen...	Reject	This paper has conflicting reviews with no strong advocate. One of the positive reviewers states the caveat that paper is "very dense to read and needs to be improved". Having looked at the paper myself I would agree with this criticism. One of the negative reviewers states that the paper gives "an incremental varia...	test	[ "AbvGpQCO0dX", "asjVpqqGTYJ", "jYKt0LmGg3V", "iRMQVwUQNi6", "6HN7FUM-Mk9", "tPZWrIo2ERd", "XuSjk189C93", "pCem9WxrBtQ", "TRe8hvsPOuy", "8l8B5OJGwTF", "BwUROjITpp", "Nvyr8NwrsSR", "o7DhDxsXKBi", "5UqARnu_2DT", "G4UdrDvOvYZ", "8nE33Jt5Obx", "TlVmjue4Z2r" ]	[ "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer" ]	[ " We thank all reviewers for their insightful reviews and helpful, constructive comments. In our general response, we would like to clarify and reiterate the motivation and the contribution of this paper.\n\nIn general, we want to learn models that can predict relationships among a set of entities. Consider simply ...	[ -1, -1, -1, 6, -1, 3, -1, 6, -1, 6, -1, -1, -1, -1, -1, -1, 3 ]	[ -1, -1, -1, 4, -1, 3, -1, 3, -1, 3, -1, -1, -1, -1, -1, -1, 4 ]	[ "iclr_2022_WKWAkkXGpWN", "6HN7FUM-Mk9", "G4UdrDvOvYZ", "iclr_2022_WKWAkkXGpWN", "5UqARnu_2DT", "iclr_2022_WKWAkkXGpWN", "o7DhDxsXKBi", "iclr_2022_WKWAkkXGpWN", "iclr_2022_WKWAkkXGpWN", "iclr_2022_WKWAkkXGpWN", "8l8B5OJGwTF", "tPZWrIo2ERd", "pCem9WxrBtQ", "TlVmjue4Z2r", "iRMQVwUQNi6", "...
iclr_2022_0Kj5mhn6sw	Gesture2Vec: Clustering Gestures using Representation Learning Methods for Co-speech Gesture Generation	Co-speech gestures are a principal component in conveying messages and enhancing interaction experiences between humans. Similarly, the co-speech gesture is a key ingredient in human-agent interaction including both virtual agents and robots. Existing machine learning approaches have yielded only marginal success in le...	Reject	PAPER: This paper describes a method to generate visual gestures by learning an intermediate representation based on gesture sequences. This proposed method builds from previous work on VAE and vector quantized VAE. DISCUSSION: The reviewers wrote some detailed reviews about the paper, bringing some valid concerns and ...	test	[ "nbeYkoTVGgI", "_BCw7d9PgLI", "zRD0C1JpL5E", "q6_xCAwvitM" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " I have read all reviews. It looks like Reviewer mN3D raised a lot of questions regarding implementation details. I had criticism on technical novelty and the limited experimental analysis. I do not see any reply from authors. I, therefore, lean towards a reject.", "In this paper, the authors propose an approach...	[ -1, 3, 6, 5 ]	[ -1, 5, 5, 4 ]	[ "q6_xCAwvitM", "iclr_2022_0Kj5mhn6sw", "iclr_2022_0Kj5mhn6sw", "iclr_2022_0Kj5mhn6sw" ]
iclr_2022_HUeyM2qVey2	Universal Joint Approximation of Manifolds and Densities by Simple Injective Flows	We analyze neural networks composed of bijective flows and injective expansive elements. We find that such networks universally approximate a large class of manifolds simultaneously with densities supported on them. Among others, our results apply to the well-known coupling and autoregressive flows. We build on the wor...	Reject	First this is the seed for a very good paper on approximating manifolds and densities using injective flows. Reviewers have done an admirable effort reviewing the paper giving detailed reviews and suggestions to improve the theory and corrections that resulted in an improvement of the paper during the rebuttal/ rev...	train	[ "FfwxSJeIPtQ", "RFDDhoUbnNF", "a_kGTK8BcW", "WYohs_Rf0Lm", "gZqtDLSfAYr", "aSlQxBzYlaq", "dZwpXUXBX_n", "k9HPLtf5P6A", "E9rFpB9eN3X", "hkcteH0hHpn", "Q5K8OxH9_oE", "5NCxeoEC8D-", "KNqHbqpNpP", "o5hrQifUKOK", "1rmclaLYCwV", "Yp3i499TD1v", "550PAlH9bkT", "GIqZk4Xi_O4", "NBLU3ZYOV-O...	[ "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "official_reviewer", "author", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_...	[ " Dear Reviewers eR1p, 2dK9, WKWr,\n\nReviewer 4sjW suggested they would further improve their score if we committed to implementing several organizational updates and if the other reviewers agreed with those updates. It would be fantastic if you could tell us whether this is the case. The main updates may be s...	[ -1, -1, 6, -1, -1, -1, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, 8 ]	[ -1, -1, 4, -1, -1, -1, -1, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 2, 4 ]	[ "iclr_2022_HUeyM2qVey2", "WYohs_Rf0Lm", "iclr_2022_HUeyM2qVey2", "aSlQxBzYlaq", "dZwpXUXBX_n", "GIqZk4Xi_O4", "5NCxeoEC8D-", "iclr_2022_HUeyM2qVey2", "5NCxeoEC8D-", "5NCxeoEC8D-", "5NCxeoEC8D-", "GIqZk4Xi_O4", "TG3rcU6wTVA", "KNqHbqpNpP", "Mz4AyfO4J_d", "a_kGTK8BcW", "a_kGTK8BcW", ...
iclr_2022_6PahjGFjVG-	Secure Distributed Training at Scale	Some of the hardest problems in deep learning can be solved via pooling together computational resources of many independent parties, as is the case for scientific collaborations and volunteer computing. Unfortunately, any single participant in such systems can jeopardize the entire training run by sending incorrect up...	Reject	This paper presents promising and ambitious work in the context of Byzantine-tolerant learning in the decentralized setup. The reviews raised several critical points of concern: completeness of technical derivations and details, experimental results and comparisons to other work, excluding several state of the art atta...	train	[ "bM4UJA7Y_re", "CbD8YgYB0i-", "cNhUUPjXY6M", "YAyOhn23W_z", "AjFlVW22nKH", "KbttF_2rWrx", "dG-JXSFuCn3", "dWSftPYtGdM", "bxafHf1zXLr", "3oD8iI1HFV", "41iElSdLerm", "xWC_avr6O4U", "YBNqY2JGuka", "Vf28laxkqQl", "SRbpUh5XGBL", "Urj7MT_H0Xe", "L1hMDML-80-", "WNCDI4rLabS", "uB5dIJy2IA...	[ "official_reviewer", "official_reviewer", "official_reviewer", "author", "official_reviewer", "official_reviewer", "author", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "author", "official_reviewer", "author", "author", "...	[ " I acknowledge I have read the author's final response.", "The paper considers the problem of decentralized training in the presence of Byzantine peers. The paper proposes a Byzantine-tolerant algorithm which uses randomly chosen peers as validators. A validator recomputes the gradient and ‘accuses’ a peer if th...	[ -1, 5, -1, -1, -1, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, 3, 6 ]	[ -1, 3, -1, -1, -1, 2, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 4, 4 ]	[ "xWC_avr6O4U", "iclr_2022_6PahjGFjVG-", "3oD8iI1HFV", "AjFlVW22nKH", "uB5dIJy2IA2", "iclr_2022_6PahjGFjVG-", "dWSftPYtGdM", "oJ6hKjYtq72", "YBNqY2JGuka", "YBNqY2JGuka", "Vf28laxkqQl", "Vf28laxkqQl", "SRbpUh5XGBL", "D7sxeZJpaH", "Urj7MT_H0Xe", "hEGqLYYbK-r", "iclr_2022_6PahjGFjVG-", ...
iclr_2022_4V4TZG7i7L_	Hierarchical Multimodal Variational Autoencoders	Humans find structure in natural phenomena by absorbing stimuli from multiple input sources such as vision, text, and speech. We study the use of deep generative models that generate multimodal data from latent representations. Existing approaches generate samples using a single shared latent variable, sometimes with m...	Reject	PAPER: This paper presents a multimodal auto-encoder architecture built on the premise that unimodal variations can be best generated when taking advantage of a shared latent space. This is operationalized by defining a hierarchical model with two primary levels: a shared structure space and unimodal variations (which ...	test	[ "XuDqEMipBS-", "D7JIExdZlTk", "jp8hkeRs18G", "H0Mdn89RK-N", "IXmS7IlqIVO", "myp9mDnQYti", "jv8YPcGfC1", "0Rik6LO6Cxi", "YUw-i64OGaN", "CBc4TqMMjx9", "B8FDWGrIot9", "IiKD7WepGbz", "H-FnGY0b0of" ]	[ "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " I thank the authors for their detailed response to reviewer feedback. \nI do not have any other questions/comments.", " Dear authors,\n\nThanks for the response. I do not have questions now.", " We thank Reviewer j3q4 for the feedback. We uploaded a revised version of the paper/code that incorporates this fee...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, 5, 5, 6 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 4, 4, 3 ]	[ "myp9mDnQYti", "jv8YPcGfC1", "IiKD7WepGbz", "YUw-i64OGaN", "YUw-i64OGaN", "CBc4TqMMjx9", "H-FnGY0b0of", "IiKD7WepGbz", "B8FDWGrIot9", "iclr_2022_4V4TZG7i7L_", "iclr_2022_4V4TZG7i7L_", "iclr_2022_4V4TZG7i7L_", "iclr_2022_4V4TZG7i7L_" ]
iclr_2022_IsHQmuOqRAG	Learning to perceive objects by prediction	The representation of objects is the building block of higher-level concepts. Infants develop the notion of objects without supervision. The prediction error of future sensory input is likely the major teaching signal for infants. Inspired by this, we propose a new framework to extract object-centric representation fro...	Reject	This paper tackles the difficult problem of learning to segment objects from an image using no supervision during training. The paper is clearly written and a new synthetic dataset is made available. Unfortunately, the reviewers raised a number of issues with the submission (missing citations and comparison to relevant...	test	[ "iYvhcd4ZbQY", "gmQ16nOPtP", "2ryQ3uUpI6t", "qKTV1jdPvs", "aU-F_hcIFgx", "B-HU3V2VrJp", "SxWPEOza_mi", "WxGgw3p0m06", "S9Z3pJlQLkU", "hQ0xM5MTe9y", "_8ejTwEW0H4", "FLQMvcWmzpu", "K8Z332m28Qn", "csQKLOEZWJ1", "WrDeHjS3hiC" ]	[ "official_reviewer", "official_reviewer", "author", "author", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " As per my review, this is work heading in a nice direction. While I was the most positive reviewer, I was also worried about the somewhat unfinished state of the paper, and addressing some of the concerns of others requires substantive extra careful validation. I am impressed with the authors' effort to address a...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 5, 5, 8 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 4, 4, 3 ]	[ "S9Z3pJlQLkU", "hQ0xM5MTe9y", "csQKLOEZWJ1", "B-HU3V2VrJp", "SxWPEOza_mi", "WxGgw3p0m06", "2ryQ3uUpI6t", "K8Z332m28Qn", "WrDeHjS3hiC", "FLQMvcWmzpu", "iclr_2022_IsHQmuOqRAG", "iclr_2022_IsHQmuOqRAG", "iclr_2022_IsHQmuOqRAG", "iclr_2022_IsHQmuOqRAG", "iclr_2022_IsHQmuOqRAG" ]
iclr_2022_KNfuensPHDU	Efficient Certification for Probabilistic Robustness	Recent developments on the robustness of neural networks have primarily emphasized the notion of worst-case adversarial robustness in both verification and robust training. However, often looser constraints are needed and some margin of error is allowed. We instead consider the task of probabilistic robustness, which a...	Reject	The paper proposes a method to improve PROVEN, which gives a certification for probabilistic robustness. However, reviewers think the paper is below the acceptance bar due to unclear motivation and insufficient experiments. In particular, a clear use case of probabilistic robustness certification is crucial for the pap...	train	[ "ZKFecastF6g", "z0TUt8aFzp7", "MvL0fWkpPmr", "QIw31h9eBU", "UC4ZEEqhTYc", "8zaMkli-W2N", "tNP9XkdzrF2", "npolf_819oW", "fxaM1-jjE0", "cL-jwwsnCDM" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Thanks for your response. I am not convinced by the additional experiment, partly because there are many verification methods that are not Monte Carlo--for instance the first paper that I linked to above, which uses semidefinite programming instead and is deterministic.\n\nRegarding correctness, the 2 category re...	[ -1, -1, 5, -1, -1, -1, -1, 3, 5, 5 ]	[ -1, -1, 4, -1, -1, -1, -1, 2, 3, 4 ]	[ "QIw31h9eBU", "8zaMkli-W2N", "iclr_2022_KNfuensPHDU", "npolf_819oW", "fxaM1-jjE0", "cL-jwwsnCDM", "MvL0fWkpPmr", "iclr_2022_KNfuensPHDU", "iclr_2022_KNfuensPHDU", "iclr_2022_KNfuensPHDU" ]
iclr_2022_1-YP2squpa7	Deep learning via message passing algorithms based on belief propagation	Message-passing algorithms based on the Belief Propagation (BP) equations constitute a well-known distributed computational scheme. It is exact on tree-like graphical models and has also proven to be effective in many problems defined on graphs with loops (from inference to optimization, from signal processing to clust...	Reject	This paper presents a method for training neural networks with belief propagation-based algorithms. The approach is to set a fully factorized prior over weights, compute a forward and backward pass of messages on a minibatch, then set the new prior to be a slightly higher temperature version of the minibatch approximat...	train	[ "w3vQi3j1f1a", "0WlAix6UkHk", "cSeLbaQBnS", "jLs-DW4zyd", "S7nBqzz2PiQ", "xSZ-_tQTauA", "NL0Svx6lBPs", "NUPQCDUEuLw", "KjpWXeWUbHV", "w4T24COthX", "iElpcKxLPgD", "KLQmpUK4_xQ", "fylQf9kFYL", "YBqh09Epq2E", "NG8qiofnc6P" ]	[ "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ " We thank the reviewer for her/his stimulating follow-up. \n\n> still there are several concerns with the current results and presentation, including the advantages of the proposed algorithm and practical usage in real applications. \n\nIn this regard, we would like to reiterate that the purpose of the work was to...	[ -1, 8, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, 3 ]	[ -1, 3, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4 ]	[ "cSeLbaQBnS", "iclr_2022_1-YP2squpa7", "NL0Svx6lBPs", "iclr_2022_1-YP2squpa7", "jLs-DW4zyd", "jLs-DW4zyd", "jLs-DW4zyd", "0WlAix6UkHk", "0WlAix6UkHk", "YBqh09Epq2E", "NG8qiofnc6P", "NG8qiofnc6P", "iclr_2022_1-YP2squpa7", "iclr_2022_1-YP2squpa7", "iclr_2022_1-YP2squpa7" ]
iclr_2022_c7S4WIlmu5	Contrastive Pre-training for Zero-Shot Information Retrieval	Information retrieval is an important component in natural language processing, for knowledge intensive tasks such as question answering and fact checking. Recently, information retrieval has seen the emergence of dense retrievers, based on neural networks, as an alternative to classical sparse methods based on term-fr...	Reject	Good premise: What unsupervised training supports IR? This is a key question for IR and is a focus for papers in TREC 2019 Deep Learning Track, for instance. Also, historically, empirical work in the IR community is a very high standard. One reviewer says the contrastive loss for learning Siamese Transformers is n...	train	[ "nsZ72lfBkE_", "GOmCoFv9Zt-", "Zbc9Ik6j8G", "5d-8qc2JICn", "xZM-YzrvAGD", "5E3IS1LMegr", "bcEtF5b9Z4h", "NygoBnumsqA", "143C1z-kdE", "jCqsLchZBhX" ]	[ "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Thanks to the authors for updating the paper and for adding some new results.\n\nI still feel that the paper writing can be improved a lot including the Introduction and Related Work sections. For example, Table 8 needs to be in the main paper and not in the appendix. There can be a better comparison with the pre...	[ -1, -1, -1, -1, -1, -1, 5, 5, 3, 3 ]	[ -1, -1, -1, -1, -1, -1, 3, 4, 5, 4 ]	[ "Zbc9Ik6j8G", "iclr_2022_c7S4WIlmu5", "143C1z-kdE", "jCqsLchZBhX", "NygoBnumsqA", "bcEtF5b9Z4h", "iclr_2022_c7S4WIlmu5", "iclr_2022_c7S4WIlmu5", "iclr_2022_c7S4WIlmu5", "iclr_2022_c7S4WIlmu5" ]
iclr_2022_YxWU4YZ4Cr	Generalization to Out-of-Distribution transformations	Humans understand a set of canonical geometric transformations (such as translation, rotation and scaling) that support generalization by being untethered to any specific object. We explored inductive biases that allowed artificial neural networks to learn these transformations in pixel space in a way that could genera...	Reject	This paper studies different inductive biases that would improve OOD generalization (and in particular under translation, rotation and scaling) for image tasks. The study is focused on a toy dataset which allows authors to have more control over the data generation process and the transformations. Authors further show ...	test	[ "LXCu9MJBRaZ", "uDJyk-E02I-", "9rmROiSECJa", "4slog3W5jbn" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ "The authors trained autoencoders and variational autoencoders in cartesian space, as well as autoencoders in log-polar space, with generated data representing a range of interesting transformations. They then tested the ability of the models to extrapolate beyond the learned transformations in pixel space. The aut...	[ 6, 3, 3, 3 ]	[ 3, 4, 3, 3 ]	[ "iclr_2022_YxWU4YZ4Cr", "iclr_2022_YxWU4YZ4Cr", "iclr_2022_YxWU4YZ4Cr", "iclr_2022_YxWU4YZ4Cr" ]
iclr_2022_6ooiNCGZa5K	On-Target Adaptation	Domain adaptation seeks to mitigate the shift between training on the source data and testing on the target data. Most adaptation methods rely on the source data by joint optimization over source and target. Source-free methods replace the source data with source parameters by fine-tuning the model on target. Either wa...	Reject	The work proposed an interesting source free adaptation setting, where one is asked to adapt a pre-trained source model to a target domain without accessing data from the source domain. While reviewers find the setup interesting and the initial results encouraging, they expressed concerns on the limited novelty of the ...	train	[ "dzrsmybMLy", "fYaFiOy7Tit", "8-ARyy8uKK", "dGrSmrZ6Fne", "Po23mUfnwi", "MgT9q60czb", "d9-Q0ec9c6", "9EZULyUBe8I", "q3LaJWbSDYU" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ " Thank you for your response. I see that some of my concerns were addressed, but for some, I did not find a satisfactory response. E.g. on missing results in the tables or not including SHOT. However, I still believe that the paper is worth publishing and prefer to keep my original rating of \"marginally above ......	[ -1, 5, -1, -1, -1, -1, -1, 3, 6 ]	[ -1, 4, -1, -1, -1, -1, -1, 3, 3 ]	[ "MgT9q60czb", "iclr_2022_6ooiNCGZa5K", "dGrSmrZ6Fne", "fYaFiOy7Tit", "9EZULyUBe8I", "q3LaJWbSDYU", "iclr_2022_6ooiNCGZa5K", "iclr_2022_6ooiNCGZa5K", "iclr_2022_6ooiNCGZa5K" ]
iclr_2022_TWANKAJ1ZCr	Learn Together, Stop Apart: a Novel Approach to Ensemble Pruning	Gradient boosting is the most popular method of constructing ensembles that allow getting state-of-the-art results on many tasks. One of the critical parameters affecting the quality of the learned model is the number of models in the ensemble, or the number of boosting iterations. Unfortunately, the problem of selecti...	Reject	While the reviewers agree that the paper contains interesting ideas and the method is elegant, it unfortunately does not meet the bar for acceptance. I strongly encourage the authors to revise their paper, in particular using the numerous comments made throughout the discussion phase; for example: * It is important th...	train	[ "kP5NH5lJ7GS", "aW-w-jYbEPv", "ZtYROoMqe4y", "dGTFRQUU2nw", "Dl0FoM8DbGo", "4ig764yXnRB", "WbDLJ3Bug9Z", "veMpzjhhf-", "m2vRCb-DfW", "4XG5rKQHqaa", "dkM81nubzAw" ]	[ "author", "author", "official_reviewer", "author", "official_reviewer", "official_reviewer", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " We apply the pairwise Wilcoxon test to the test sets’ results obtained from the default and pruned models. Namely, the test set was divided into parts, and the qualities of two models on each part form one observation pair. Then we apply the pairwise Wilcoxon test to the obtained set of pairs.\n\nThe method of se...	[ -1, -1, -1, -1, -1, -1, -1, -1, 5, 6, 5 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, 4, 4, 5 ]	[ "ZtYROoMqe4y", "4ig764yXnRB", "dGTFRQUU2nw", "m2vRCb-DfW", "veMpzjhhf-", "WbDLJ3Bug9Z", "dkM81nubzAw", "4XG5rKQHqaa", "iclr_2022_TWANKAJ1ZCr", "iclr_2022_TWANKAJ1ZCr", "iclr_2022_TWANKAJ1ZCr" ]
iclr_2022_fHPdmN3I0tY	Decoupled Kernel Neural Processes: Neural Network-Parameterized Stochastic Processes using Explicit Data-driven Kernel	Neural Processes (NPs) are a class of stochastic processes parametrized by neural networks. Unlike traditional stochastic processes (e.g., Gaussian processes), which require specifying explicit kernel functions, NPs implicitly learn kernel functions appropriate for a given task through observed data. While this data-dr...	Reject	This paper proposes Decoupled Kernel Neural Processes (DKNPs), a new neural stochastic process, which learns a separate mean and kernel function to directly model the covariance between output variables. Numerical experiments on 1-D regression and 2-D image completion are provided. There was a concern that the origin...	test	[ "aO9T-JX1Apo", "_Vk5n6lc595", "b_bdPrddBs", "wkXteJrNgjR", "s7TB52GFbE5", "4yig48SLXVJ", "H1TFlsElDaj", "Sfs60y8ndso", "YNmP3P6akSh", "AjmlmRMTSR-", "cDhjuXBd676", "-E2FGUzd6rN" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Thanks to the authors for their response. The lack of novelty remains the major issue of this work though I agree this work is well motivated. The novelty is limited because based on my understanding, with a single target sample, the model reduces to the attentional neural processes without a probabilistic enco...	[ -1, -1, 5, -1, -1, -1, -1, -1, -1, 6, 5, 8 ]	[ -1, -1, 4, -1, -1, -1, -1, -1, -1, 2, 5, 4 ]	[ "Sfs60y8ndso", "4yig48SLXVJ", "iclr_2022_fHPdmN3I0tY", "-E2FGUzd6rN", "4yig48SLXVJ", "b_bdPrddBs", "AjmlmRMTSR-", "cDhjuXBd676", "iclr_2022_fHPdmN3I0tY", "iclr_2022_fHPdmN3I0tY", "iclr_2022_fHPdmN3I0tY", "iclr_2022_fHPdmN3I0tY" ]
iclr_2022_FH_mZOKFX-b	Takeuchi's Information Criteria as Generalization Measures for DNNs Close to NTK Regime	Generalization measures are intensively studied in the machine learning community for better modeling generalization gaps. However, establishing a reliable generalization measure for statistical singular models such as deep neural networks (DNNs) is challenging due to the complex nature of the singular models. We foc...	Reject	While the reviewers acknowledge the broad experimental work done in this paper, they all find several issues, which in their combination show that the paper is simply not in a good enough shape. This impression has not changed during the rebuttal phase and as a result, this is a clear reject.	train	[ "aqEnCZXjXFp", "rfqZVrqrtTC", "tXFAco0BI2b", "YLq98g3VZF0", "2BOV3Ki3n7B", "wHTVpHmE0z", "HBitwGXLc4R", "9D4t70-nhu0", "PJG5CofOpo4", "wR4vhbrSGoe" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer" ]	[ "The authors experimentally investigate the ability of Takeuchi's information criterion to predict generalization gaps in various neural network architectures, with a particular focus on the Neural Tangent Kernel (NTK) regime. This was previously done only on small scale networks [1], and the authors extend it to s...	[ 3, 5, 3, -1, -1, -1, -1, -1, -1, 3 ]	[ 3, 3, 3, -1, -1, -1, -1, -1, -1, 4 ]	[ "iclr_2022_FH_mZOKFX-b", "iclr_2022_FH_mZOKFX-b", "iclr_2022_FH_mZOKFX-b", "wR4vhbrSGoe", "tXFAco0BI2b", "rfqZVrqrtTC", "aqEnCZXjXFp", "aqEnCZXjXFp", "rfqZVrqrtTC", "iclr_2022_FH_mZOKFX-b" ]
iclr_2022_U1edbV4kNu_	SWARM Parallelism: Training Large Models Can Be Surprisingly Communication-Efficient	Many deep learning applications benefit from using large models with billions of parameters. These models can only be trained with specialized distributed training algorithms that require low-latency and high-bandwidth interconnect. As a result, large models are typically trained in dedicated GPU clusters that can be e...	Reject	Overall, the reviewers thought this paper suggested an important problem. However, there were many concens. Particularly, the multiple reviewers felt it was unclear when the new approach is better than prior work. The reviewers had difficulty connecting the experiments to the paper's main claims.	train	[ "_0feOQ4atJ0", "Npg77t_Zaas", "B_hjhBVilwM", "074eoQ_9jo", "BEE1q0RHvIN", "KKCCZKq_SnO", "yoqnKT5HRjU", "GcoWMVzMi0S", "phJr9AsoqMH", "Qhe2R6bzpJb", "JwoL6gSvO2r", "WaWXSGa7kcs", "TbZsKtxiXwf", "jFx0peY4h4M", "jDYv_fcchnf", "H-n4i-ggD1g", "mOQ5iPdCsL", "5Cu8F4-bPN9", "r7Oub1Qs17k...	[ "official_reviewer", "author", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", ...	[ "The paper describes a novel parallelism strategy for training large neural networks: compute devices are grouped into “swarms”, pipeline parallelism is applied within each swarm, and pipeline parallelism is applied between swarms. Communication between swarms is randomized and swarms are periodically rebalanced, m...	[ 6, -1, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 6 ]	[ 4, -1, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 4 ]	[ "iclr_2022_U1edbV4kNu_", "iclr_2022_U1edbV4kNu_", "jFx0peY4h4M", "iclr_2022_U1edbV4kNu_", "GcoWMVzMi0S", "Qhe2R6bzpJb", "H-n4i-ggD1g", "iclr_2022_U1edbV4kNu_", "074eoQ_9jo", "TbZsKtxiXwf", "WaWXSGa7kcs", "iclr_2022_U1edbV4kNu_", "ErH0_xnKZMI", "jDYv_fcchnf", "074eoQ_9jo", "_0feOQ4atJ0"...
iclr_2022_niZImJIrqVt	Mean-Variance Efficient Reinforcement Learning by Expected Quadratic Utility Maximization	Risk management is critical in decision making, and mean-variance (MV) trade-off is one of the most common criteria. However, in reinforcement learning (RL) for sequential decision making under uncertainty, most of the existing methods for MV control suffer from computational difficulties owing to calculating the gradi...	Reject	This is a borderline paper with some reviewers voted for acceptance and some think it is not still ready. What is clear is more efforts by the authors is needed to make the paper appealing to reviewers with different interests. Changes such as better writing, more in depth literature review, more convincing experiments...	train	[ "_GGkyuJ6nrb", "QcJu0B9m6hY", "qduzJbSXzWK", "f4Tlpjje3X", "1lVVzEa_qZi", "ah2pb5QZGx", "QXDx14Rj-pW", "Hpn7Ec-Bxiy", "BTbxdbeZN81", "tuHJc1G4Dh", "4l62K6Pq85J", "VIefLLokzv5", "xMC-XbrO6qk", "-9NFss2WFQR", "vAvN2m5SB-s", "ipPw1u0duTu", "6udkGIiBRyY" ]	[ "author", "official_reviewer", "official_reviewer", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Thank you for sharing the interesting papers. They are online learning, not RL; that is, their setting is different from ours. e.g., they consider an adversary with switching costs or problem setting different from MDPs. On the other hand, our problem formulation is RL on MDP by definition.\n\n**Note that online ...	[ -1, -1, 8, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 6, 5 ]	[ -1, -1, 4, -1, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 4, 4 ]	[ "QcJu0B9m6hY", "-9NFss2WFQR", "iclr_2022_niZImJIrqVt", "ah2pb5QZGx", "iclr_2022_niZImJIrqVt", "BTbxdbeZN81", "iclr_2022_niZImJIrqVt", "6udkGIiBRyY", "tuHJc1G4Dh", "xMC-XbrO6qk", "qduzJbSXzWK", "ipPw1u0duTu", "1lVVzEa_qZi", "vAvN2m5SB-s", "iclr_2022_niZImJIrqVt", "iclr_2022_niZImJIrqVt"...
iclr_2022_X1y1ur-NCh_	Did I do that? Blame as a means to identify controlled effects in reinforcement learning	Identifying controllable aspects of the environment has proven to be an extraordinary intrinsic motivator to reinforcement learning agents. Despite repeatedly achieving State-of-the-Art results, this approach has only been studied as a proxy to a reward-based task and has not yet been evaluated on its own. We show that...	Reject	This paper presents an attempt to infer controllable aspects of the environment dynamics by imposing an architecture on a pixel prediction model, such that the prediction is the sum of an action-aware and action-agnostic prediction, trained on a fixed dataset gathered under a uniform random policy, in such a way as to ...	train	[ "Vr4aDEhNcNO", "Phw_3mU-P15", "h45EifFV1Rx", "Q8vioEZzNQK", "aWPJsxLtVm", "n_QJWM9id-J", "vvhGXWSY9G", "YG0M0JFsy19", "YHv0VzczGdB", "uFbvr6tFs3U", "40rzT4HQf8D", "iwi09W-cGI", "rtczN8w89R1", "CiOeZP1ldFv", "X7gr54TzJJ", "jkj9q99MTVN", "hlf8VLx4cOL", "7HqqvGFrBwe", "qCfHTcxdDuZ",...	[ "official_reviewer", "official_reviewer", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "official_reviewer", "official_reviewer", "author", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "a...	[ " I appreciate the authors' detailed answers to my questions. I understand now how despite a low F1 score, CEN is still correctly identifying the controlled objects in MZR. The follow-up experiment on MZR is also encouraging, in that engineering the network architecture could enable CEN to scale to domains with mor...	[ -1, -1, -1, 5, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, 6 ]	[ -1, -1, -1, 4, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 4 ]	[ "qCfHTcxdDuZ", "h45EifFV1Rx", "uFbvr6tFs3U", "iclr_2022_X1y1ur-NCh_", "40rzT4HQf8D", "iclr_2022_X1y1ur-NCh_", "YG0M0JFsy19", "YHv0VzczGdB", "iwi09W-cGI", "X7gr54TzJJ", "rtczN8w89R1", "CiOeZP1ldFv", "7HqqvGFrBwe", "hlf8VLx4cOL", "jkj9q99MTVN", "CIhz_-Mm1hT", "n_QJWM9id-J", "Q8vioEZz...
iclr_2022_IJ-88dRfkdz	SoftHebb: Bayesian inference in unsupervised Hebbian soft winner-take-all networks	State-of-the-art artificial neural networks (ANNs) require labelled data or feedback between layers, are often biologically implausible, and are vulnerable to adversarial attacks that humans are not susceptible to. On the other hand, Hebbian learning in winner-take-all (WTA) networks, is unsupervised, feed-forward, and...	Reject	The authors provide an analysis of soft-winner-take-all (WTA) networks with Hebbian local learning as a generative probabilistic mixture model. They then present experiments on comparably simple data sets, MNIST and F-MNIST. Results are compared to hard WTA networks and an MLP of the same size (single hidden layer) tra...	train	[ "81zTIfqitqY", "qXYj6I06BrH", "1A-2aBAJ4eW", "2bD0t66sSJg", "9PrbjPM1dMj", "gzdgNka-x5A", "6OV_0e19wX", "mYtjlaDQC3z", "5xcF-FYIfkP", "TxMv_yen2SV", "jJeZs4KOB4n", "5hxbwr6KwNY", "E0jwqeJRXQS", "-6FBLYpOkG", "ucumRLmKUB", "oVMz2y1q6FE", "2vvME4cSnT", "Z0mdmzBmgPu", "4AJwUxNQgB4",...	[ "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "official_reviewer" ...	[ "The work presents an interesting equivalency between a generative probabilistic mixture model and a winner-take-all Hebbian local learning rule-based learning. The approach seems promising, especially the adversarial robustness results compared to a multi-layer perceptron. \n\nUPDATED SCORE\nI would like to th...	[ 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, -1, 6, -1, -1, -1, 6 ]	[ 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, -1, 3, -1, -1, -1, 4 ]	[ "iclr_2022_IJ-88dRfkdz", "2vvME4cSnT", "Z0mdmzBmgPu", "81zTIfqitqY", "Z0mdmzBmgPu", "ZZyeZYRo9oH", "Hhv9cpRLup3", "81zTIfqitqY", "ucumRLmKUB", "Hhv9cpRLup3", "iclr_2022_IJ-88dRfkdz", "wkRO7sZpBQT", "iclr_2022_IJ-88dRfkdz", "oVMz2y1q6FE", "81zTIfqitqY", "iclr_2022_IJ-88dRfkdz", "-6FBL...
iclr_2022_D9E8MKsfhw	An Empirical Investigation of the Role of Pre-training in Lifelong Learning	The lifelong learning paradigm in machine learning is an attractive alternative to the more prominent isolated learning scheme not only due to its resemblance to biological learning, but also its potential to reduce energy waste by obviating excessive model re-training. A key challenge to this paradigm is the phenomeno...	Reject	The paper presents an empirical study on the impact of pertained model on lifelong learning. It concludes that the generic pertaining can benefit the lifelong learning duet the flatter loss landscape and evaluates on CV and NLP tasks. The paper is well written with detailed analysis. However, there is concerns on its l...	train	[ "LNFcMGQnlr", "DHmoUcI2sb5", "MxYCBJ0rl-9", "cbcF0r8Twk", "BOp7UI0ijg", "nLB-YQ_T5m0", "zJSp13hbeo3", "Ul6uFydd9Fw", "08a_xLngqSl", "wx_EPFxkBI" ]	[ "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " [Task similarity/dissimilarity]\n\nFor a fair comparison between pre-trained and randomly initialized models, we wanted to explicitly control for and remove the overlap between pre-training and downstream tasks. For NLP experiments, this is less of a concern as Transformer models are pre-trained on generic la...	[ -1, -1, -1, -1, -1, -1, -1, 6, 5, 5 ]	[ -1, -1, -1, -1, -1, -1, -1, 4, 4, 4 ]	[ "wx_EPFxkBI", "wx_EPFxkBI", "iclr_2022_D9E8MKsfhw", "08a_xLngqSl", "08a_xLngqSl", "wx_EPFxkBI", "Ul6uFydd9Fw", "iclr_2022_D9E8MKsfhw", "iclr_2022_D9E8MKsfhw", "iclr_2022_D9E8MKsfhw" ]
iclr_2022_5qwA7LLbgP0	Disentangling Sources of Risk for Distributional Multi-Agent Reinforcement Learning	In cooperative multi-agent reinforcement learning, state transitions, rewards, and actions can all induce randomness (or uncertainty) in the observed long-term returns. These randomnesses are reflected from two risk sources: (a) agent-wise risk (i.e., how cooperative our teammates act for a given agent) and (b) environ...	Reject	This paper proposes a method of multi-agent reinforcement learning that separately deals with the risk associated with uncertainties of the other agents and the risk associated with the uncertainties of the environment. This allows for example to be agent-wise risk seeking and environment-wise risk averse. The propose...	train	[ "rDG-wVjDxDW", "jRhJq8t2v1c", "gDhzJE74k86", "vg2KQiLKm3", "yWvILey8w50", "QyQQT-mL4vn", "Z6YEaCWGV3", "VgGue5OOvLk", "sMTCrbJUci_", "L_dZMw8oxSL", "7MnKfKlQnjp", "vBD4E745Cev", "Y60RQJ0RLBr", "5GJYZEkp5Z7", "E2jIYd1aVk7", "TvXHSSiyblu", "LtyjtKIb8h6", "FUjOj-C5wqB", "mvmuY4RA8z7...	[ "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", ...	[ " We highly appreciate your insightful and helpful comments. We are glad to hear that our response helped you understand our method.", " We highly appreciate your insightful and helpful comments. We are glad to hear that our response helped you understand our method. Thank you very much for raising the score.\n\n...	[ -1, -1, 6, 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, 3, 3 ]	[ -1, -1, 4, 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, 3, 4, 4 ]	[ "L_dZMw8oxSL", "yWvILey8w50", "iclr_2022_5qwA7LLbgP0", "iclr_2022_5qwA7LLbgP0", "Z6YEaCWGV3", "5GJYZEkp5Z7", "sMTCrbJUci_", "7MnKfKlQnjp", "vBD4E745Cev", "FUjOj-C5wqB", "LtyjtKIb8h6", "Y4MdQlb2ZA", "iclr_2022_5qwA7LLbgP0", "yfZdIVmWZYb", "iclr_2022_5qwA7LLbgP0", "JVcqQA7_LX-", "bTkSa...
iclr_2022_ZkC8wKoLbQ7	Understanding and Preventing Capacity Loss in Reinforcement Learning	The reinforcement learning (RL) problem is rife with sources of non-stationarity that can destabilize or inhibit learning progress. We identify a key mechanism by which this occurs in agents using neural networks as function approximators: \textit{capacity loss}, whereby networks trained to predict a sequence of target...	Accept (Spotlight)	The paper analyzes the learning behavior of deep networks inside RL algorithms, and proposes an interesting hypothesis: that many of the observed difficulties in deep RL methods stem from _capacity loss_ of the trained network (that is, the network loses the ability to adapt quickly to fit new functions). As the paper ...	train	[ "KXaimdcTMr", "yYH31aTkkOd", "fDttjpL0hSP", "UFl3vR8oX2Y", "Ohl6Cgx2j78", "xdMjZOdV_wW", "IqxfMJwkDLO", "hUmglKXzB5A", "nqoBb8b8vEF", "Rok1rh90NAm", "5rvKxPUrJgu", "7vojLMY6Jfq", "5Jnq4qbjsUh", "HA14Ho3btrC", "k-f-O_9K14u", "jRtDNLqFKSG", "WjrI_KXM3x6", "JRoU1ZVxG6W", "aL41iVjoXQ...	[ "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "official_reviewer", "author", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", ...	[ " Thanks to the reviewer for their engagement in the discussion period. As the discussion period is ending today, we will not be able to respond to any follow-up comments the reviewer may have. We refer the reviewer to our discussion with reviewer Mqcv if our response has flagged any additional questions, as this d...	[ -1, -1, 6, -1, 8, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 8, 3 ]	[ -1, -1, 4, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4 ]	[ "HA14Ho3btrC", "UFl3vR8oX2Y", "iclr_2022_ZkC8wKoLbQ7", "hUmglKXzB5A", "iclr_2022_ZkC8wKoLbQ7", "k-f-O_9K14u", "nqoBb8b8vEF", "nqoBb8b8vEF", "Rok1rh90NAm", "JoDAxKCHs-", "HA14Ho3btrC", "HA14Ho3btrC", "JhnaIrVcrzy", "qkpB38yAN2D", "hnlneN0S3hI", "fDttjpL0hSP", "qkpB38yAN2D", "qkpB38y...
iclr_2022_BNIt2myzSzS	IA-MARL: Imputation Assisted Multi-Agent Reinforcement Learning for Missing Training Data	Recently, multi-agent reinforcement learning (MARL) adopts the centralized training with decentralized execution (CTDE) framework that trains agents using the data from all agents at a centralized server while each agent takes an action from its observation. In the real world, however, the training data from some agent...	Reject	The paper tackles the problem of missing data in centralized training multi-agent RL approaches. The authors propose 1) using generative adversarial imputation networks for imputing missing data and 2) discarding training data where data from multiple consecutive timesteps is missing. Reviewers agreed that the problem...	train	[ "aDnCXVCBWQ5", "xdqyiN_h3IE", "_xW7CTbmn8z", "hzBXURX0Jr", "7Z8WXc-iVTN", "6fJNOtUMFm", "NTgf1RhI_7s", "0ihTcyJ98u", "qKY0caTQSdW" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ " I have read the author response and looked at the revised version of the submission.\n\n### Writing\n\nThe submission does not appear to have made any changes to its writing. Even the two sentences explicitly given above remain uncorrected.\n\n> We have taken a close look at the paper Lyu et al. (2021). Although ...	[ -1, 3, 3, -1, -1, -1, -1, 1, 5 ]	[ -1, 3, 4, -1, -1, -1, -1, 4, 4 ]	[ "7Z8WXc-iVTN", "iclr_2022_BNIt2myzSzS", "iclr_2022_BNIt2myzSzS", "_xW7CTbmn8z", "0ihTcyJ98u", "xdqyiN_h3IE", "qKY0caTQSdW", "iclr_2022_BNIt2myzSzS", "iclr_2022_BNIt2myzSzS" ]
iclr_2022_2cpsEstmH1	Beyond Examples: Constructing Explanation Space for Explaining Prototypes	As deep learning has been successfully deployed in diverse applications, there is ever increasing need for explaining its decision. Most of the existing methods produced explanations with a second model that explains the first black-box model, but we propose an inherently interpretable model for more faithful explanati...	Reject	This paper proposes to create an explanation space to describe the relationships between input data and prototypes (and also between the prototypes themselves). It constructs such a space suing VAEs and conducts experiments to validate the effectiveness and interpretability of the method. Strengths: - The proposed met...	val	[ "2RYCoAK15-l", "Iow4giJt1B", "IsVaZPUZuOe", "5CirhILdKvU", "Gj82c4JTS2X", "P0-QMCyzlpz", "uiyrJCOYpgM", "1UgOD9coijQ", "aiZRUXTeKRa", "HpwXhrGLsdj", "p5cjN17Rlc", "zqVT0K3Bpin", "Ov0gIz437Qx", "OD56MlOkic_f", "Ik7nb-6tys_", "_VN-oFyUFW_", "Q8F4SBBnsiT", "1Qf6RZ8bFZJ", "pAhBJPK4z1...	[ "author", "author", "author", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " \\\nThank you very much for reviewing our paper thoughtfully. \n\nThe following are our comments on your thoughtful comment.\n\n\\\n**Summary Of The Review: The idea is interesting but not a lot of novelty, as some existing papers (e.g. Li's Deep Learning for Case-Based Reasoning through Prototypes paper) already...	[ -1, -1, -1, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 8, 5, 3 ]	[ -1, -1, -1, -1, 2, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 3, 3 ]	[ "as4UzWZTxRk", "P0-QMCyzlpz", "p5cjN17Rlc", "P0-QMCyzlpz", "iclr_2022_2cpsEstmH1", "HpwXhrGLsdj", "as4UzWZTxRk", "GBojo4gjNi", "p5cjN17Rlc", "p5cjN17Rlc", "Ov0gIz437Qx", "iclr_2022_2cpsEstmH1", "Gj82c4JTS2X", "Gj82c4JTS2X", "Gj82c4JTS2X", "pAhBJPK4z1H", "GBojo4gjNi", "GBojo4gjNi", ...
iclr_2022_fRb9LBWUo56	On the benefits of deep RL in accelerated MRI sampling	Deep learning approaches have shown great promise in accelerating magnetic resonance imaging (MRI), by reconstructing high quality images from highly undersampled data. While previous sampling methods relied on heuristics, recent work has improved the state-of-the-art (SotA) with deep reinforcement learning (RL) sampli...	Reject	While all reviewers acknowledge the relevance of such an evaluation work for the MRI reconstruction field, they all agree that the contribution has a limited fit with a ML conference like ICLR. The work is solid experimentally and will surely interest the audience of conferences like ISMRM or MICCAI. For this reason, t...	train	[ "TdgbSlmSM6x", "oTIdRWBqpjO", "MfH3MAt38CC", "LvpuPYFrrxL", "j0TFDTgsOiK", "RAbvZsub5i7", "C0whVKhl4qq", "8yS95d-w1Zu", "iJpisQm0KU5", "xSuAbr2KXoq", "2ianBd4bvj9", "q-csZw_41D", "KpbjYwTCGd", "6dRptuMQNgK" ]	[ "author", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Dear reviewer,\n\nWe have added Appendix H with a comparison to our best effort reproduction to Pineda et al. in the setting of Bakker et al., as suggested by reviewer 8cse. The SSIM and PSNR AUCs against the sampling rate are shown in the table below. \n\nAs we have noted, the computational and time expense by ...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 3, 3, 5 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 2, 4, 2 ]	[ "LvpuPYFrrxL", "LvpuPYFrrxL", "iclr_2022_fRb9LBWUo56", "RAbvZsub5i7", "KpbjYwTCGd", "6dRptuMQNgK", "q-csZw_41D", "2ianBd4bvj9", "xSuAbr2KXoq", "iclr_2022_fRb9LBWUo56", "iclr_2022_fRb9LBWUo56", "iclr_2022_fRb9LBWUo56", "iclr_2022_fRb9LBWUo56", "iclr_2022_fRb9LBWUo56" ]
iclr_2022_clwYez4n8e8	Logarithmic Unbiased Quantization: Practical 4-bit Training in Deep Learning	Quantization of the weights and activations is one of the main methods to reduce the computational footprint of Deep Neural Networks (DNNs) training. Current methods enable 4-bit quantization of the forward phase. However, this constitutes only a third of the training process. Reducing the computational footprint of th...	Reject	This paper proposes a method for 4-bit quantized training of NNs (forward and backward), obtaining SOTA 4-bit training quantization, motivated by an analysis of rounding schemes (an important aspect) in quantized training. The main concerns from the reviewers were that the approach was not practical (both a general con...	train	[ "zeRSc5NhzVe", "NQk8sg_ZUW5", "RPitzHGTTdY", "rvfkHZQaAFW", "tZEPUcASz4", "W-NNyfRSmQP", "3KyO1OtFzS", "i3Qud3KPH59", "_zSzPUy-UQ9", "OJxXzG5BtbS", "xUQn0QLSaCa", "1dks1zrZRdN", "g5UyJ3eArE", "uorvFYHOyL", "kXEG2xsYY-d", "jTVQpziG_Kk", "HrkimuDCAlN", "XlXob-ZJhvE", "6UspeQ2Trez",...	[ "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "official_reviewer", "official_reviewer", "author", "official_reviewer", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", ...	[ " Before this discussion ends, we would like to thank all the reviewers for their positive and helpful feedback and to summarize our current perspective. As indicated by the reviewers' responses so far, we believe we have clarified all the reviewers' concerns, except one: the issue of data movement, raised by revie...	[ -1, -1, -1, -1, -1, -1, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, 5, 6 ]	[ -1, -1, -1, -1, -1, -1, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4, 4 ]	[ "iclr_2022_clwYez4n8e8", "tZEPUcASz4", "1dks1zrZRdN", "kXEG2xsYY-d", "OJxXzG5BtbS", "i3Qud3KPH59", "i3Qud3KPH59", "iclr_2022_clwYez4n8e8", "i3Qud3KPH59", "xUQn0QLSaCa", "HrkimuDCAlN", "g5UyJ3eArE", "6UspeQ2Trez", "i3Qud3KPH59", "AF9gPAFVip4", "fOa8Aab2yxk", "fOa8Aab2yxk", "37HR_5dg...
iclr_2022_UF5cHSBycOt	Learning to Pool in Graph Neural Networks for Extrapolation	Graph neural networks (GNNs) are one of the most popular approaches to using deep learning on graph-structured data, and they have shown state-of-the-art performances on a variety of tasks. However, according to a recent study, a careful choice of pooling functions, which are used for the aggregation and readout operat...	Reject	This paper focuses on the extrapolation ability of graph neural networks and proposes a new pooling function based on vector norm. The proposed method can be applied to replace the commonly used pooling function like max/mean/sum, and is proved able for extrapolation in a simple example. Overall, all reviewers tend t...	test	[ "MnK0v34wG3A", "qQCdqd3sxXc", "uml31o7wWB2", "74qQV7ARF3I", "MYxStq-Za1j", "1qu56IyzC6y", "GHSJimsV935", "zr0ve1Qi_zA", "1Bc8qTZAR6L", "LPVr52IfADW", "NIcZL8Jwo-7", "Ht68D6afPWE", "QodWFRzLLAL", "puu8VmzgKEo", "41wpobrjd--", "4qTjNOVT4M", "Ixn9lHB-ah", "IUo8xQegW6w" ]	[ "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "official_reviewer", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " We would like to emphasize that our main contribution is extrapolation, rather than achieving the highest score in benchmarking tasks unrelated to extrapolation. In this work, we verified that simply replacing pooling functions with GNP enables GNNs to extrapolate well on many node-level, graph-level, and set...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 3, 3, 5 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 4, 4, 4 ]	[ "iclr_2022_UF5cHSBycOt", "Ht68D6afPWE", "NIcZL8Jwo-7", "QodWFRzLLAL", "1qu56IyzC6y", "GHSJimsV935", "zr0ve1Qi_zA", "LPVr52IfADW", "puu8VmzgKEo", "IUo8xQegW6w", "Ixn9lHB-ah", "4qTjNOVT4M", "41wpobrjd--", "iclr_2022_UF5cHSBycOt", "iclr_2022_UF5cHSBycOt", "iclr_2022_UF5cHSBycOt", "iclr_...
iclr_2022_3MjOIZ2CF9	An evaluation of quality and robustness of smoothed explanations	Explanation methods play a crucial role in helping to understand the decisions of deep neural networks (DNNs) to develop trust that is critical for the adoption of predictive models. However, explanation methods are easily manipulated through visually imperceptible perturbations that generate misleading explanations. T...	Reject	The paper conducts a series of empirical studies to evaluate the robustness of smoothed attribution methods. Although the reviewers think this is an important direction, there are several concerns about the experimental settings, such as the sample size and the models to be tested. Also, one of the main finding that Lp...	train	[ "DO5qeXLAOJd", "FpejhkxUVZ5", "bU1c7s6i57p", "5AEPo7_Z5x", "CuxckkZV6u8", "npdstUQdYyb", "p39aIguN1FB", "AbEL5cfZaq", "dhVSttjRPRj", "WXjbn7xORsX", "69xrWvzGTzt", "Mdo7HWDdPAo", "sX_CN369RXm", "0wvWru6kLCA", "9zZoUqu4-rv", "clx0T6GO-np" ]	[ "author", "official_reviewer", "author", "author", "official_reviewer", "author", "official_reviewer", "author", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Dear reviewer,\n\nThanks for your reply.\n\nThe smoothing methods we investigated in this work, did not have an explicit motivation to be only robust against explanation attacks based on additive models. $\\beta$-smoothing [2] method performs a global smoothing and smooth gradient, and uniform gradient [1] perfor...	[ -1, -1, -1, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, 6, 5, 5 ]	[ -1, -1, -1, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4, 4 ]	[ "FpejhkxUVZ5", "Mdo7HWDdPAo", "69xrWvzGTzt", "Mdo7HWDdPAo", "iclr_2022_3MjOIZ2CF9", "p39aIguN1FB", "AbEL5cfZaq", "dhVSttjRPRj", "WXjbn7xORsX", "CuxckkZV6u8", "clx0T6GO-np", "9zZoUqu4-rv", "0wvWru6kLCA", "iclr_2022_3MjOIZ2CF9", "iclr_2022_3MjOIZ2CF9", "iclr_2022_3MjOIZ2CF9" ]
iclr_2022_Rx9luEzcSoy	Lottery Image Prior	Deep Neural Networks (DNNs), either pre-trained (e.g., GAN generator) or untrained (e.g., deep image prior), could act as overparameterized image priors that help solve various image inverse problems. Since traditional image priors have much fewer parameters, those DNN-based priors naturally invite the curious question...	Reject	The paper studies the lottery ticket hypothesis in the context of deep image priors. Deep image priors are convolutional neural networks that are imposed as a prior for image reconstruction problems. A deep image prior can be an un-trained convolutional network, or it can be a trained generator, and the paper considers...	train	[ "J4HjKnz3rWM", "5FDvke7U-yx", "8_rxM4cKPYn", "dRc7XlXIUkk", "HxgChH0m2E", "5uXXjikU6hL", "Ft4KPEhzi0", "EoPaphyo2RZ", "p1xfNm3AXD", "SYJB4oSMlE0", "7i69Xnj7fZk", "pOVmujQ3qK" ]	[ "author", "official_reviewer", "official_reviewer", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ " Thanks for your constructive feedback and positive re-evaluation of our work.\n\n1. We added visualization results on the frequency intensities of the Fourier transformation of the Baby image from Set-5 (Figure 15 at the end of the Supplementary). Specifically, we performed the visualization for the ground-truth ...	[ -1, 6, -1, -1, -1, 6, -1, -1, -1, -1, 8, 6 ]	[ -1, 3, -1, -1, -1, 4, -1, -1, -1, -1, 5, 4 ]	[ "8_rxM4cKPYn", "iclr_2022_Rx9luEzcSoy", "dRc7XlXIUkk", "5FDvke7U-yx", "SYJB4oSMlE0", "iclr_2022_Rx9luEzcSoy", "5uXXjikU6hL", "5FDvke7U-yx", "7i69Xnj7fZk", "pOVmujQ3qK", "iclr_2022_Rx9luEzcSoy", "iclr_2022_Rx9luEzcSoy" ]
iclr_2022_5sP_PUUS78v	SeqPATE: Differentially Private Text Generation via Knowledge Distillation	Protecting the privacy of user data is crucial when training neural text generation models, which may leak sensitive user information during generation. Differentially private (DP) learning algorithms provide guarantees on identifying the existence of a training sample from model outputs. PATE is a DP learning algorith...	Reject	This was a very borderline decision. Here are the major factors involved in the decision. 1. The concurrent works by Li et al and Yu et al. It is unclear about the relationship/strength between those results and the ones in the present paper. However, in accordance with the ICLR policy on simultaneous work, we ignore...	train	[ "ZmG14SUTvWN", "njrkrzRTbJF", "WsCbxomPo_M", "AUcOnQJwD8I", "EDfj5FJVm24", "I07KKPAQ6ih", "uHlCcHZrDjz", "mpsDOFMDeds", "Ed1HXlWSfML", "YJOOZgGTBP", "nMLjd2T51j_", "0DkfxeM6D6", "KZdUysUc74", "AiTt9vPBe2N", "pius3i2DTi1", "o2ZhHve7paD", "gCfH32rb70", "0pIFqY2iAsh", "CkQogpVk-g-",...	[ "author", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "official_reviewer", "author", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer",...	[ " Thank you very much for your positive comments and suggestions!\n\nAs you mentioned, we've clarified most of your concerns, and the new experiments look convincing. May I know if there is anything else we need to address and make the paper more satisfactory to you? I am willing to solve any concern or problem...	[ -1, -1, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 6, 8 ]	[ -1, -1, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 3, 4 ]	[ "nMLjd2T51j_", "WsCbxomPo_M", "EDfj5FJVm24", "iclr_2022_5sP_PUUS78v", "uHlCcHZrDjz", "W71Y7i2Fmjw", "mpsDOFMDeds", "07UpZraqbr", "0DkfxeM6D6", "Tms4vZ-U6Q9", "0DkfxeM6D6", "0pIFqY2iAsh", "CkQogpVk-g-", "XN1TskBXRdf", "o2ZhHve7paD", "gCfH32rb70", "W71Y7i2Fmjw", "APdZbRppfBm", "AYw...
iclr_2022_ZUXZKjfptc9	Bit-aware Randomized Response for Local Differential Privacy in Federated Learning	In this paper, we develop BitRand, a bit-aware randomized response algorithm, to preserve local differential privacy (LDP) in federated learning (FL). We encode embedded features extracted from clients' local data into binary encoding bits, in which different bits have different impacts on the embedded features. Based ...	Reject	The paper proposes BitRand, a bit-aware randomized response algorithm, to preserve local differential privacy in federated learning. The main idea is to take into account the bit indices and prioritise higher order bits focussed towards achieving a utility which is higher than other algorithms which are oblivious to th...	train	[ "cM0vSl96TzU", "enpXEjtkxej", "h4IvnjFSlMA", "toAERPga4gQ", "4Uz9tuzOw6", "mRN_eMJ5KSp", "jCtdCcoqXxc", "Nv0tTTMPs5", "GPRjPo-BMk", "IWE_9ECGF5t", "c8Uf8I4aWf", "WCvNGr4_cx-", "GFwxn1P-j9I", "pW-mLKd6RKS", "25-iBf9lQoH", "0VT6MHt0jRh", "AWtp2QGPLgH", "wcVDQQEXN9P", "twdaiUj1sa-",...	[ "author", "official_reviewer", "official_reviewer", "author", "author", "official_reviewer", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", ...	[ " We are thankful for the reviewer's feedback and do appreciate for recognizing the significance of our work. We will reorganize our paper to highlight the core contribution of our works better.", "The paper tries to address the question whether we can have different noise scale for different bits for a high dime...	[ -1, 6, -1, -1, -1, -1, -1, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, 3 ]	[ -1, 3, -1, -1, -1, -1, -1, -1, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 4 ]	[ "h4IvnjFSlMA", "iclr_2022_ZUXZKjfptc9", "gPOryOcI42", "GPRjPo-BMk", "mRN_eMJ5KSp", "twdaiUj1sa-", "Nv0tTTMPs5", "E_h7cHKHPuA", "iclr_2022_ZUXZKjfptc9", "GFwxn1P-j9I", "iclr_2022_ZUXZKjfptc9", "E_h7cHKHPuA", "25-iBf9lQoH", "E_h7cHKHPuA", "GPRjPo-BMk", "Sn3t12Idnaz", "Sn3t12Idnaz", "...
iclr_2022_dLDzuxaN0Hd	Unsupervised Pose-Aware Part Decomposition for 3D Articulated Objects	Articulated objects exist widely in the real world. However, previous 3D generative methods for unsupervised part decomposition are unsuitable for such objects, because they assume a spatially fixed part location, resulting in inconsistent part parsing. In this paper, we propose PPD (unsupervised Pose-aware Part Decomp...	Reject	The submission received split reviews: two reviewers recommended accepts, and the other two rejects. The AC went through the reviews, responses, and discussions carefully. The AC appreciates the authors' effort during the response period and agreed that the revision has addressed some of the concerns of the reviewers. ...	train	[ "dA3qgqfPVO5", "aAFdZFu2TWV", "2KtdFlRFLze", "nsu7iRJwEE-", "AUNQ_26lEd", "ckjSsqAMs-j", "S1IX95F337F", "jqDaaodl2C3", "RFRguE-xti8", "7Zsg0jF6i-G", "sjgcW_R4uZ", "VoOYKXgEfWJ", "vxKnUpP66Sc", "xfG8ikvATpO" ]	[ "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " We would like to thank all the reviewers for the suggestions and questions to improve our paper.\n\n### Major revision\n\n1. Added the new baseline, Neural Parts [Paschalidou+ 2021] ($\\textcolor{red}{R2}$, $\\textcolor{green}{R3}$, $\\textcolor{magenta}{R4}$).\n2. Eased claim on generative applications ($\\textc...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 8, 5, 8, 5 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4, 2, 4 ]	[ "iclr_2022_dLDzuxaN0Hd", "2KtdFlRFLze", "xfG8ikvATpO", "vxKnUpP66Sc", "ckjSsqAMs-j", "S1IX95F337F", "jqDaaodl2C3", "VoOYKXgEfWJ", "sjgcW_R4uZ", "iclr_2022_dLDzuxaN0Hd", "iclr_2022_dLDzuxaN0Hd", "iclr_2022_dLDzuxaN0Hd", "iclr_2022_dLDzuxaN0Hd", "iclr_2022_dLDzuxaN0Hd" ]
iclr_2022_0J98XyjlQ1	D$^2$-GCN: Data-Dependent GCNs for Boosting Both Efficiency and Scalability	Graph Convolutional Networks (GCNs) have gained an increasing attention thanks to their state-of-the-art (SOTA) performance in graph-based learning tasks. However, their sheer number of node features and large adjacency matrix limit their deployment into real-world applications, as they impose the following challenges:...	Reject	The paper proposes Data-Dependent GCN (D2-GCN), which improves the efficiency of vanilla GCN by node-wise skipping, edgewise skipping, and bit-wise skipping. Gate functions are learned to prune the unimportant neighbor nodes in combinations, unimportant edge connections, and in the bit-precision. The proposed method bo...	train	[ "2JPlqUFjyjO", "S7GnF7T9OAd", "CLfQfs8ohPO", "47awNPkJFp3", "aycF7YDG0TR", "B_4Z6nWxsry", "OvwVTVl4Nc1", "WwuOrQJL9V", "9QOVUU6cTef", "_Gz07lJAKGQ", "KfK8uis-PnP", "YPGAL29oQ1y", "6mF3Wr6ni3v", "ELPn2TQIW1s" ]	[ "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ "The paper attempts to address the training efficiency and scalability of GCNs. To be specific, a so-called Data-Dependent dynamic GCN framework is proposed, in which node-wise skepping, edge-wise skipping, an bit-wise skipping are integrated via gate function to squeeze out (or reduce) the unimportant neighbor no...	[ 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 6, 3 ]	[ 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 4, 3 ]	[ "iclr_2022_0J98XyjlQ1", "B_4Z6nWxsry", "47awNPkJFp3", "_Gz07lJAKGQ", "KfK8uis-PnP", "2JPlqUFjyjO", "6mF3Wr6ni3v", "9QOVUU6cTef", "YPGAL29oQ1y", "KfK8uis-PnP", "ELPn2TQIW1s", "iclr_2022_0J98XyjlQ1", "iclr_2022_0J98XyjlQ1", "iclr_2022_0J98XyjlQ1" ]
iclr_2022_JVWB8QRUOi-	Learning Homophilic Incentives in Sequential Social Dilemmas	Promoting cooperation among self-interested agents is a long-standing and interdisciplinary problem, but receives less attention in multi-agent reinforcement learning (MARL). Game-theoretical studies reveal that altruistic incentives are critical to the emergence of cooperation but their analyses are limited to non-seq...	Reject	This paper propose a novel framework to increase cooperation in second-order social dilemmas. This is based on encouraging homophilic incentives. Reviewers agree that the paper does not meet the standards of publication yet. In particular, they worry that the assumptions made are so restrictive as to make model inappli...	train	[ "xdzrusDUdQz", "IgcIqvKUCwE", "KfIcVy-2Ubs", "7Oks6LcAdWS", "ZK2jwAkpuJ", "mx3nkqMnKu", "efGmExFsFUQ" ]	[ "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " We thank all reviewers for their reviews and suggestions of our work. All the comments are helpful and have been reflected in the updated paper. Here we briefly list the updates.\n\n[Appendix B.4] To show the mechanisms behind other baselines, we plot the gradient field of Inequity Aversion.\n\n**[Appendix B....	[ -1, -1, -1, -1, 5, 3, 3 ]	[ -1, -1, -1, -1, 3, 5, 4 ]	[ "iclr_2022_JVWB8QRUOi-", "efGmExFsFUQ", "mx3nkqMnKu", "ZK2jwAkpuJ", "iclr_2022_JVWB8QRUOi-", "iclr_2022_JVWB8QRUOi-", "iclr_2022_JVWB8QRUOi-" ]
iclr_2022_2RNpZ8S4alJ	KINet: Keypoint Interaction Networks for Unsupervised Forward Modeling	Object-centric representation is an essential abstraction for physical reasoning and forward prediction. Most existing approaches learn this representation through extensive supervision (e.g, object class and bounding box) although such ground-truth information is not readily accessible in reality. To address this, we ...	Reject	This submission received four high-quality reviews. After the discussion period, all reviewers agreed that this submission is not strong enough to be accepted. Concerns include the novelty of the proposed method wrt related work and the limited experiments. The AC agrees. The AC also finds it disappointing that the aut...	train	[ "c_kFaDZaMGQ", "xzX-gN1N2oS", "GwR92LVh8bH", "8o_MZruQN2L", "a51D-lEJrM", "0dkha2BBfos", "GxJdjlipD1Q", "kmt9eWrphT", "jJgTGtGbTe", "6jtCwyGVpad", "A54-tVW5ER4", "7t5g70Aq_V", "w6ewb9txIjE", "vXG3_v-WfcH" ]	[ "official_reviewer", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Overall, I think what the paper does is quite similar to V-CDN, and making it end-to-end does not seem to be a significant contribution.\n\nRegarding the inference of adjacency matrix, in Variational Graph Auto-Encoders, the groundtruth adjacency matrix is provided, so the decoder can be trained to reconstruct th...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 3, 5, 3 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4, 4, 4 ]	[ "6jtCwyGVpad", "8o_MZruQN2L", "vXG3_v-WfcH", "GwR92LVh8bH", "w6ewb9txIjE", "w6ewb9txIjE", "w6ewb9txIjE", "7t5g70Aq_V", "A54-tVW5ER4", "A54-tVW5ER4", "iclr_2022_2RNpZ8S4alJ", "iclr_2022_2RNpZ8S4alJ", "iclr_2022_2RNpZ8S4alJ", "iclr_2022_2RNpZ8S4alJ" ]
iclr_2022_dK_t8oN8G4	Neurosymbolic Deep Generative Models for Sequence Data with Relational Constraints	There has been significant recent 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, and many such models perform well on local coherence at the cost of global cohe...	Reject	This work proposes an interesting approach for learning the relational constraints of a dataset and then generating according to those constraints. Learning the constraints via a constrained optimization problem is an interesting contribution. The application of constrained generation is also interesting and can be app...	train	[ "6g5FbGHp52v", "Q1nVfTiyHBA", "P6V018IPl0c", "veJA89KEn3", "JtlLszZ07F", "b1uCJE1wuF", "PyIBFYc7nOO", "XO57qjTMzR", "ZngIfqg9FUn" ]	[ "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Regarding applying our technique to other domains: Our experiments are tailored to tasks where there are syntactic relationships between different parts of the input, which occur naturally in music and poetry. Both these directions are important applications in their own right; for instance, there is a great deal...	[ -1, -1, -1, -1, -1, 6, 6, 5, 5 ]	[ -1, -1, -1, -1, -1, 3, 4, 4, 3 ]	[ "ZngIfqg9FUn", "XO57qjTMzR", "PyIBFYc7nOO", "b1uCJE1wuF", "iclr_2022_dK_t8oN8G4", "iclr_2022_dK_t8oN8G4", "iclr_2022_dK_t8oN8G4", "iclr_2022_dK_t8oN8G4", "iclr_2022_dK_t8oN8G4" ]
iclr_2022_Vy5WbmrVPaD	Pretext Tasks Selection for Multitask Self-Supervised Speech Representation Learning	Through solving pretext tasks, self-supervised learning leverages unlabeled data to extract useful latent representations replacing traditional input features in the downstream task. In audio/speech signal processing, a wide range of features where engineered through decades of research efforts. As it turns out, learni...	Reject	The paper proposes a method for selecting a group of pretext tasks out of a set of candidates in order to optimize self training for downstream performance. The method relies on Hilbert Schmidt Independence Criterion (HSIC) and uses a few data samples to select weights for the given set of tasks The paper demonstrates ...	train	[ "oI9NbpHeCY", "xsOcHkwaQl_", "zyZZXVkaf4R", "nzPVizifsXO", "HDGPERnf1kx", "ch0rKEDJmwI", "PQZNB6xGy5v", "ioziMXkwZSk", "88v4otCNY4g", "RC0YOYH-KcM" ]	[ "author", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Thank you first for your reply. Here are precise answers to dissipate your final concerns.\n\n> Comment: One of the claims of the paper is that adapting the choices in the self supervised pipeline to a specific downstream task improves the final downstream performance. I agree with the claim that refining the pre...	[ -1, -1, -1, -1, -1, -1, -1, 3, 6, 3 ]	[ -1, -1, -1, -1, -1, -1, -1, 5, 3, 3 ]	[ "xsOcHkwaQl_", "nzPVizifsXO", "ioziMXkwZSk", "ioziMXkwZSk", "iclr_2022_Vy5WbmrVPaD", "RC0YOYH-KcM", "88v4otCNY4g", "iclr_2022_Vy5WbmrVPaD", "iclr_2022_Vy5WbmrVPaD", "iclr_2022_Vy5WbmrVPaD" ]
iclr_2022_rbPg0zkHGi	Deep Active Learning with Noise Stability	Uncertainty estimation for unlabeled data is crucial to active learning. With a deep neural network employed as the backbone model, the data selection process is highly challenged due to the potential over-confidence of the model inference. Existing methods usually resort to multi-pass model training or adversarial tra...	Reject	The submission considers a new acquisition function for active learning. The method considers the sensitivity of the prediction for a given datapoint with respect to parameter perturbations. Points with the largest variance under these perturbations are selected for labelling. The method is simple and the empirical re...	val	[ "1jA9_aEPcTR", "JoCCDp_ZhF", "IQC--4QY0-2", "Cza6gCZn3Xp", "uVW9T81JG6c", "acN1pH-p3Nm", "fcXl8iz3KTM", "ADrD1bqkxqa", "Na5GAkymtsY", "Hmh9oZZcdUK" ]	[ "official_reviewer", "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer" ]	[ " I appreciate the authors' efforts in answering my questions and adding new numbers. \n\nHowever, the concerns are still there, so I will keep my score. \n\n- There are partial comparisons on Imagenet. I understand there was not enough time, but that should not be an excuse. I would encourage completing these resu...	[ -1, -1, 6, 5, -1, -1, -1, -1, -1, 3 ]	[ -1, -1, 4, 3, -1, -1, -1, -1, -1, 5 ]	[ "acN1pH-p3Nm", "uVW9T81JG6c", "iclr_2022_rbPg0zkHGi", "iclr_2022_rbPg0zkHGi", "Na5GAkymtsY", "Hmh9oZZcdUK", "iclr_2022_rbPg0zkHGi", "IQC--4QY0-2", "Cza6gCZn3Xp", "iclr_2022_rbPg0zkHGi" ]
iclr_2022_morSrUyWG26	AutoOED: Automated Optimal Experimental Design Platform with Data- and Time-Efficient Multi-Objective Optimization	We present AutoOED, an Automated Optimal Experimental Design platform powered by machine learning to accelerate discovering solutions with optimal objective trade-offs. To solve expensive multi-objective problems in a data-efficient manner, we implement popular multi-objective Bayesian optimization (MOBO) algorithms wi...	Reject	The reviews are of good quality. The responses by the authors are commendable, but reviewers remain of the opinion that the scientific contribution of the paper is limited, no matter how strong the software engineering contribution may be.	train	[ "g8Cc97bI3x0", "ZHXl95aTMBb", "JqKjWWzhBUQ", "coVEOKMNMEV", "I1Gre2_c6cm", "Xry4pWYYXwa", "v4FVeBgguch", "ql6sYgBOxVN", "tJ8gjQfv9rO", "uWTo-xyrDP", "g-ATKhCc_6h", "FbPneI9lMi", "kv3slUskf_o" ]	[ "author", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "public", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Thank you for your thoughtful comments and suggestions!\n\nBefore clarifications of other technical capabilities, we would like to first emphasize that the GUI/visualization features are extremely important especially for general users (which are the majority of people who need OED), which is the primary motivati...	[ -1, -1, -1, 5, -1, -1, -1, -1, -1, -1, 6, 5, 5 ]	[ -1, -1, -1, 5, -1, -1, -1, -1, -1, -1, 3, 4, 5 ]	[ "uWTo-xyrDP", "iclr_2022_morSrUyWG26", "ql6sYgBOxVN", "iclr_2022_morSrUyWG26", "kv3slUskf_o", "FbPneI9lMi", "g-ATKhCc_6h", "coVEOKMNMEV", "iclr_2022_morSrUyWG26", "iclr_2022_morSrUyWG26", "iclr_2022_morSrUyWG26", "iclr_2022_morSrUyWG26", "iclr_2022_morSrUyWG26" ]
iclr_2022_ibNr25jJrf	Direct Evolutionary Optimization of Variational Autoencoders With Binary Latents	Many types of data are generated at least partly by discrete causes that are sparsely active. To model such data, we here investigate a deep generative model in the form of a variational autoencoder (VAE) which can learn a sparse, binary code for its latents. Because of the latents’ discrete nature, standard VAE traini...	Reject	This paper presents a new approach for learning binary latent variable models using evolutionary optimization. Pros: * A new perspective to learning binary latent variables is proposed using evolutionary algorithms. * The proposed method works well on auxiliary tasks such as zero-shot denoising and inpainting. Cons:...	train	[ "4bCqixoHcnm", "NRI2A5wkdz", "jLrUbCgDbjv", "Bwf0vqE4dJ0", "95-QC-lEghH", "uIRuTFOGO4L", "gMuvZVv3fL_", "jV6Wy0Y0f2", "oBusEqQje3k", "7RhE8vkv9e2", "X3Gvk4-BNlM", "WpNygmNoqsw", "YsbeIiUhQuX" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ "This paper proposes an approach to learn VAEs with binary latent variables. Rather than relying on commonly used continuous relaxation techniques, the authors completely forgo the encoder neural network and parameterize the approximate posterior using a subset of states, so that the approximate posterior is propor...	[ 5, -1, 8, -1, -1, -1, -1, -1, -1, -1, 6, 8, 5 ]	[ 4, -1, 4, -1, -1, -1, -1, -1, -1, -1, 2, 3, 4 ]	[ "iclr_2022_ibNr25jJrf", "gMuvZVv3fL_", "iclr_2022_ibNr25jJrf", "uIRuTFOGO4L", "4bCqixoHcnm", "YsbeIiUhQuX", "jLrUbCgDbjv", "X3Gvk4-BNlM", "WpNygmNoqsw", "iclr_2022_ibNr25jJrf", "iclr_2022_ibNr25jJrf", "iclr_2022_ibNr25jJrf", "iclr_2022_ibNr25jJrf" ]
iclr_2022_46lmrnVBHBL	Explanatory Learning: Beyond Empiricism in Neural Networks	We introduce Explanatory Learning (EL), an explanation-driven machine learning framework to use existing knowledge buried in symbolic sequences expressed in an unknown language. In EL, the burden of interpreting explanations is not left to humans or human-coded compilers, as done in Program Synthesis. Rather, EL calls ...	Reject	This paper presents an explanation-based learning approach that learns from both observations (examples) and explanations paired with examples. It proposes to learn an interpreter that can map from natural language sentences to examples. The authors also develop an evaluation environment and protocols for the tasks. S...	train	[ "lWAJIHqUuxm", "Xvooir0jeKw", "QSeKKdyrFl", "nA2I3_8y34d", "qs3d7g_Lg-R", "JXWstTzxy4x", "RwFTkzBY6l7", "-ak4svuFJnF", "CMkTxVrI-02", "E0NydlafUIi", "Pp3mc_v0s0X", "Dg9dwUel3C", "dn4i8wQeWZu", "ycV2WZFpV3s", "GQe6XBHtK55", "Zr3AJ3S8ArV", "xDBGlCI1bK", "7TeA54ZbKwP", "JuWRoCKfuG2"...	[ "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " We thank the reviewer for the relevant reference and for engaging in the fascinating discussion about the nature of language learning. This is indeed one of the main goals of our paper; we wish to generate interest in this discussion among as many ML researchers as possible.\nWe believe that we addressed the othe...	[ -1, -1, -1, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 8, 3 ]	[ -1, -1, -1, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 2, 4, 4 ]	[ "QSeKKdyrFl", "nA2I3_8y34d", "Xvooir0jeKw", "7TeA54ZbKwP", "iclr_2022_46lmrnVBHBL", "iclr_2022_46lmrnVBHBL", "E0NydlafUIi", "CMkTxVrI-02", "qs3d7g_Lg-R", "-ak4svuFJnF", "RwFTkzBY6l7", "dn4i8wQeWZu", "C-6LEcN_gZ0", "GQe6XBHtK55", "tqDYuVHUrFS", "xDBGlCI1bK", "7TeA54ZbKwP", "JuWRoCKf...
iclr_2022_Zae_OHNq-y	Imbalanced Adversarial Training with Reweighting	Adversarial training has been empirically proven to be one of the most effective and reliable defense methods against adversarial attacks. However, the majority of existing studies are focused on balanced datasets, where each class has a similar amount of training examples. Research on adversarial training with imbalan...	Reject	Three out of the four reviewers raised various concerns on motivation clarify, result significance, and unclear writing. While the authors provided their rebuttals, unfortunately no reviewer seems to have changed their mind. AC reads the paper and agreed this paper perhaps needs major revision before publishing in a ma...	train	[ "x_PF35xVf88", "d-FUZB782A-", "r2w5Jm_5Pr3", "y34JiseEJOC", "JRd64FsWTVh", "sEFjvSmhezv", "VCZDjsMRuQc", "Kmjs5rf5624", "disyzGvYIYX", "r4XC0CJvUc", "Bt1BXVPvjJ", "y4uYjWJLlmv" ]	[ "official_reviewer", "author", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Thank the authors for the detailed response.", " Thanks for your valuable feedback. We address them in detail as follows.\n\nQ1. Proposed model is only evaluated on one dataset, CIFAR10. \n\nAnswer: Due to the limited space, we only provided part of experimental results on imbalanced CIFAR10 dataset in the...	[ -1, -1, -1, -1, 3, -1, -1, -1, -1, 3, 3, 8 ]	[ -1, -1, -1, -1, 4, -1, -1, -1, -1, 4, 4, 4 ]	[ "disyzGvYIYX", "r4XC0CJvUc", "y34JiseEJOC", "VCZDjsMRuQc", "iclr_2022_Zae_OHNq-y", "VCZDjsMRuQc", "Bt1BXVPvjJ", "JRd64FsWTVh", "y4uYjWJLlmv", "iclr_2022_Zae_OHNq-y", "iclr_2022_Zae_OHNq-y", "iclr_2022_Zae_OHNq-y" ]
iclr_2022_eZ-xMLuKPc	Surgical Prediction with Interpretable Latent Representation	Given the risks and cost of surgeries, there has been significant interest in exploiting predictive models to improve perioperative care. However, due to the high dimensionality and noisiness of perioperative data, it is challenging to develop accurate, robust and interpretable encoding for surgical applications. We pr...	Reject	The reviews are of good quality. The responses by the authors are commendable, but ICLR is selective and reviewers still believe that the research would be better as two separate papers: one about the problem and solution from an ML perspective, and the other about the application to surgery. Papers that provide a new ...	val	[ "qWG0uSvWeM", "jcVYPrujNor", "os5hb0xnblI", "QRNd-jUXNwl", "JCe0Kw5pDg3", "2OQx3-59Vy", "G7HM45GiIh7" ]	[ "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " We appreciate the valuable feedback from reviewer CEnW and we are happy to use this chance to discuss further. Here we summarize the questions by reviewer CEnW and try our best to answer each of them. \n\n\nComment 1.1. The marginal improvement of the proposed methods over the previously published methods may no...	[ -1, -1, -1, -1, 6, 5, 3 ]	[ -1, -1, -1, -1, 4, 4, 4 ]	[ "JCe0Kw5pDg3", "G7HM45GiIh7", "jcVYPrujNor", "2OQx3-59Vy", "iclr_2022_eZ-xMLuKPc", "iclr_2022_eZ-xMLuKPc", "iclr_2022_eZ-xMLuKPc" ]
iclr_2022_r9cpyzP-DQ	Learning Efficient and Robust Ordinary Differential Equations via Diffeomorphisms	Advances in differentiable numerical integrators have enabled the use of gradient descent techniques to learn ordinary differential equations (ODEs), where a flexible function approximator (often a neural network) is used to estimate the system dynamics, given as a time derivative. However, these integrators can be uns...	Reject	This submission proposes a new manner to learn ordinary differential equations, aiming to improve their efficiency. While judging it interesting, the reviewers are quite split on this work. Overall there was no strong consensus to accept, nor anyone willing to champion this work. The main stated weaknesses are - The ...	train	[ "inY5epg89J", "EkOgQSuqb2", "0Wrnrst6_kp", "OoU_UPzEYm", "bFvZ-Y_RfsP", "7vmlnoh_pEz", "jvY4vFj1TWL", "CLPk5Mi9WjH", "2fWBezbXWUa" ]	[ "author", "author", "author", "official_reviewer", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " We would like to thank you for your positive feedback, and use this opportunity to respond to some of the comments and clarify any misunderstandings.\n\n> “from the limited experiments presented it is not clear how applicable the method is for complex practical problems”\n\nWe'd like to clarify to the reviewer th...	[ -1, -1, -1, -1, -1, 6, 3, 6, 3 ]	[ -1, -1, -1, -1, -1, 3, 4, 5, 4 ]	[ "7vmlnoh_pEz", "2fWBezbXWUa", "CLPk5Mi9WjH", "bFvZ-Y_RfsP", "jvY4vFj1TWL", "iclr_2022_r9cpyzP-DQ", "iclr_2022_r9cpyzP-DQ", "iclr_2022_r9cpyzP-DQ", "iclr_2022_r9cpyzP-DQ" ]
iclr_2022_T_p1vd88T87	Neural Implicit Representations for Physical Parameter Inference from a Single Video	Neural networks have recently been used to model the dynamics of diverse physical systems. While existing methods achieve impressive results, they are limited by their strong demand for training data and their weak generalization abilities. To overcome these limitations, in this work we propose to combine neural implic...	Reject	The paper proposes a framework for learning the physical parameters of a physical system’s dynamics from a video. The model combines a differentiable neural ODE solver (NODE) with neural implicit representations through a local coordinate-based network which reconstruct the frames based on the ODE solution. Both the st...	train	[ "MQrx20EUBM2", "dwll6QDtQhU", "jVtiLx855xm", "fXOZDmi1W1S", "lmsma1fJoHQ", "yNjYknjH8I8", "b69hKSJDL9i", "CHx7BQrzXWk", "SD-jclBhzrR", "QHtDSYwMy5L", "lW7dO34aBV2", "dvpWVrkhAkH", "uRdiK7C5PYA", "4WSERryawXB" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ "The paper proposes a method which is capable of obtaining physical models of motion using only visual observations of the motion from a single video (no training data). This is done by combining a differentiable ODE solver with a coordinate-based network which reconstructs the video frames based on the ODE solutio...	[ 6, -1, -1, -1, 5, -1, -1, -1, -1, -1, -1, 3, 3, 3 ]	[ 2, -1, -1, -1, 3, -1, -1, -1, -1, -1, -1, 3, 4, 3 ]	[ "iclr_2022_T_p1vd88T87", "yNjYknjH8I8", "CHx7BQrzXWk", "QHtDSYwMy5L", "iclr_2022_T_p1vd88T87", "MQrx20EUBM2", "4WSERryawXB", "uRdiK7C5PYA", "dvpWVrkhAkH", "lmsma1fJoHQ", "iclr_2022_T_p1vd88T87", "iclr_2022_T_p1vd88T87", "iclr_2022_T_p1vd88T87", "iclr_2022_T_p1vd88T87" ]
iclr_2022_rGg-Qcyplgq	Distributional Perturbation for Efficient Exploration in Distributional Reinforcement Learning	Distributional reinforcement learning aims to learn distribution of return under stochastic environments. Since the learned distribution of return contains rich information about the stochasticity of the environment, previous studies have relied on descriptive statistics, such as standard deviation, for optimism in fac...	Reject	This paper studies efficient algorithms for distributional reinforcement learning. The motivation stems from the need of risk neutrality, since other existing approaches might have one-sided risk tendencies. The algorithms proposed in this paper are based on sampling from a distributional perturbation rather than using...	train	[ "1JEZmD38lVB", "pJuaai396Szb", "WpvUbr5QOYv", "yp-GFhA-kZQo", "h14-a2Yfk8U", "ChNFEqGHMwN", "NCQZpALE3wx", "OGspts0Wu2B" ]	[ "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " We thank all the reviewers for their valuable comments and constructive feedback!\n\nWe highlighted the revised sentences in red color for the updated version.\n\n* Clearer explanation and correct the grammatical errors for all reviewers.\n* Clear notation and proof with better readability for **reviewer 61i6...	[ -1, -1, -1, -1, -1, 6, 5, 6 ]	[ -1, -1, -1, -1, -1, 3, 4, 3 ]	[ "iclr_2022_rGg-Qcyplgq", "OGspts0Wu2B", "NCQZpALE3wx", "NCQZpALE3wx", "ChNFEqGHMwN", "iclr_2022_rGg-Qcyplgq", "iclr_2022_rGg-Qcyplgq", "iclr_2022_rGg-Qcyplgq" ]
iclr_2022_ET1UAOYeU42	Edge Partition Modulated Graph Convolutional Networks	Graph convolutional networks (GCNs), which propagate the node features through the edges and learn how to transform the aggregated features under label supervision, have achieved great success in supervised feature extraction for both graph-level and node-level classification tasks. However, GCNs typically treat the gr...	Reject	The weaknesses of the paper can briefly be summarised as follows: i) the suggested motivation is not so clear, and in addition the experimental results (by themselves questionable in the way they are obtained) do not support the main claim of the paper that "...edges are generated by aggregating the node interactions o...	train	[ "6uV0rLrfLrE", "hBGCexSvzy7", "5a_qoEHaHcS", "gQtsuYhYiAm", "SQqRVwncNd4", "m3JxwhKb90a", "YseTAUAZ4jh", "6GmqqPd7-2u", "roKxApjXeqp", "jGhxmHzTVoi", "nubRJmmwQCv" ]	[ "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " We appreciate all reviewers for the comments and suggestions. Here we would like to clarify some concerns shared by more than one reviewer. \n\n* The motivation of method / architecture (LpaE, 6jg3, 1iDH):\n \n We find partitioning the graph into multiple graphs by relation and learning each relation with a sepa...	[ -1, -1, -1, -1, -1, -1, -1, 8, 5, 3, 3 ]	[ -1, -1, -1, -1, -1, -1, -1, 3, 3, 4, 5 ]	[ "iclr_2022_ET1UAOYeU42", "6GmqqPd7-2u", "nubRJmmwQCv", "nubRJmmwQCv", "nubRJmmwQCv", "jGhxmHzTVoi", "roKxApjXeqp", "iclr_2022_ET1UAOYeU42", "iclr_2022_ET1UAOYeU42", "iclr_2022_ET1UAOYeU42", "iclr_2022_ET1UAOYeU42" ]
iclr_2022_TEt7PsVZux6	I-PGD-AT: Efficient Adversarial Training via Imitating Iterative PGD Attack	Adversarial training has been widely used in various machine learning paradigms to improve the robustness; while it would increase the training cost due to the perturbation optimization process. To improve the efficiency, recent studies leverage Fast Gradient Sign Method with Random Start (FGSM-RS) for adversarial trai...	Reject	This paper aims to improve the efficiency of adversarial training. Specifically, by analyzing the differences between the adversarial perturbations generated by FGSM-RS and the adversarial perturbations generated by PGD, this paper proposes a new single-step attacker I-PGD (which imitates PGD by creating diverse advers...	train	[ "MWMYVLcSS6Q", "7-Tb2sB4ozH", "czqFhJqZ4gq", "r5u-Gq6LgdC", "BSfDeLFbpre", "1CiNDnJaPE-", "49HiMMZjlM9", "7QT4oCuZyHH", "C5vNCxnwHYn", "_lGSPrq8o75", "EWegQXiZOWY" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ " I thank the authors for their detailed response. However, the response still does not address some of my main concerns about the paper. \n\nFor Q1, the authors admit that \"in the approximation, it might go far away from the boundary with the iteration progressing on\". Then the authors argue that this will not c...	[ -1, 6, 5, -1, -1, -1, -1, -1, -1, 3, 6 ]	[ -1, 4, 5, -1, -1, -1, -1, -1, -1, 5, 4 ]	[ "C5vNCxnwHYn", "iclr_2022_TEt7PsVZux6", "iclr_2022_TEt7PsVZux6", "_lGSPrq8o75", "1CiNDnJaPE-", "EWegQXiZOWY", "7-Tb2sB4ozH", "czqFhJqZ4gq", "r5u-Gq6LgdC", "iclr_2022_TEt7PsVZux6", "iclr_2022_TEt7PsVZux6" ]
iclr_2022_qXa0nhTRZGV	Understanding Sharpness-Aware Minimization	Sharpness-Aware Minimization (SAM) is a recent training method that relies on worst-case weight perturbations. SAM significantly improves generalization in various settings, however, existing justifications for its success do not seem conclusive. First, we analyze the implicit bias of SAM over diagonal linear networks,...	Reject	As evident by the title the paper focuses on understanding sharpness-aware minimization which is a contemporary training procedure based on minimizing the worse case perturbation of the weights in ball. It has been observed that SAM improves the generalization and this paper aims to demystify this success. They also p...	train	[ "qnl5ONoMfTs", "TedlVgpEhMx", "Rd3f_CpOgM", "VVdiWPmLkes", "kGRTryM_6Pw", "FXpkaewTp94", "Uq_BAFMfBvH", "2KUxG_22A1", "yoi2e7HLLv", "h45YVWiF91o", "fiNE7UUeRtJ", "ZV7Pp9Kp9HI" ]	[ "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Here are additional results for unnormalized SAM for a ResNet-18 trained on CIFAR-100 using the same hyperparameters as in Figure 10 for CIFAR-10. We report the test error with early stopping based on a validation set:\n\n\nStandard CIFAR-100\n\n\| Method \| Test error \|\n\| -------- \|-------\|\n\| ERM \| 2...	[ -1, -1, -1, -1, -1, -1, -1, -1, 3, 6, 3, 3 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, 3, 4, 4, 3 ]	[ "TedlVgpEhMx", "Rd3f_CpOgM", "Uq_BAFMfBvH", "ZV7Pp9Kp9HI", "fiNE7UUeRtJ", "fiNE7UUeRtJ", "h45YVWiF91o", "yoi2e7HLLv", "iclr_2022_qXa0nhTRZGV", "iclr_2022_qXa0nhTRZGV", "iclr_2022_qXa0nhTRZGV", "iclr_2022_qXa0nhTRZGV" ]
iclr_2022_EMLJ_mTz_z	Convolutional Neural Network Dynamics: A Graph Perspective	The success of neural networks (NNs) in a wide range of applications has led to increased interest in understanding the underlying learning dynamics of these models. In this paper, we go beyond mere descriptions of the learning dynamics by taking a graph perspective and investigating the relationship between the graph ...	Reject	This paper proposes to represent a deep neural network as a graph and analyze its learning dynamics as a time series of weighted graphs corresponding to the neural network. As the graph representations, the authors propose to use a rolled representation in addition to a unrolled representation. Then, they proposed to u...	train	[ "Q5NeRs91Dck", "ds-HjgNKHrV", "z_U91umAmBI", "uXGOq-YK9YI", "oTduTQnG88l", "UJ5CKUrOukk", "6XZ-FUL-YFJ", "u3OOzrWLAl2", "bJo5eCo-7PT", "jEVpuJ8riGR", "Qan1BkfZmC", "aO2E-aaTlU5", "0JhaUOpsnR1", "aiIs0ChP9I", "7aaq-5JDNcq", "n8iXRfjaU3x", "WnMUHTF5Ygm", "vzwDzawyGhD", "3afiHwTtQcV...	[ "author", "official_reviewer", "official_reviewer", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Given that our paper updates and response adequately addressed the vast majority of your concerns, we are wondering if you would be willing to increase your score? Thank you again for your time and feedback.", "This paper tries to model the learning dynamics of a (deep) neural network using graphs. In particula...	[ -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 8, 5, 3 ]	[ -1, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 3, 5 ]	[ "UJ5CKUrOukk", "iclr_2022_EMLJ_mTz_z", "aO2E-aaTlU5", "UJ5CKUrOukk", "iclr_2022_EMLJ_mTz_z", "u3OOzrWLAl2", "3afiHwTtQcV", "bJo5eCo-7PT", "6XZ-FUL-YFJ", "Qan1BkfZmC", "vzwDzawyGhD", "0JhaUOpsnR1", "ds-HjgNKHrV", "WnMUHTF5Ygm", "iclr_2022_EMLJ_mTz_z", "iclr_2022_EMLJ_mTz_z", "iclr_202...
iclr_2022_UFYYol-bRq	ANCER: Anisotropic Certification via Sample-wise Volume Maximization	Randomized smoothing has recently emerged as an effective tool that enables certification of deep neural network classifiers at scale. All prior art on randomized smoothing has focused on isotropic $\ell_p$ certification, which has the advantage of yielding certificates that can be easily compared among isotropic metho...	Reject	The paper proposes the anisotropic version of randomized smoothing. Evaluation metrics based on the volume of the certified region are proposed, allowing comparisons with the certified regions provided from isotropic randomized smoothing. Experimental results show the usefulness of introducing anisotropic randomized sm...	train	[ "7YCKLcuu1Dl", "c6z56HLtBhh", "SBYIWhscajn", "d8HMqSY6qfR", "71-VSP6UNO1", "T5ibc0uvwh", "Od_8gx8IPx", "EfpBJlYSUZt", "sSHBfsW3lt", "yL33_IiKXZr", "HiOAD2NP_-0", "psfUDrwhhrk", "VrVVPk-S9RD", "n6kdXXDShOs", "F_hutPQKTuw", "-Y5ATj8s4KN", "PDQLTOMKK8n", "IY8-Z0T0hGh" ]	[ "official_reviewer", "official_reviewer", "author", "official_reviewer", "official_reviewer", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ "The authors extend the standard (isotropic) l_1/l_2 randomized smoothing certificate to anisotropic smoothing distributions (and anisotropic regions). They propose to maximize the volume of the certified region for each sample independently, using a memory procedure to preserve soundness. The paper is well written...	[ 6, -1, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, 8 ]	[ 5, -1, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 4 ]	[ "iclr_2022_UFYYol-bRq", "psfUDrwhhrk", "7YCKLcuu1Dl", "iclr_2022_UFYYol-bRq", "Od_8gx8IPx", "iclr_2022_UFYYol-bRq", "EfpBJlYSUZt", "yL33_IiKXZr", "IY8-Z0T0hGh", "HiOAD2NP_-0", "d8HMqSY6qfR", "VrVVPk-S9RD", "n6kdXXDShOs", "F_hutPQKTuw", "7YCKLcuu1Dl", "PDQLTOMKK8n", "iclr_2022_UFYYol-...
iclr_2022_Y0cGpgUhSvp	Prioritized training on points that are learnable, worth learning, and not yet learned	We introduce reducible held-out loss selection (RHOLS), a technique for faster model training which selects a sequence of training points that are “just right”. We propose a tractable information-theoretic acquisition function—the reducible heldout loss—to efficiently choose training points that maximize information ab...	Reject	The paper provides a method to accelerate training by choosing a subset of points. After the initial submission, the reviewers raised a major concern about the practicality of the method. In the rebuttal phase the authors provided additional experiments on a large datasets that addressed this concern. That being said, ...	train	[ "j7DBlbEN3OY", "Ee3brV4vcl9", "hAffkO1zP6", "bymQ9B8Xd1W", "DM5MwnaUSG9", "0eaxDwNXQom", "wE1nF6RkX8V", "x2wDyjXJ-3U", "wELbQ_VOGBD", "--fZ8CkXWq", "RhiDDlVyChk", "TpguYVFiXg", "uajLCBQseuB", "svp52ekuMqE", "b6Z3LbjDxg", "9uX5sUcrHje", "xkBSVX9EXQ8", "Lu8Gl8NPIwx", "0L_1VURW7v3",...	[ "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", ...	[ " Thank you for considering our response and raising your score. Please find the updated manuscript with highlighted changes [here](https://www.dropbox.com/s/l66q8x1srp0z1a8/ICLR2021_CurricuLM_Paper-3.pdf?dl=0). We have changed a lot; the main changes are summarised at the bottom of this response. \n\nYour main rem...	[ -1, -1, -1, 5, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 6 ]	[ -1, -1, -1, 4, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4 ]	[ "DM5MwnaUSG9", "j7DBlbEN3OY", "wE1nF6RkX8V", "iclr_2022_Y0cGpgUhSvp", "svp52ekuMqE", "iclr_2022_Y0cGpgUhSvp", "xkBSVX9EXQ8", "--fZ8CkXWq", "xacYgUg1d75", "wELbQ_VOGBD", "iclr_2022_Y0cGpgUhSvp", "zB4IQPU-dxY", "9uX5sUcrHje", "b6Z3LbjDxg", "bymQ9B8Xd1W", "0eaxDwNXQom", "uajLCBQseuB", ...
iclr_2022_hEiwVblq4P	Proper Straight-Through Estimator: Breaking symmetry promotes convergence to true minimum	In the quantized network, its gradient shows either vanishing or diverging. The network thus cannot be learned by the standard back-propagation, so that an alternative approach called Straight Through Estimator (STE), which replaces the part of the gradient with a simple differentiable function, is used. While STE is k...	Reject	I agree with the reviewers that this work is not well-presented, and it seriously lacks rigor and experimental support. The writing of this work also needs significant improvement. The authors made many claims without offering rigorous proofs, and hand-waved their argument throughout without strong empirical support. I...	train	[ "IU4pWImT_st", "dO_0DmTMMn", "jgKa-ZDt2U", "SAAf0YOu7Vd", "NBTECBA1Iuf", "PRFXj5mnfgI", "KphSYM0R_Qa", "R1W6Xgdrite", "carg3Auhj3F", "WSbJsDudp-B" ]	[ "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ "Training quantized neural network by the biased Straight-Through estimator (STE) is a common practice in the NN quantization literature. However, due to the biased nature of STE, there is no principled guidance in making algorithmic choices for such training methods. This work focuses on the choice of surrogate di...	[ 5, -1, -1, -1, -1, -1, -1, 5, 3, 3 ]	[ 3, -1, -1, -1, -1, -1, -1, 3, 5, 4 ]	[ "iclr_2022_hEiwVblq4P", "jgKa-ZDt2U", "carg3Auhj3F", "IU4pWImT_st", "R1W6Xgdrite", "WSbJsDudp-B", "R1W6Xgdrite", "iclr_2022_hEiwVblq4P", "iclr_2022_hEiwVblq4P", "iclr_2022_hEiwVblq4P" ]
iclr_2022_huXTh4GF2YD	Distance-Based Background Class Regularization for Open-Set Recognition	In open-set recognition (OSR), classifiers should be able to reject unknown-class samples while maintaining robust closed-set classification performance. To solve the OSR problem based on pre-trained Softmax classifiers, previous studies investigated offline analyses, e.g., distance-based sample rejection, which can li...	Reject	This paper tackles the open-set recognition problem, specifically the subset that looks at rejecting test data that with unknown classes that are related to the training data. The proposed approach uses an existing distance-based classifier (based on LDA) combined with a new background class regularizer. Results, compa...	train	[ "uHq8yTxFTRs", "xaNkVWnqvi", "cHgQKaH-wjq", "wXdBl-4mcc", "D72Lp5b3gOn", "0F2gZjVJcLZ", "7OfgO_3n0De", "7oJL7fbK9p", "l-QYasoHAM", "1Wftk-liyI", "5-vdYoMaDOX", "XBWeIKmzq-R", "l2689yvmv9u", "oUpPPn_pdeJ", "a4nOuiWEM_3", "4YfKKXDbFhF", "se7Ap1HJGnj", "ORd1Vgkb7sF" ]	[ "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " On L_bg,k being not sufficient\n\nIn our setting where class-wise anchors are independently sampled from the standard Gaussian distribution, the Euclidean distance between two class-wise anchors is $\\sqrt{2D}$ in expectation ($D$ is the dimension of latent feature vectors).\n\nFor instance, the Euclidean dis...	[ -1, -1, -1, -1, -1, -1, -1, 6, -1, -1, -1, -1, -1, -1, -1, 5, 8, 5 ]	[ -1, -1, -1, -1, -1, -1, -1, 5, -1, -1, -1, -1, -1, -1, -1, 3, 4, 4 ]	[ "cHgQKaH-wjq", "cHgQKaH-wjq", "0F2gZjVJcLZ", "oUpPPn_pdeJ", "l2689yvmv9u", "7OfgO_3n0De", "5-vdYoMaDOX", "iclr_2022_huXTh4GF2YD", "iclr_2022_huXTh4GF2YD", "7oJL7fbK9p", "4YfKKXDbFhF", "7oJL7fbK9p", "se7Ap1HJGnj", "ORd1Vgkb7sF", "se7Ap1HJGnj", "iclr_2022_huXTh4GF2YD", "iclr_2022_huXTh...
iclr_2022_m7zsaLt1Sab	Finding One Missing Puzzle of Contextual Word Embedding: Representing Contexts as Manifold	The current understanding of contextual word embedding interprets the representation by associating each token to a vector that is dynamically modulated by the context. However, this “token-centric” understanding does not explain how a model represents context itself, leading to a lack of characterization from such a p...	Reject	This paper proposes a theory for understanding the context representation in pretrained language models. The strengths of the paper, as identified by reviewers, are in the importance of an attempt to explain contextualization in language models, and in the novelty of using the category theory to model the connection be...	val	[ "ppzjXt6U1xv", "vm19E7Q3sfk", "t858PUbWu5y", "E9TdeW8HaC", "wBy8iiVp-Q", "YPg-WYkj3L-", "1eOQ7OJ7AYe", "fc1Dk6t15aM", "xNPG3EUaQO7", "lQVu-4rTO-1", "qhH6GkY7hHZ", "csSSa8yk2Kl", "u3RhN7iBEZo", "zgag91rhFOt", "qrohqIINOan", "qLXqNgwKSoQ", "WD5oqb2jx5R", "iZuZyJGakqH" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_rev...	[ " I would like to thank the authors for the efforts put into the rebuttal. I agree with the other reviewers that the paper still needs to be improved in clarity in order to be considered for publication. For example, start from basic/widely-used definitions of concepts (_e.g._, defining tokens, contexts with rigoro...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 3, 3, 3, 1 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 2, 3, 4, 2, 3 ]	[ "lQVu-4rTO-1", "qLXqNgwKSoQ", "qhH6GkY7hHZ", "zgag91rhFOt", "iclr_2022_m7zsaLt1Sab", "zgag91rhFOt", "qLXqNgwKSoQ", "xNPG3EUaQO7", "iZuZyJGakqH", "WD5oqb2jx5R", "qrohqIINOan", "iclr_2022_m7zsaLt1Sab", "iclr_2022_m7zsaLt1Sab", "iclr_2022_m7zsaLt1Sab", "iclr_2022_m7zsaLt1Sab", "iclr_2022_...
iclr_2022_YDqIYJBQTQs	Unsupervised Object Learning via Common Fate	Learning generative object models from unlabelled videos is a long standing problem and is required for causal scene modeling. We decompose this problem into three easier subtasks, and provide candidate solutions for each of them. Inspired by the Common Fate Principle of Gestalt Psychology, we first extract (noisy) mas...	Reject	This paper studies the challenging problem of object-centric generation of visual scenes. While the paper has some novel ideas that make it interesting, its (quantitative and qualitative) comparison with existing methods is currently premature to allow drawing conclusions with sufficient evidence. Instead of claiming ...	train	[ "1wRmsfVeBDG", "-PRjQ6dKT-I", "3XXLZtCmL1A", "5nWMm_66_b", "5619J-aaVrj", "9TKDhA68jwF", "gSOFAN1ABTL", "SuR1oy79BT8", "cam8xTfGZJ", "Vl461qMUSuR", "YSub6UwwHR", "njK-2VFVbLN", "AfAXr7JHtVl", "WzAUlNtab_n" ]	[ "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Given the author's rebuttal, I would like to keep my rating.\n\nWhile the idea is novel and the results look good, the comparison with the baseline needs to be more thorough.\nThe hyperparameter needs to be carefully tuned.\nFor example, the glimpse shape prior in SPACE should reflect the fact that fishes tend to...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 6, 3 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 3, 4 ]	[ "3XXLZtCmL1A", "iclr_2022_YDqIYJBQTQs", "SuR1oy79BT8", "gSOFAN1ABTL", "9TKDhA68jwF", "cam8xTfGZJ", "Vl461qMUSuR", "YSub6UwwHR", "WzAUlNtab_n", "AfAXr7JHtVl", "njK-2VFVbLN", "iclr_2022_YDqIYJBQTQs", "iclr_2022_YDqIYJBQTQs", "iclr_2022_YDqIYJBQTQs" ]
iclr_2022_UTTrevGchy	Learning Diverse Options via InfoMax Termination Critic	We consider the problem of autonomously learning reusable temporally extended actions, or options, in reinforcement learning. While options can speed up transfer learning by serving as reusable building blocks, learning reusable options for unknown task distribution remains challenging. Motivated by the recent success ...	Reject	This paper proposes InfoMax Termination Critic (IMTC), a new approach for learning option termination conditions with the aim of discovering more diverse options. IMTC relies on a scalable approximation of the gradient of a mutual information objective with respect to the termination function parameters. Reviewers lik...	val	[ "A2RDqgB-NEz", "4xekqPTiiQf", "FdRUjK8oSFQ", "l-yonqFFuLL", "4OmkJxFFVJw", "3beBl3rT7vH", "VhM_HizVun7", "R6fUYcaQAlQ", "lPwTsXbNd0c", "8GjXr1CSBVP", "3frRj4N_cck", "X4FfHup_yfx", "9-z2u4j5Riv" ]	[ "author", "official_reviewer", "official_reviewer", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ " Thank you for checking our revised manuscript. We really appreciate your efforts. \n\nHere let us discuss your point a bit further.\n\n> I still have my doubts whether the current definition is generally useful or would be limiting depending on the scenario. It would be great if follow up work could expand on th...	[ -1, 6, -1, -1, 6, -1, -1, -1, -1, -1, -1, 5, 5 ]	[ -1, 4, -1, -1, 4, -1, -1, -1, -1, -1, -1, 3, 4 ]	[ "FdRUjK8oSFQ", "iclr_2022_UTTrevGchy", "R6fUYcaQAlQ", "3beBl3rT7vH", "iclr_2022_UTTrevGchy", "lPwTsXbNd0c", "iclr_2022_UTTrevGchy", "4xekqPTiiQf", "4OmkJxFFVJw", "9-z2u4j5Riv", "X4FfHup_yfx", "iclr_2022_UTTrevGchy", "iclr_2022_UTTrevGchy" ]
iclr_2022_W3-hiLnUYl	On the Practicality of Deterministic Epistemic Uncertainty	A set of novel approaches for estimating epistemic uncertainty in deep neural networks with a single forward pass has recently emerged as a valid alternative to Bayesian Neural Networks. On the premise of informative representations, these deterministic uncertainty methods (DUMs) achieve strong performance on detecting...	Reject	The paper performs an empirical evaluation of deterministic methods for the quantification of epistemic uncertainty. There is no new algorithm. The main contribution is the empirical evaluation. This empirical evaluation will be useful for the community. It is an independent evaluation that casts some doubts on the ...	test	[ "piYOUeKnmBp", "JBJqcf5p0Wa", "XGr4R8M6KEE", "vAPlMjhYtP7", "wXIKTIFDiVk", "t3t1PMjITK", "iN1uFJeKEdW", "UdnnREmOI2f", "6nhoBy53Fqf", "2mhTItdDcr5", "uZWKQhmxr_", "b8qWC9KUyUv", "MU26PgWlJIJ", "dk8VofQFtRU" ]	[ "official_reviewer", "author", "official_reviewer", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ "\nThis paper reviews and compares a class of epistemic uncertainty estimation methods that avoid sampling (hence called deterministic) as well as multiple models (like ensembles) because of their memory footprint. The reasons invoked to include some of the cited DUM approaches are not very convincing, since the po...	[ 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 8, 5 ]	[ 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 5, 4 ]	[ "iclr_2022_W3-hiLnUYl", "XGr4R8M6KEE", "uZWKQhmxr_", "iN1uFJeKEdW", "t3t1PMjITK", "6nhoBy53Fqf", "iclr_2022_W3-hiLnUYl", "piYOUeKnmBp", "dk8VofQFtRU", "MU26PgWlJIJ", "b8qWC9KUyUv", "iclr_2022_W3-hiLnUYl", "iclr_2022_W3-hiLnUYl", "iclr_2022_W3-hiLnUYl" ]
iclr_2022_inSTvgLk2YP	MeshInversion: 3D textured mesh reconstruction with generative prior	Recovering a textured 3D mesh from a single image is highly challenging, particularly for in-the-wild objects that lack 3D ground truths. Prior attempts resort to weak supervision based on 2D silhouette annotations of monocular images. Since the supervision lies in the 2D space while the output is in the 3D space, such...	Reject	This submission received 4 diverging ratings: 3, 5, 5, 6. On the positive side, reviewers appreciated the novelty of the approach and strong empirical performance. At the same time, all negatively-inclined reviewers mentioned unfair comparisons with baselines (which was partially addressed in the rebuttal), flaws in th...	train	[ "JCLJSNbv7Gu", "cGmhHhyq8Gs", "aAlnYNSG7a-", "cxxiT6-obFu", "lk6Ezlp7fbE", "17Io2mLrGFS", "ysayD88Qeo", "XRhi4g90MRv", "-A6xHE_NzVd", "P_HLghkIC1", "Dfe2hngPMz-", "HPWC8n7yVpj", "woNf2PqFZfH", "-CdgRN7wO70", "2paTLDn5Z_M", "6EP7BNWKOzr", "yEOX6EIcGjU" ]	[ "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" ]	[ " Thank you for addressing my concerns in detail.\nHowever, I think particularly when comparing to the baselines which were optimized at test time the performance of this is approach is not strong enough wrt 3D reconstruction. Further, I share the concerns of the other reviewers regarding evaluation metrics for 3D ...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, 5, 3, 5 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4, 4, 3 ]	[ "HPWC8n7yVpj", "XRhi4g90MRv", "Dfe2hngPMz-", "-A6xHE_NzVd", "Dfe2hngPMz-", "woNf2PqFZfH", "lk6Ezlp7fbE", "-CdgRN7wO70", "P_HLghkIC1", "yEOX6EIcGjU", "6EP7BNWKOzr", "17Io2mLrGFS", "2paTLDn5Z_M", "iclr_2022_inSTvgLk2YP", "iclr_2022_inSTvgLk2YP", "iclr_2022_inSTvgLk2YP", "iclr_2022_inST...
iclr_2022_y_tIL5vki1l	LatentKeypointGAN: Controlling GANs via Latent Keypoints	Generative adversarial networks (GANs) have attained photo-realistic quality in image generation. However, how to best control the image content remains an open challenge. We introduce LatentKeypointGAN, a two-stage GAN which is trained end-to-end on the classical GAN objective with internal conditioning on a set of sp...	Reject	The paper proposes an unconditional GAN that learns a set of structured keypoints as the intermediate representation. It was shown that these learned keypoints may be used to control the image synthesis output. The paper received a mixed rating before the rebuttal, with one reviewer rating the paper marginally above th...	train	[ "tt4k99sl3Lo", "S8DfXxtFHr", "C1biRTIUfuI", "C14tirVhP-J", "a3XPFXKdib", "G3GU36pdeGQ", "tg81y-MbfU7", "FZ6QnB2alBG", "Zmq3__oJiHE", "IXlpG7rD6KC", "Mdh-ut5rMOc", "IMXnrznwraN", "B6a3fFc7QNv", "SombpJzncm2", "lwjysiYkEpk", "dcioB14wnqH", "vnakjXi18Ri", "DIy9PtDpDig", "WT8UnTPkioU...	[ "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " We applied the PPL following the implementation of StyleGAN2 and the LPIPS implementation of https://github.com/richzhang/PerceptualSimilarity. As we suspected, PPL scores seem not transferable across neural network architectures. The PPL when measuring path length in the entire noise input is 2117.0, which is al...	[ -1, -1, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, 5, 5 ]	[ -1, -1, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4, 4 ]	[ "C1biRTIUfuI", "a3XPFXKdib", "FZ6QnB2alBG", "iclr_2022_y_tIL5vki1l", "Zmq3__oJiHE", "dcioB14wnqH", "B6a3fFc7QNv", "SombpJzncm2", "Mdh-ut5rMOc", "C14tirVhP-J", "IXlpG7rD6KC", "iclr_2022_y_tIL5vki1l", "WT8UnTPkioU", "lwjysiYkEpk", "DIy9PtDpDig", "vnakjXi18Ri", "iclr_2022_y_tIL5vki1l", ...
iclr_2022_TTnjervir3J	DATA-DRIVEN EVALUATION OF TRAINING ACTION SPACE FOR REINFORCEMENT LEARNING	Training action space selection for reinforcement learning (RL) is conflict-prone due to complex state-action relationships. To address this challenge, this paper proposes a Shapely-inspired methodology for training action space categorization and ranking. To reduce exponential-time Shapely computations, the methodolog...	Reject	This paper proposes a method for finding the action space in reinforcement learning problems, characterizing the search space into dispensable and indispensable actions through a Monte Carlo approximation. Reviewers are unanimous that the paper is not fit for publication at this stage. While it tackles an interesting ...	train	[ "2o4I6mrd8d", "xR5IQRQieVS", "orbC3iprVzm", "gCZZw0xGAnA", "B-BOTGFNdD-", "Yq7XMq4hp3J", "AuLRhSHWkD7", "MpSV2E0SBV", "DyEo8Kv-Nnv", "njqBggY3Nnz", "7N2ypB5hgr", "uCDdvDeQOnA" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Thank you for providing an updated version of the paper; the presentation of the paper has indeed improved. Having said that, I completely agree with other reviewers that the quality of the paper in terms of its technical contributions and empirical evaluation still does not match the standard of ICLR. Therefore,...	[ -1, -1, -1, -1, -1, -1, -1, -1, 3, 1, 1, 3 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, 3, 4, 4, 4 ]	[ "MpSV2E0SBV", "B-BOTGFNdD-", "Yq7XMq4hp3J", "AuLRhSHWkD7", "uCDdvDeQOnA", "7N2ypB5hgr", "njqBggY3Nnz", "DyEo8Kv-Nnv", "iclr_2022_TTnjervir3J", "iclr_2022_TTnjervir3J", "iclr_2022_TTnjervir3J", "iclr_2022_TTnjervir3J" ]
iclr_2022_i1ogYhs0ByT	Transformer with a Mixture of Gaussian Keys	Multi-head attention is a driving force behind state-of-the-art transformers which achieve remarkable performance across a variety of natural language processing (NLP) and computer vision tasks. It has been observed that for many applications, those attention heads learn redundant embedding, and most of them can be rem...	Reject	The paper proposes using a mixture of Gaussian models for transformer keys (MGK) so that the posterior distribution of key given query matches the attention scores in the transformer architecture under some assumption. Similarly but in reverse, the query given key under a MoG also matches transformer attention score. T...	test	[ "j35rbryf8Kl", "0NOZ2KoxvhG", "ptTEVgLYfwG", "VrKR1xBRQZt", "HsBLuXrdUrY", "QKSIqhToM9K", "NdfyrJKuSwf", "h4ECMEraUL", "Nv7PgEOAOlH", "qdKTgUgnlAZ", "UMMp8rzW9yg", "LuBwKeuZILl", "dV11Qt-20xh", "97Yg9zWUXj-", "rQiW3qcO6QZ", "O2u7NTNBnrY", "WxCT12cXMNN", "tsZ4LzAj8NG", "OC58DQGcfp...	[ "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", ...	[ " Thanks for your response and we appreciate your endorsement. ", "This paper proposes a transformer with a mixture of Gaussian keys (Transformer-MGK) in order to avoid training redundant attention heads in the transformer modules. The authors formulate the attention mechanism as a posterior distribution $p(t_j =...	[ -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 8, -1, 8, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 6 ]	[ -1, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 4 ]	[ "ptTEVgLYfwG", "iclr_2022_i1ogYhs0ByT", "Cf8LOXaeAB", "97Yg9zWUXj-", "rQiW3qcO6QZ", "TFcncGYXls_", "0NOZ2KoxvhG", "TFcncGYXls_", "0NOZ2KoxvhG", "CM4fpVWAds5", "2vYBXncgzYt", "5RYuRnGSSPp", "O2u7NTNBnrY", "WxCT12cXMNN", "OC58DQGcfp", "NjaeL358YYC", "9BjjAQhyBha", "iclr_2022_i1ogYhs0...
iclr_2022_g6UqpVislvH	Generalized Fourier Features for Coordinate-Based Learning of Functions on Manifolds	Recently, positional encoding of input coordinates has been found crucial to enable learning of high-frequency functions with multilayer perceptrons taking low-dimensional coordinate values. In this setting, sinusoids are typically used as a basis for the encoding, which is commonly referred to as "Fourier Features". ...	Reject	Positional encoding of the input coordinates using Fourier basis [as described in (1)] is a common tool in the context of multilayer perceptrons (MLP). The author propose to replace the Fourier basis with one on manifolds M (2), such as the classical spherical harmonics (M=S^2), the Fourier basis on M=SO(3) or on M=S^2...	train	[ "ZGlNYCPjWvu", "GA45TssCNJe", "e5A2WTjiCuu", "ILP9WGlQJ1W", "bwHfVsMf5ZW", "sedDynPKFW", "r5V6UoVGhO2", "F5Mghtfo3rQ", "Q8A-1VhahsP", "IeY38tVhLcP", "fJqD15IwMAF", "8uo-tbx-Kfg", "JApdcaWPgM7", "VjO5lUuK59S" ]	[ "official_reviewer", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Dear Authors, \n\nThank you for your careful responses. However, I am not convinced. I stand by my original assessment to this paper ( i.e not good enough, reject), which is also consistent with the majority of the reviewers' comments. Most of the reviewers' comments are long, sensible and insightful, and th...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, 10, 3, 3, 5, 6 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4, 4, 4, 4 ]	[ "bwHfVsMf5ZW", "e5A2WTjiCuu", "ILP9WGlQJ1W", "VjO5lUuK59S", "fJqD15IwMAF", "JApdcaWPgM7", "8uo-tbx-Kfg", "IeY38tVhLcP", "iclr_2022_g6UqpVislvH", "iclr_2022_g6UqpVislvH", "iclr_2022_g6UqpVislvH", "iclr_2022_g6UqpVislvH", "iclr_2022_g6UqpVislvH", "iclr_2022_g6UqpVislvH" ]
iclr_2022_8rCMq0yJMG	Source-Target Unified Knowledge Distillation for Memory-Efficient Federated Domain Adaptation on Edge Devices	To support local inference on an edge device, it is necessary to deploy a compact machine learning model on such a device. When such a compact model is applied to a new environment, its inference accuracy can be degraded if the target data from the new environment have a different distribution from the source data that...	Reject	This paper considers a domain adaptation setting where a source domain model trained on a server is adapted on a client using target domain dataset. The paper considers the setting where the client only has a modest memory footprint (e.g., an edge device) and uses a recently proposed technique "TinyTL" (NeurIPS 2020) w...	train	[ "_2LsQcSFSE", "M8tP3t7Jm6j", "mOOtnpGS2LR", "VAPrYe3nxxw", "_5v0MNADrdn", "caInZq3_s8X", "LAByZx17vfE", "d-b1T388acK", "k7-XQLtl5_H", "EQQlBhU5D9h", "vr5pMmhj2K", "cbCRL4jwqat", "KYis8Aij9ow", "5yw7RJFZQo2", "CNvjOYol05Y", "eZRtLzRuDoO", "hwtsDbA9pDC", "sv8mIZ5aYTP", "HTqdbYW5hxd...	[ "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "public", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " We thank all the reviewers for their constructive suggestions on our paper.\n\nWe have revised our paper based on the reviewers' comments, and the revised paper has been posted. \nMoreover, additional experiments have been conducted to further evaluate our scheme, and the new experimental results have been added ...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 8, 5, 6 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4, 3, 3 ]	[ "iclr_2022_8rCMq0yJMG", "LAByZx17vfE", "k7-XQLtl5_H", "EQQlBhU5D9h", "vr5pMmhj2K", "18GnKg6pODh", "HTqdbYW5hxd", "sv8mIZ5aYTP", "5yw7RJFZQo2", "CNvjOYol05Y", "cbCRL4jwqat", "hwtsDbA9pDC", "5yw7RJFZQo2", "hwtsDbA9pDC", "hwtsDbA9pDC", "iclr_2022_8rCMq0yJMG", "iclr_2022_8rCMq0yJMG", "...
iclr_2022_aBAgwom5pTn	Dynamic and Efficient Gray-Box Hyperparameter Optimization for Deep Learning	Gray-box hyperparameter optimization techniques have recently emerged as a promising direction for tuning Deep Learning methods. However, the multi-budget search mechanisms of existing prior works can suffer from the poor correlation among the performances of hyperparameter configurations at different budgets. As a rem...	Reject	This paper presents a new method for performing Bayesian optimization for hyperparameter tuning that uses learning curve trajectories to reason about how long to train a model for (thus "grey box" optimization) and whether to continue training a model. The reviewers seem to find the paper clear, well-motivated and the...	train	[ "u4U2XwR9VbF", "n9PJ7Cu9hKG", "57SO3gv3Uxb", "4IA7xiuVMY8", "-LaNSj9iGY", "MlhgfpdLL9b", "Z-JtqiDnJ9", "WTioUfkUis", "gekGDcc_flp", "4Xa367pjtTZ", "jg9t2aT2MTv", "oUMiJV8NyjC", "QmSZq8GqZgC", "RKQjtbwpcSk", "lXv7bQwUdRi", "gcdhEMCJU1k", "eK81Qb0mtrg", "N6xUPpe8mMk", "rdBdAVq7TS",...	[ "author", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", "author", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "officia...	[ " Thanks a lot for increasing your score after considering the additional experiments and analyses of our rebuttal. Following your recommendations, we will incorporate the contents of the analyses on Appendix D, Figs 14,15,16, into the main manuscript for the camera ready version.", "The paper presents an extensi...	[ -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 6, 3, 3 ]	[ -1, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 3, 4, 5 ]	[ "n9PJ7Cu9hKG", "iclr_2022_aBAgwom5pTn", "iclr_2022_aBAgwom5pTn", "H82D2oIhVMN", "Ry7WjHog53S", "iclr_2022_aBAgwom5pTn", "WTioUfkUis", "gekGDcc_flp", "4Xa367pjtTZ", "jg9t2aT2MTv", "n9PJ7Cu9hKG", "gcdhEMCJU1k", "quoyrfXUwxd", "oUMiJV8NyjC", "iclr_2022_aBAgwom5pTn", "6_Ue3O37NKD", "GrWN...
iclr_2022_wronZ3Mx_d	Transfer Learning for Bayesian HPO with End-to-End Meta-Features	Hyperparameter optimization (HPO) is a crucial component of deploying machine learning models, however, it remains an open problem due to the resource-constrained number of possible hyperparameter evaluations. As a result, prior work focus on exploring the direction of transfer learning for tackling the sample ineffici...	Reject	The submission describes a method for tuning machine learning pipeline hyperparameters using transfer learning from related tuning tasks. In particular, the method uses learned meta features to construct a covariance function for a GP. This was an extremely difficult case and could have gone either way. It was the clo...	train	[ "2hiJAJbK80", "bQ3cu8_CB-1", "YQu-CcKJEyy", "I1inmjLFAgG", "dmpwcedmskr", "quh0AATDW7c", "dL0VM7CmlnI", "sddS4zFwVhu", "ACh8-8_v4g7", "s5YXDh6JTIC", "iePuAayYlht", "ugBZjesfHbQ", "sxobcUFGaR", "tpL7d7GHNFZ", "Tp8HW85E8o9", "0o_i4eyh5vA", "pCCikUxC0E", "kX5r4fcxIW", "GX0IeSgi_7P",...	[ "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_rev...	[ " Thank you again for engaging the in discussion:\n\n- In addition to the mean normalized regret, we present the critical difference diagram which explicitly shows that DKLM (NT) is statistically significant compared to FSBO (NT)\n- Thank you for point out the evaluation of ABLR w.r.t. to wall-clock time. We have r...	[ -1, -1, -1, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 6, 8, 5 ]	[ -1, -1, -1, -1, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 3, 3, 3 ]	[ "bQ3cu8_CB-1", "Tp8HW85E8o9", "sddS4zFwVhu", "kX5r4fcxIW", "iclr_2022_wronZ3Mx_d", "sxobcUFGaR", "sddS4zFwVhu", "s5YXDh6JTIC", "pCCikUxC0E", "ugBZjesfHbQ", "iclr_2022_wronZ3Mx_d", "eHYFVqRznP_", "dmpwcedmskr", "BgxV24OoTU8", "GX0IeSgi_7P", "jbKkZNd12bJ", "BgxV24OoTU8", "dmpwcedmskr...
iclr_2022_NHHM1jjrH1	An Optimization Perspective on Realizing Backdoor Injection Attacks on Deep Neural Networks in Hardware	State-of-the-art deep neural networks (DNNs) have been proven to be vulnerable to adversarial manipulation and backdoor attacks. Backdoored models deviate from expected behavior on inputs with predefined triggers while retaining performance on clean data. Recent works focus on software simulation of backdoor injection ...	Reject	The work studied the problem of inserting backdoor into a deployed model through bit flip. Some important concerns have been proposed by reviewers, including: the incorrect claim of the treat model, the potential defenses are only discussed but not validated, experimental setups and analysis (e.g., the sensitivity tes...	train	[ "OIhcJEf4hlH", "ZhItYX5WC_F", "59AAlz_2ARA", "oQ-DSPnqtvn", "dCKkoWANbu_", "f5uJXI6uSWQ", "aTE-hYtsFR", "89X8T6OiGZF", "WDX2xpv8zr7", "XuZTU5OD_qt", "cc9EO3mZV2B", "k4KnCl2WIYf", "39Gnd1N8Aln", "vnXzlAjMoqW", "XdJ54trExP", "d_tEwnyC55", "81R2QUkTkut", "xrRMNt-UVD", "zgVsCSNXFsE" ...	[ "official_reviewer", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " I very much appreciate the detailed responses from the authors. In this case I confirm my score. ", " We are grateful to all reviewers for their constructive feedback. We have uploaded the revised version of the paper according to their reviews. The modifications are explained in our responses under each review...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 5, 5, 5 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 5, 4, 3 ]	[ "f5uJXI6uSWQ", "iclr_2022_NHHM1jjrH1", "oQ-DSPnqtvn", "xrRMNt-UVD", "d_tEwnyC55", "zgVsCSNXFsE", "d_tEwnyC55", "d_tEwnyC55", "81R2QUkTkut", "xrRMNt-UVD", "xrRMNt-UVD", "81R2QUkTkut", "81R2QUkTkut", "zgVsCSNXFsE", "zgVsCSNXFsE", "iclr_2022_NHHM1jjrH1", "iclr_2022_NHHM1jjrH1", "iclr_...
iclr_2022_kroqZZb-6s	Cluster-based Feature Importance Learning for Electronic Health Record Time-series	The recent availability of Electronic Health Records (EHR) has allowed for the development of algorithms predicting inpatient risk of deterioration and trajectory evolution. However, prediction of disease progression with EHR is challenging since these data are sparse, heterogeneous, multi-dimensional, and multi-modal ...	Reject	This paper has been independently assessed by four expert reviewers. Two of them recommend acceptance (one straight, one marginal), and two rejection (both marginal). Among the main limitations of the presented work, found by the reviewers, was the limited reproducibility of the results due to the use of private data. ...	train	[ "YiHSxTdYom4", "b1FMdt74G7", "9SHeCcG0z7o", "Lpw87vZTXEv", "r3I8U-aU9Va", "NT9c6B1YTfk", "fRJOTtgPkb", "qSUvNTY1Tp7", "OFDKinL1cS", "yRbK2ub8ep_", "dgGhtPyLBW9", "F797-iwcfhu" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer" ]	[ " I want to thank the authors for clarifying my questions. I updated my score based on the discussion and revision.", " Thank you authors for clarifying my questions and addressing my considerations. \nI have updated my novelty score based on the discussion and the revised manuscript.", "The paper describes a c...	[ -1, -1, 6, 8, -1, 5, -1, -1, -1, -1, -1, 5 ]	[ -1, -1, 3, 4, -1, 5, -1, -1, -1, -1, -1, 3 ]	[ "OFDKinL1cS", "yRbK2ub8ep_", "iclr_2022_kroqZZb-6s", "iclr_2022_kroqZZb-6s", "dgGhtPyLBW9", "iclr_2022_kroqZZb-6s", "iclr_2022_kroqZZb-6s", "F797-iwcfhu", "NT9c6B1YTfk", "9SHeCcG0z7o", "Lpw87vZTXEv", "iclr_2022_kroqZZb-6s" ]
iclr_2022_NeRrtif_hfa	Better state exploration using action sequence equivalence	Incorporating prior knowledge in reinforcement learning algorithms is mainly an open question. Even when insights about the environment dynamics are available, reinforcement learning is traditionally used in a \emph{tabula rasa} setting and must explore and learn everything from scratch. In this paper, we consider the ...	Reject	This well-written paper introduces an improved exploration strategy by exploiting knowledge about sequences of actions that lead to the same state. The idea is straightforward and easy to understand and apply, which makes it potentially interesting. An important downside is the limited applicability of the method, as t...	train	[ "HCiKSbUe9j5", "2XHL3KOo_7K", "4z7PJarcez", "Xx8ug0OYCPI", "FqrQp_p29nU", "qIKB_oAHZtC", "_FbWQRTBFM_", "6i_IFHX8KJ4", "oepR33aM6ug" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ "This paper considers a property of MDPs where multiple sequences of states end up at the same action. The main contributions of this paper are (1) a formalization of the action sequence phenomenon, (2) an algorithm to build a local dynamics graph, (3) a simple exploration procedure based on the local dynamics grap...	[ 5, -1, -1, -1, -1, -1, -1, 5, 8 ]	[ 4, -1, -1, -1, -1, -1, -1, 4, 4 ]	[ "iclr_2022_NeRrtif_hfa", "_FbWQRTBFM_", "qIKB_oAHZtC", "iclr_2022_NeRrtif_hfa", "oepR33aM6ug", "6i_IFHX8KJ4", "HCiKSbUe9j5", "iclr_2022_NeRrtif_hfa", "iclr_2022_NeRrtif_hfa" ]
iclr_2022_qw674L9PfQE	CLOOB: Modern Hopfield Networks with InfoLOOB Outperform CLIP	Contrastive learning with the InfoNCE objective is exceptionally successful in various self-supervised learning tasks. Recently, the CLIP model yielded impressive results on zero-shot transfer learning when using InfoNCE for learning visual representations from natural language supervision. However, InfoNCE as a lower ...	Reject	Four experts reviewed the paper and provided mixed recommendations. All reviewers found the experimental results strong, but they have different views about the technical novelty. Three reviewers considered the technical novelty as a weakness of the paper, but Reviewer z4BR was less concerned about it than the other tw...	train	[ "z9dOuh7u6i", "FyhEo0rQDGv", "IUrS2cAZWhw", "LrQNAMGg9KF", "Ac7kmeU4vzq", "Pt3OjnlxHU", "_TW3wCHrJW6", "kKotGPIZuda", "z4UzBUnI7e", "9ZmpcSc0K-c", "t5C0ysDSnz8", "f6y3rTtEgLZ", "udoxSN8j-ic", "NKmXd3A4w9U", "SYXWvkOc4i", "kERvpacwqcW", "4IucyYqtew", "-9V7LlEPEdu", "SLj_0CXosPZ" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "...	[ " I would like to thank the authors for responding to my comments. I feel they have addressed my first question concerning the relationship between CLIP and InfoLOOB. On the other hand, despite their comments, I believe my concerns re. the novelty still stand, to some extent.\nAs I do not believe a paper that achie...	[ -1, -1, 8, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 6, 5 ]	[ -1, -1, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4, 4 ]	[ "kERvpacwqcW", "LrQNAMGg9KF", "iclr_2022_qw674L9PfQE", "IUrS2cAZWhw", "f6y3rTtEgLZ", "SYXWvkOc4i", "NKmXd3A4w9U", "_TW3wCHrJW6", "iclr_2022_qw674L9PfQE", "-9V7LlEPEdu", "-9V7LlEPEdu", "-9V7LlEPEdu", "SLj_0CXosPZ", "SLj_0CXosPZ", "4IucyYqtew", "IUrS2cAZWhw", "iclr_2022_qw674L9PfQE", ...
iclr_2022_djhu4DIZZHR	NAIL: A Challenging Benchmark for Na\"ive Logical Reasoning	Logical reasoning over natural text is an important capability towards human level intelligence. Existing datasets are either limited and inadequate to train and evaluate logical reasoning capability (e.g., LogiQA and ReClor), or not oriented for logical reasoning (e.g., SQuAD and HotpotQA). In this paper, we focus on ...	Reject	This paper introduces a dataset, based on preexisting standardized tests, of elimination/grid-completion-style logical reasoning puzzles expressed in text; available in both Chinese and English (with some of the text coming from semi-automatic translation). The early pretrained MLMs BERT and RoBERTa perform poorly. T...	train	[ "LVtdrKhz8QW", "TXWaJqjr7oC", "UAUFhLj8G1k", "hqUx04V2YJ7", "vBQJK39JHS", "jfAJZWwXFzj", "uulmp5R3bD1", "nPY8TDiqrz", "gbAemX2Rp8K", "2Lxx4k16GRT", "WmZqRDm9QL5", "iWZQ_qEUl-q", "82WtF9DR_nT", "rEdzcjZef4H", "L1sXtJ4ZF3F", "XHOcb4Ovis", "H2uij283uiA" ]	[ "official_reviewer", "official_reviewer", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ "This paper proposes a new challenge for large pretrained models (which can of course extend to any other model claiming to store language knowledge): naive logical reasoning. The main contribution of the paper is the construction of a dataset in Chinese and English (termed NAIL) with items for training and evaluat...	[ 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 8, 5, 3 ]	[ 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 3, 5 ]	[ "iclr_2022_djhu4DIZZHR", "gbAemX2Rp8K", "hqUx04V2YJ7", "uulmp5R3bD1", "iclr_2022_djhu4DIZZHR", "iclr_2022_djhu4DIZZHR", "nPY8TDiqrz", "XHOcb4Ovis", "2Lxx4k16GRT", "LVtdrKhz8QW", "iWZQ_qEUl-q", "L1sXtJ4ZF3F", "rEdzcjZef4H", "H2uij283uiA", "iclr_2022_djhu4DIZZHR", "iclr_2022_djhu4DIZZHR"...
iclr_2022_ciSap6Cw5mk	MANDERA: Malicious Node Detection in Federated Learning via Ranking	Federated learning is a distributed learning paradigm which seeks to preserve the privacy of each participating node's data. However, federated learning is vulnerable to attacks, specifically to our interest, model integrity attacks. In this paper, we propose a novel method for malicious node detection called MANDERA. ...	Reject	This manuscript proposes a ranking approach to identify Byzantine agents in federated learning. Distinct from existing methods, the mitigation is implemented by computing ranks for each gradient, then computing rank statistics across agents. The primary intuition is that adversarial agents can be identified by examinin...	train	[ "SuvrZLPjx1", "uPoicEM1taP", "h-GN1vlNU8B", "dMAJubftWS", "WGG3xl8bUlz", "IcR8C7hASdD", "9ZKHHamI5KC", "N18B0_Yhxxx", "CnY4p0opL6", "_cMenkKfMwO", "HAr_ZDZ3y2Q", "hvS5AVLp1IL", "IDi1X-wJSiI", "2dlrZP-0cmt" ]	[ "author", "official_reviewer", "official_reviewer", "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "official_reviewer", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ " Thank you for your positive comments. We are appreciated if you could update your score to reflect your positive view on our paper. Below are the further clarifications. We would be happy to add these clarifications in the camera ready version.\n\n1). What configuration of FLTrust was used? The size of the valida...	[ -1, -1, -1, -1, -1, 3, -1, -1, -1, -1, -1, -1, 6, 8 ]	[ -1, -1, -1, -1, -1, 3, -1, -1, -1, -1, -1, -1, 3, 3 ]	[ "uPoicEM1taP", "N18B0_Yhxxx", "dMAJubftWS", "WGG3xl8bUlz", "_cMenkKfMwO", "iclr_2022_ciSap6Cw5mk", "HAr_ZDZ3y2Q", "CnY4p0opL6", "hvS5AVLp1IL", "IcR8C7hASdD", "2dlrZP-0cmt", "IDi1X-wJSiI", "iclr_2022_ciSap6Cw5mk", "iclr_2022_ciSap6Cw5mk" ]
iclr_2022_6j9YOwh8itH	Unified Recurrence Modeling for Video Action Anticipation	Forecasting future events based on evidence of current conditions is an innate skill of human beings, and key for predicting the outcome of any decision making. In artificial vision for example, we would like to predict the next human action before it is actually performed, without observing the future video frames ass...	Reject	This paper presents work on action anticipation. The reviewers appreciated the message passing based method. However, concerns were raised regarding novelty, effectiveness, presentation, empirical results, and magnitude of impact for ICLR. The reviewers considered the authors' response in their subsequent discussion...	val	[ "NYPb20Wlq7a", "WmeUqBXQDp-", "QW2zlxD60gN", "gO8yv5e5nS7", "viGsjKiO4Dm", "VU0FGRcvExJ", "Uo_0U_6priw", "OBB4D071bZG", "cDyBSBdlWjq", "FABoBHe7u7k", "3EIYxOBuC_m" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ "This paper addressed an interesting problem of action anticipation. To this end, the authors proposed a unified recurrence model that generates the graph representation of the video sequence. Extensive experimental results have shown the effectiveness of the proposed method. Strength:\n\n+The message passing metho...	[ 3, -1, -1, -1, -1, -1, -1, -1, 5, 6, 3 ]	[ 4, -1, -1, -1, -1, -1, -1, -1, 5, 3, 3 ]	[ "iclr_2022_6j9YOwh8itH", "viGsjKiO4Dm", "Uo_0U_6priw", "3EIYxOBuC_m", "NYPb20Wlq7a", "FABoBHe7u7k", "cDyBSBdlWjq", "iclr_2022_6j9YOwh8itH", "iclr_2022_6j9YOwh8itH", "iclr_2022_6j9YOwh8itH", "iclr_2022_6j9YOwh8itH" ]
iclr_2022_Ehhk6jyas6v	On The Quality Assurance Of Concept-Based Representations	Recent work on Explainable AI has focused on concept-based explanations, where deep learning models are explained in terms of high-level units of information, referred to as concepts. In parallel, the field of disentanglement learning has explored the related notion of finding underlying factors of variation in the dat...	Reject	This paper considers the question of whether recent concept-based learning algorithms, as well disentangled representation learning algorithms, result in high-quality representations. In particular, the authors consider what high-quality should mean in terms of the relationship with ground truth concepts and the abilit...	train	[ "J8RDzsktD5J", "Wlfsql39mfg", "5RjLl7ahkyZ", "4YRC-uJSng5", "qF1QJn4ulE", "a0ud9FzaHA", "0iqsQ5_B5bB", "SpYrYWWtAQ", "txtXwLK1mem", "JB6NoZjps1g", "cv7YDc7qFiZ", "yvROdG4Zo8O", "DHMbCif1GMv", "qWBUkTMlW4m", "8_NK8OUhp21", "fPJ2mn9yJh" ]	[ "official_reviewer", "author", "author", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Thanks to the authors for their response. A couple thoughts on what they've discussed:\n\nAbout the choice of AUC. If the main motivation for using AUC is to bound the metric in [0, 1], that's not a very good reason. The mutual information $I(X; Y)$ is bounded by $0 \\leq I(X; Y) \\leq H(Y)$, and if $Y$ is di...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 5, 5 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4, 4 ]	[ "qF1QJn4ulE", "5RjLl7ahkyZ", "a0ud9FzaHA", "DHMbCif1GMv", "8_NK8OUhp21", "SpYrYWWtAQ", "iclr_2022_Ehhk6jyas6v", "fPJ2mn9yJh", "JB6NoZjps1g", "qWBUkTMlW4m", "yvROdG4Zo8O", "SpYrYWWtAQ", "qF1QJn4ulE", "iclr_2022_Ehhk6jyas6v", "iclr_2022_Ehhk6jyas6v", "iclr_2022_Ehhk6jyas6v" ]
iclr_2022_lycl1GD7fVP	Neural tangent kernel eigenvalues accurately predict generalization	Finding a quantitative theory of neural network generalization has long been a central goal of deep learning research. We extend recent results to demonstrate that, by examining the eigensystem of a neural network's "neural tangent kernel," one can predict its generalization performance when learning arbitrary function...	Reject	Summary: Study generalization in kernel regression discussing the NTK case and experiments on finite width nets. Strengths: - Mix of theoretical and empirical results in an important topic. - Advances a promising recent line of work. Weaknesses: - Concerns about novelty and lack of comparison with existing...	train	[ "6xyhJzFqy37", "il8eEjqDwgc", "FrBRmgsj4H6", "31iK7bz2rNW", "fOEuMKJaWDP", "RBtOee6CSiZ", "RrYoxskokQu", "l_QoWQw8OhU", "1AC9KqpCmS6", "MvVarRzzZEL", "58iKIrZzsnl", "0Rf9ntaMDu8", "rbaOXP6Jb6D", "vxodXAItVf-", "Y74X_Gm7KVB" ]	[ "author", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "official_reviewer", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " We thank the reviewer for elaborating upon their critiques. Our high-level response is quite simple:\n\n* We do not claim to be the first to show that \"a function is more learnable the more weight it places in high eigenvalue modes\"! This is a well-established picture (as we discussed in our Introduction!),...	[ -1, -1, -1, 8, -1, -1, -1, -1, -1, -1, -1, -1, 3, 6, 5 ]	[ -1, -1, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, 5, 2, 3 ]	[ "FrBRmgsj4H6", "iclr_2022_lycl1GD7fVP", "fOEuMKJaWDP", "iclr_2022_lycl1GD7fVP", "58iKIrZzsnl", "Y74X_Gm7KVB", "31iK7bz2rNW", "RBtOee6CSiZ", "RrYoxskokQu", "vxodXAItVf-", "0Rf9ntaMDu8", "rbaOXP6Jb6D", "iclr_2022_lycl1GD7fVP", "iclr_2022_lycl1GD7fVP", "iclr_2022_lycl1GD7fVP" ]
iclr_2022_74cDdRwm4NV	Learning to Shape Rewards using a Game of Two Partners	Reward shaping (RS) is a powerful method in reinforcement learning (RL) for overcoming the problem of sparse or uninformative rewards. However, RS typically relies on manually engineered shaping-reward functions whose construction is time consuming and error-prone. It also requires domain knowledge which runs contrar...	Reject	This paper tackled the reward shaping problem under the framework of Markov games. The authors proposed reward shaping algorithms for RL with mild theoretical guarantees. The AC agrees with the reviewers that the empirical performance is ambiguous. The paper should be substantially improved before being accepted.	train	[ "cBe4MCPn9FH", "j_-nDhXCt5l", "SaEZVKwTddr", "hapKw_lfnIp", "OlLu3v7r3N", "PYq3fql-N4f", "0ZThfscltp", "Gh12xqvPdry", "Xmt-n0xNoCX", "NrgD5ykt70P", "A2AplRd1R-q", "FI3Rtgh74P", "-CKZP3smUd9", "RX6Cge3Ly7L", "PiGwIkeGFN5", "O_MIDZOobdS" ]	[ "official_reviewer", "author", "author", "official_reviewer", "author", "official_reviewer", "author", "author", "author", "official_reviewer", "official_reviewer", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Thanks for clarifying the point about Prop 3! I see now how the argument is structured. ", " Dear Reviewers\n\nThank you for your careful reading of our paper and your insightful comments. We would like the authors and AC to consider the novelty of the framework which introduces a new way of performing reward-s...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, -1, -1, -1, 3, 6, 5 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, -1, -1, -1, 3, 4, 3 ]	[ "SaEZVKwTddr", "iclr_2022_74cDdRwm4NV", "hapKw_lfnIp", "0ZThfscltp", "PYq3fql-N4f", "Gh12xqvPdry", "PiGwIkeGFN5", "O_MIDZOobdS", "A2AplRd1R-q", "iclr_2022_74cDdRwm4NV", "FI3Rtgh74P", "NrgD5ykt70P", "RX6Cge3Ly7L", "iclr_2022_74cDdRwm4NV", "iclr_2022_74cDdRwm4NV", "iclr_2022_74cDdRwm4NV"...
iclr_2022_Fj1Tpym9KxH	A Closer Look at Smoothness in Domain Adversarial Training	Domain adversarial training has been ubiquitous for achieving invariant representations and is used widely for various domain adaptation tasks. In recent times methods converging to smooth optima have shown improved generalization for supervised learning tasks like classification. In this work, we analyze the effect o...	Reject	This paper studies the loss landscape of domain adversarial neural networks for domain adaptation. First, the authors show that smooth minima with respect to adversarial loss leads to sub-optimal generalization on the target domain. Then, they suggest to enforce smoothness only with respect to the task loss. 3 reviewer...	train	[ "K833SA6H2c", "AsLmszM_d8k", "bAXv5WYu5r", "A_9kWZCPcID", "p9UpIekGoXv", "sPMMmfVzVT9", "nNpINL8Omq", "WIuv7GBZbbb", "j_Tr-GHan7R", "MXDw2NM45jM", "qpFGQ01xFcq", "gJ8PmvqNyND", "cMtZWScmDTI", "bv6Ylyedv5p", "FVwisolwsz", "VOrOOdMyoF2" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ " Dear authors, thank you for your effort in the rebuttal and for the extra experiments. In light of this new empirical evidence to support some of the claims in the manuscript, I increased the scores of \"Correctness\" and \"Empirical Novelty And Significance\" from 2 to 3.", "This work studies the loss landscap...	[ -1, 5, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5, 8 ]	[ -1, 4, 2, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 4 ]	[ "p9UpIekGoXv", "iclr_2022_Fj1Tpym9KxH", "iclr_2022_Fj1Tpym9KxH", "nNpINL8Omq", "A_9kWZCPcID", "FVwisolwsz", "qpFGQ01xFcq", "iclr_2022_Fj1Tpym9KxH", "gJ8PmvqNyND", "VOrOOdMyoF2", "j_Tr-GHan7R", "AsLmszM_d8k", "bAXv5WYu5r", "cMtZWScmDTI", "iclr_2022_Fj1Tpym9KxH", "iclr_2022_Fj1Tpym9KxH" ...
iclr_2022_oxwsctgY5da	A Branch and Bound Framework for Stronger Adversarial Attacks of ReLU Networks	Strong adversarial attacks are important for evaluating the true robustness of deep neural networks. Most existing attacks find adversarial examples via searching in the input space, e.g., using gradient descent. In this work, we formulate an adversarial attack using a branch-and-bound (BaB) procedure on ReLU neural ne...	Reject	This work proposes a branch and bound framework for adversarial attacks, based on the MIP formulation of attacking against ReLU networks. It adopts several heuristic tricks to accelerate the attack efficiency, and shows better attack performance on hard examples, compared to off-the-shelf MIP solvers. The main concer...	train	[ "QZL_1vpyPYa", "SOabP4SExvj", "dxZkqd9xiDZ", "AltWylkyzle", "IDBBEPi4_4J", "29wh3RV-Kr2", "U0WZONrv0ZL", "o-66NBo-6oy", "NLEdW4tWoLa", "52SRlF0by_", "mKhQoKYswNw", "myk54evW31E", "X5XcsLIS4MT", "TJ-ooPs-8Bi", "gTDzePzX8qo", "5KNCTBP--RW", "JH6sTxxmnaX", "PxDVg28--fl", "UqowN5vlVV...	[ "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "official_reviewer", "official_reviewer", "author", "official_reviewer", "author", "author", "author", "author", "author", "...	[ "In this paper, the authors propose to solve the adversarial attack problem of ReLU neural network by using a branch-and-bound procedure, and search adversarial examples within the activation space corresponding to binary variables of a mixed-integer programming formulation. The experiments showcase that such adver...	[ 8, -1, 8, -1, 6, -1, -1, -1, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1 ]	[ 2, -1, 3, -1, 3, -1, -1, -1, -1, 2, -1, -1, -1, -1, -1, -1, -1, -1, -1 ]	[ "iclr_2022_oxwsctgY5da", "myk54evW31E", "iclr_2022_oxwsctgY5da", "dxZkqd9xiDZ", "iclr_2022_oxwsctgY5da", "IDBBEPi4_4J", "mKhQoKYswNw", "dxZkqd9xiDZ", "IDBBEPi4_4J", "iclr_2022_oxwsctgY5da", "PxDVg28--fl", "X5XcsLIS4MT", "QZL_1vpyPYa", "iclr_2022_oxwsctgY5da", "dxZkqd9xiDZ", "QZL_1vpyPY...
iclr_2022_mNLLDtkAy4X	Escaping Stochastic Traps with Aleatoric Mapping Agents	When extrinsic rewards are sparse, artificial agents struggle to explore an environment. Curiosity, implemented as an intrinsic reward for prediction errors, can improve exploration but fails when faced with action-dependent noise sources. We present aleatoric mapping agents (AMAs), a neuroscience inspired solution mod...	Reject	Authors present a method to disentangle epistemic from aleatoric uncertainty for avoiding the noisy TV problem during self-driven exploration. This is an important area where we need more ideas and experiments. The authors present a biologically inspired approach and through experiments. Although it doesn't present the...	train	[ "b4KwQAmfnB", "ZX7JwNVadc", "-YzSH22pHL", "bmU1ElffiB", "sKzNey_xnwd", "_BBrP5d4sEe", "c0Ic3lKhZm6", "w-aOT8huq-a", "PI-hcdBNkb-", "xSMp7YQZoKC", "43KiiNh59jI", "7s3k4XFIopa", "Xde5kwQ1HSA", "wHh-opY0yu0", "cDTCwQkDgkm", "a9sM5ziTWCg", "wjeBuj5Tmt1" ]	[ "official_reviewer", "author", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ " Thank you for your response.\n\nIdeally, I would like to see:\n1) The experiments, similar to Burda et al '19 and similar to your Space invaders without noisy TVs and Mario without noisy TVs where your approach (AMA) outperforms the existing approaches (e.g. Pixel MSE, IDF, and RND) as measured by the e...	[ -1, -1, -1, -1, -1, -1, -1, -1, 6, -1, -1, -1, -1, -1, -1, 3, 5 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, 5, -1, -1, -1, -1, -1, -1, 4, 4 ]	[ "ZX7JwNVadc", "-YzSH22pHL", "sKzNey_xnwd", "w-aOT8huq-a", "43KiiNh59jI", "xSMp7YQZoKC", "iclr_2022_mNLLDtkAy4X", "cDTCwQkDgkm", "iclr_2022_mNLLDtkAy4X", "Xde5kwQ1HSA", "7s3k4XFIopa", "wjeBuj5Tmt1", "PI-hcdBNkb-", "a9sM5ziTWCg", "a9sM5ziTWCg", "iclr_2022_mNLLDtkAy4X", "iclr_2022_mNLLD...
iclr_2022_kxARp2zoqAk	Information-Aware Time Series Meta-Contrastive Learning	Various contrastive learning approaches have been proposed in recent years and achieve significant empirical success. While effective and prevalent, contrastive learning has been less explored for time series data. A key component of contrastive learning is to select appropriate augmentations imposing some priors to co...	Reject	This paper presents a method which selects feasible data augmentations suitable for contrastive time series representation learning. The topic in this paper is timely and interesting. One of 4 reviewers did not complete the review, not responding to a few reminders. So, one emergency reviewer, who is an expert in meta-...	test	[ "IloMGGLqnn", "8lYsTvo8_Xe", "vLwMwITmOF8", "bk4LtiuQyWr", "e0AKIJ16w1W", "iU1W2OwuQSn", "HbZsv1F0_tt", "6wR_Y2xGlB_", "Jm6b0UxGH8", "M3g-6ZE0Zwn", "XgV0r77qqmR", "Ecuc-ZknNPU", "R3kX_NOrV7W", "0XkgJESdAyS", "etr8_qRhWoj", "P3rBcTZLP5D", "dPg_YHszGl", "t37fo1GDK1p", "wdModhMBe8M"...	[ "author", "official_reviewer", "author", "author", "author", "official_reviewer", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", ...	[ " Thanks for the reviewer's comments. We will revise our paper to clearly state our contributions, including information-aware adaptive augmentation and new contrastive loss in the next version. To further address the review's concerns on the advantages of the proposed adaptive selection, we have conducted more com...	[ -1, -1, -1, -1, -1, 5, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 10, 6 ]	[ -1, -1, -1, -1, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 3, 5 ]	[ "8lYsTvo8_Xe", "vLwMwITmOF8", "Jm6b0UxGH8", "iU1W2OwuQSn", "iU1W2OwuQSn", "iclr_2022_kxARp2zoqAk", "XgV0r77qqmR", "M3g-6ZE0Zwn", "etr8_qRhWoj", "P3rBcTZLP5D", "dPg_YHszGl", "iclr_2022_kxARp2zoqAk", "_6fca2LP2sO", "_6fca2LP2sO", "_6fca2LP2sO", "wdModhMBe8M", "wdModhMBe8M", "7hlDg0gw...
iclr_2022_mF5tmqUfdsw	Zeroth-Order Actor-Critic	Evolution based zeroth-order optimization methods and policy gradient based first-order methods are two promising alternatives to solve reinforcement learning (RL) problems with complementary advantages. The former work with arbitrary policies, drive state-dependent and temporally-extended exploration, possess robustne...	Reject	The paper proposes a new reinforcement learning actor-critic type algorithm for parameterized policy spaces. The actor builds gradient estimates derived from perturbations of the policy (in the spirit of simultaneous perturbation stochastic approximation (SPSA) or Flaxman-Kalai-McMahan's "Gradient Descent without a Gra...	train	[ "Eh9lpT1TAL", "s9NwPir81b", "yk5p5xy5Fho", "qfnFUgs8PdU", "t76WGkhsqIA", "RciKhFIDEvW", "gnWnpS22jhf", "w68G57glWV", "CT3KE_L04Tp", "pyGbQ-t39uF", "DIB7SHGrx6g", "ISQjuwyeL7C", "tFMTL2t9V4G" ]	[ "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "official_reviewer" ]	[ " We thank the reviewers for their thorough and insightful feedback. We are glad that the reviewers gave generally positive comments on our work:\n\n1. Our idea and the proposed algorithm is novel (Reviewer cTPK), interesting (Reviewer eX28), well-motivated and elegant (Reviewer wk6f);\n\n2. Experiment on several b...	[ -1, -1, 6, 6, -1, -1, -1, -1, -1, -1, -1, -1, 6 ]	[ -1, -1, 4, 3, -1, -1, -1, -1, -1, -1, -1, -1, 3 ]	[ "iclr_2022_mF5tmqUfdsw", "w68G57glWV", "iclr_2022_mF5tmqUfdsw", "iclr_2022_mF5tmqUfdsw", "CT3KE_L04Tp", "gnWnpS22jhf", "DIB7SHGrx6g", "tFMTL2t9V4G", "pyGbQ-t39uF", "yk5p5xy5Fho", "ISQjuwyeL7C", "qfnFUgs8PdU", "iclr_2022_mF5tmqUfdsw" ]
iclr_2022_rRg0ghtqRw2	That Escalated Quickly: Compounding Complexity by Editing Levels at the Frontier of Agent Capabilities	Deep Reinforcement Learning (RL) has recently produced impressive results in a series of settings such as games and robotics. However, a key challenge that limits the utility of RL agents for real-world problems is the agent's ability to generalize to unseen variations (or levels). To train more robust agents, the fiel...	Reject	This paper tackles the problem of Unsupervised Environment Design to train more robust agents. The proposed method trains RL agents by generating a curriculum of training tasks to enable agents to generalize to many tasks. The key contribution is an algorithm to generate this curriculum by incremental edits of the grid...	train	[ "GPmWaGlg5Ae", "ioOr-o5kewC", "ZiimU1XgPeb", "WLtjdN7IjX", "NyXb3yQZf3", "xuinsBMOCxC", "hejdUFP8MF", "xqSLpEPgS9i", "DPF3DHPSGQ1", "Rtbx_TBcWt8", "uVmt8m_cigh", "992MgT4JOh", "EN9XSKn4Ws9", "FZ2oimYEKK", "d70yJweg2vl", "s8eSac5sLqn" ]	[ "official_reviewer", "official_reviewer", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " I thank the authors for the clarifications. My main concern is still w.r.t. the statement made about the optimality of the algorithm in Remark 1.\n\n> We agree with the reviewer that these two approaches will not guarantee a perfect estimate of the true regret, which is often computationally intractable. \n\nI ag...	[ -1, -1, -1, -1, -1, -1, 6, -1, -1, -1, -1, -1, -1, 5, 5, 5 ]	[ -1, -1, -1, -1, -1, -1, 4, -1, -1, -1, -1, -1, -1, 3, 4, 3 ]	[ "Rtbx_TBcWt8", "uVmt8m_cigh", "xuinsBMOCxC", "iclr_2022_rRg0ghtqRw2", "xqSLpEPgS9i", "DPF3DHPSGQ1", "iclr_2022_rRg0ghtqRw2", "EN9XSKn4Ws9", "s8eSac5sLqn", "d70yJweg2vl", "FZ2oimYEKK", "EN9XSKn4Ws9", "hejdUFP8MF", "iclr_2022_rRg0ghtqRw2", "iclr_2022_rRg0ghtqRw2", "iclr_2022_rRg0ghtqRw2"...
iclr_2022_9zcjXdavnX	Sampling from Discrete Energy-Based Models with Quality/Efficiency Trade-offs	Energy-Based Models (EBMs) allow for extremely flexible specifications of probability distributions. However, they do not provide a mechanism for obtaining exact samples from these distributions. Monte Carlo techniques can aid us in obtaining samples if some proposal distribution that we can easily sample from is avail...	Reject	This paper proposes a new approximate sampling approach called Quasi Rejection Sampling (QRS) to exploit global proposal distributions without requiring to know a bound on the associated importance ratio, and providing a trade-off between the approximation quality of the sampler and its efficiency. QRS is demonstrated ...	train	[ "OneLxjUeZy8", "DsY1hT8A28k", "k65wbXaBOw3", "lCEH5aylyI-", "4jzKHGma4An", "-C2O1VPZRvG", "ZeUxVHROACo", "VbVIRXyfba", "ohGu5shAh3F", "ymiDt5M8KMO" ]	[ "official_reviewer", "author", "author", "author", "official_reviewer", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ "This paper focuses on the challenge of sampling from unnormalized probability distributions (or Energy-Based Models) -- typically those found in the context of NLP when dealing with the task of controlled generation (post-hoc conditioning of a pre-trained unconditional model). Previous approaches for this have bee...	[ 6, -1, -1, -1, -1, -1, -1, -1, 5, 1 ]	[ 4, -1, -1, -1, -1, -1, -1, -1, 3, 5 ]	[ "iclr_2022_9zcjXdavnX", "ymiDt5M8KMO", "OneLxjUeZy8", "ohGu5shAh3F", "ZeUxVHROACo", "ZeUxVHROACo", "iclr_2022_9zcjXdavnX", "ZeUxVHROACo", "iclr_2022_9zcjXdavnX", "iclr_2022_9zcjXdavnX" ]
iclr_2022_AkJyAE46GA	Pretrained models are active learners	An important barrier to the safe deployment of machine learning systems is the risk of \emph{task ambiguity}, where multiple behaviors are consistent with the provided examples. We investigate whether pretrained models are better active learners, capable of asking for example labels that \textit{disambiguate} between t...	Reject	The paper shows that active learning is an emergent property of pre-trained models. They show that simple uncertainty sampling improves sample efficiency by 6 times (up to 6x fewer samples for the same accuracy). This is an interesting and important observation that has practical implications. Initially, there were v...	train	[ "zW5QJc8OWoG", "5Sd5-i7sXZ7", "dq1pwmJIhWj", "p70bNpeArXW", "zN__ic7dJPd", "t3k0SUDrnRU", "rWhOYj2aGw", "OSfC7wdI4Qp", "ZZ5avPVTkb4", "rEDildbz3HQ", "hrMZdCfV2bW", "QiYLLQB1dzz" ]	[ "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Thanks much for the rebuttal (and submission). First of all, I did read the updated version of this submission and do think it is a better paper and do appreciate the 'dose-response' experiment. In my estimation, this submission is now somewhere between a marginal and borderline reject, but I still think it is a ...	[ -1, -1, -1, -1, -1, -1, -1, -1, 8, 3, 5, 3 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, 3, 4, 5, 4 ]	[ "zN__ic7dJPd", "iclr_2022_AkJyAE46GA", "QiYLLQB1dzz", "hrMZdCfV2bW", "rEDildbz3HQ", "ZZ5avPVTkb4", "OSfC7wdI4Qp", "iclr_2022_AkJyAE46GA", "iclr_2022_AkJyAE46GA", "iclr_2022_AkJyAE46GA", "iclr_2022_AkJyAE46GA", "iclr_2022_AkJyAE46GA" ]
iclr_2022_0U0C2pXfTZl	SLASH: Embracing Probabilistic Circuits into Neural Answer Set Programming	The goal of combining the robustness of neural networks and the expressivity of symbolic methods has rekindled the interest in Neuro-Symbolic AI. Recent advancements in Neuro-Symbolic AI often consider specifically-tailored architectures consisting of disjoint neural and symbolic components, and thus do not exhibit des...	Reject	The most positive reviewers have not decided to step forward to champion the paper. Others have a negative impression which has not sufficiently changed after the answers from authors. Actually, it is acknowledge that there have been many modifications, but they are not happy enough with this situation: modifications (...	train	[ "9NAaz91OPZv", "9RG7PhbLyWP", "fy6pQmC1HnK", "3Tf1LBLPsNC", "z9B1FqOrhRe", "DH6WYTlhFRY", "l2K8ME50pB", "3wor9PDLNOx", "xNGGOCwKwYX", "TnDumpmkzzq", "AQZCzKMnoHS", "MRh9h6HzA_", "mU4ICYsvLjV", "LFvZaGA_wKd", "U6AO511qm0o", "vHaKXs0kyar", "WckUVWVJeZ", "4QjBSCmCKV4", "EnCgI_bMOta"...	[ "author", "official_reviewer", "author", "author", "author", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "author",...	[ " Yes you are right. The different training examples can indeed realize different subsets of variables (overlapping as well) based on the query.\n\nIf this answers all the questions of the reviewer satisfactorily then we would appreciate if the reviewer can reconsider the overall evaluation.", " > Further, we wou...	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, 8, 5, 3 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 3, 2, 4, 5 ]	[ "9RG7PhbLyWP", "z9B1FqOrhRe", "MRh9h6HzA_", "mU4ICYsvLjV", "DH6WYTlhFRY", "3wor9PDLNOx", "tvJl27DuIZX", "Dc-EKvrcKBF", "IyC6_23J7-d", "AQZCzKMnoHS", "LFvZaGA_wKd", "vHaKXs0kyar", "4QjBSCmCKV4", "tvJl27DuIZX", "0w1RICzj7DZ", "tvJl27DuIZX", "TnL6N5Hb5Hu", "tvJl27DuIZX", "0w1RICzj7D...
iclr_2022_eqRTPB134q0	Invariance in Policy Optimisation and Partial Identifiability in Reward Learning	It is challenging to design a reward function for complex, real-world tasks. Reward learning algorithms let one instead infer a reward function from data. However, multiple reward functions often explain the data equally well, even in the limit of infinite data. Prior work has focused on situations where the reward fun...	Reject	The paper formally studies the problem of partial identifiability when inferring a reward function from a given data source (e.g., expert demonstrations or trajectory preferences). To formally characterize this ambiguity in a data source, the paper proposes considering the infinite-limit data regime, which bounds the r...	train	[ "lzvuhbHt3O", "g1l5lSK_gU_", "rJ4JnqDcE62", "tO_zCsa868J", "wwweB8ZYszu", "_nW2H25ObA0", "0SgNi00z3Yp", "_38VozoIBos", "0yxU8SAhpxw", "bSeQrLlEgWb", "1cL6fLwzyPR", "lCzVhHmE3O5" ]	[ "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ "This paper provides a theoretical framework to understand the relationship between data source/downstream tasks and reward functions. Some clarification questions:\n\n1. At the end of page 3, you mentioned the maximum entropy policy as the fixed point to $\\pi_\\beta = \\pi_{\\beta}^{\\pi_\\beta}$. Could you prov...	[ 8, -1, -1, -1, -1, -1, -1, -1, -1, 3, 3, 8 ]	[ 4, -1, -1, -1, -1, -1, -1, -1, -1, 4, 1, 4 ]	[ "iclr_2022_eqRTPB134q0", "iclr_2022_eqRTPB134q0", "lzvuhbHt3O", "lzvuhbHt3O", "lzvuhbHt3O", "lzvuhbHt3O", "lCzVhHmE3O5", "1cL6fLwzyPR", "bSeQrLlEgWb", "iclr_2022_eqRTPB134q0", "iclr_2022_eqRTPB134q0", "iclr_2022_eqRTPB134q0" ]
iclr_2022_qfLJBJf_DnH	Brain insights improve RNNs' accuracy and robustness for hierarchical control of continually learned autonomous motor motifs	We study the problem of learning dynamics that can produce hierarchically organized continuous outputs consisting of the flexible chaining of re-usable motor ‘motifs’ from which complex behavior is generated. Can a motif library be efficiently and extendably learned without interference between motifs, and can these mo...	Reject	This manuscript presents a method to allow RNNs to chain together sequences of behaviors. Reviewers had numerous concerns but the most important is that the problem posed here is solved by a simple method: resetting the state of the RNN before processing a motif. Overall, reviewers noted a few key topics, although thi...	train	[ "kuKyzTy8bF7", "xAb-UStqxF0", "VLo_Ko3XMAj", "zq3hrWYUia", "PCkfY5yJSTB", "927TReGGLr", "nRxgH8KnfMm", "geqqARyt4j6", "VpztyguVdsK", "k8w3RN8q94m", "cffZV65ix-F", "Vg1xZqbEOXH" ]	[ "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " I have reviewed the authors' comments and looked over the updated manuscript. I appreciate their efforts and they have addressed parts of my concerns, but I still stand by my original scores.", " To address the reviewers' comments, we have uploaded an improved draft that both (i) implements necessary clarificat...	[ -1, -1, -1, -1, -1, -1, -1, -1, 5, 3, 6, 3 ]	[ -1, -1, -1, -1, -1, -1, -1, -1, 3, 4, 2, 3 ]	[ "927TReGGLr", "iclr_2022_qfLJBJf_DnH", "zq3hrWYUia", "Vg1xZqbEOXH", "cffZV65ix-F", "nRxgH8KnfMm", "k8w3RN8q94m", "VpztyguVdsK", "iclr_2022_qfLJBJf_DnH", "iclr_2022_qfLJBJf_DnH", "iclr_2022_qfLJBJf_DnH", "iclr_2022_qfLJBJf_DnH" ]
iclr_2022_H-sddFpZAp4	ModeRNN: Harnessing Spatiotemporal Mode Collapse in Unsupervised Predictive Learning	Learning predictive models for unlabeled spatiotemporal data is challenging in part because visual dynamics can be highly entangled in real scenes, making existing approaches prone to overfit partial modes of physical processes while neglecting to reason about others. We name this phenomenon \textit{spatiotemporal mode...	Reject	The paper presents a novel architecture, ModeRNN, for unsupervised video prediction by learning spatiotemporal attention in the latent subspace (slots). ModeRNN effectively learns modular features using a set of mode slots and adaptively aggregates the slot features with learnable importance weights. The paper has pro...	train	[ "28OGcVH0r8F", "7obihHBY4hk", "czQm9V0j8Ce", "kFnkPFUNSaF", "a6qsz3QcxfF", "Q1oqYk3b1Rt", "S4ibZi6Y6Hk", "n6VFVR8EOW", "QLEHZCXPYB", "Qph6D3lQtIB", "c2jrokZm7db" ]	[ "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer" ]	[ "This paper proposes a mechanism to reduce spatiotemporal mode collapse in unsupervised predictive learning. To achieve this goal, the proposed method is built upon the idea that different latent modes in the same data domain should share a set of hidden representation subspaces, which can be represented with vario...	[ 5, 8, -1, -1, -1, -1, -1, -1, -1, -1, 5 ]	[ 3, 3, -1, -1, -1, -1, -1, -1, -1, -1, 4 ]	[ "iclr_2022_H-sddFpZAp4", "iclr_2022_H-sddFpZAp4", "kFnkPFUNSaF", "c2jrokZm7db", "7obihHBY4hk", "28OGcVH0r8F", "Q1oqYk3b1Rt", "QLEHZCXPYB", "iclr_2022_H-sddFpZAp4", "iclr_2022_H-sddFpZAp4", "iclr_2022_H-sddFpZAp4" ]
iclr_2022_4tOrvK-fFOR	Sound Source Detection from Raw Waveforms with Multi-Scale Synperiodic Filterbanks	Accurately estimating sound sources' temporal location, spatial location and semantic identity label from multi-channel sound raw waveforms is crucial for an agent to understand the 3D environment acoustically. Multiple sounds form a complex waveform mixture in time, frequency and space, so accurately detecting them re...	Reject	This work studies the task of sound source localization from multi-channel audio. An approach to design a wavelet-like filter bank for audio feature extraction is proposed. After discussion, all reviewers have given this work borderline ratings. Concerns were raised about the quality of the writing, missing related wo...	test	[ "JbOQ4a5Pwj-", "H6L6zNnY2mM", "2rHTAoPSpb2", "Zggl88fNuyq", "6kKiFRpNxjy", "Z4YxshYGHyI", "KuSS41-mH4W", "baDzUwd7Ljd", "RxVAILVb_kN", "izdSNk36KZ3", "47NGToJhMvs" ]	[ "author", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " We thank the reviewer for the further comments. We also provide more feedback to make our paper clearer.\n\nQ1: Time-Frequency resolution and uncertainty principle appears a bit superfluous. \n\nA1: The TF resolution and uncertainty principle discussion serves as the motivation of synperiodic filter bank...	[ -1, -1, 6, -1, -1, -1, -1, -1, 6, 5, 6 ]	[ -1, -1, 5, -1, -1, -1, -1, -1, 3, 4, 2 ]	[ "H6L6zNnY2mM", "baDzUwd7Ljd", "iclr_2022_4tOrvK-fFOR", "2rHTAoPSpb2", "RxVAILVb_kN", "iclr_2022_4tOrvK-fFOR", "47NGToJhMvs", "izdSNk36KZ3", "iclr_2022_4tOrvK-fFOR", "iclr_2022_4tOrvK-fFOR", "iclr_2022_4tOrvK-fFOR" ]
iclr_2022_GsH-K1VIyy	Data-Driven Offline Optimization for Architecting Hardware Accelerators	To attain higher efficiency, the industry has gradually reformed towards application-specific hardware accelerators. While such a paradigm shift is already starting to show promising results, designers need to spend considerable manual effort and perform large number of time-consuming simulations to find accelerators t...	Accept (Poster)	The authors give an effective framework PRIME to tackle the challenges of automating hardware design optimization. This problem is of importance to the community. Overall, the reviewers thought the paper gave a nice clean approach to the problem and that the community would be interested with these results.	train	[ "V5XM5iD48Zj", "isV66KqhsFb", "opiML2IkAQ6", "9UuxmH63N7P", "up5VnIUrhlN", "ou8iwg-AS0V", "FGX8hroRrnD", "1TgebXBl1dy", "pT--Yyww9DE", "XMbjHwHkxKv" ]	[ "author", "official_reviewer", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Dear Reviewer 7gpC,\n\nWe have now updated the paper to add an ablation study in a new Appendix D (changes in $\\textcolor{magenta}{magenta}$) aimed at identifying the minimal subset of training applications that still allow us to generalize to all the nine applications considered in the paper. While exhausti...	[ -1, 8, -1, -1, -1, -1, -1, 6, 6, 8 ]	[ -1, 4, -1, -1, -1, -1, -1, 3, 3, 4 ]	[ "ou8iwg-AS0V", "iclr_2022_GsH-K1VIyy", "isV66KqhsFb", "XMbjHwHkxKv", "opiML2IkAQ6", "pT--Yyww9DE", "1TgebXBl1dy", "iclr_2022_GsH-K1VIyy", "iclr_2022_GsH-K1VIyy", "iclr_2022_GsH-K1VIyy" ]
iclr_2022_psQ6wcNXjS1	EBM Life Cycle: MCMC Strategies for Synthesis, Defense, and Density Modeling	This work presents strategies to learn an Energy-Based Model (EBM) according to the desired length of its MCMC sampling trajectories. MCMC trajectories of different lengths correspond to models with different purposes. Our experiments cover three different trajectory magnitudes and learning outcomes: 1) shortrun sampli...	Reject	This paper proposed a strategy to train EBMs according to the length of MCMC trajectories required. The paper covers three settings with the different length of MCMC: image synthesis, adversarial defense, and density estimation. The reviewers generally find that there are interesting ideas and promising results in the ...	train	[ "ukZgSJNnh5h", "VDbyp85m-o", "sm2k8nCrkID", "GJEWa87fwl", "dW1pziE9UTm", "kQkegsFo2w", "uIz5rt4nPVh", "umQ-ZNKeETJ", "7brW300127m", "6LRmhlCZVe", "e24mo1Ton0_", "JcQB3bDmPQ", "6N9PUwHWMzq", "PZnZomD5GHp" ]	[ "official_reviewer", "official_reviewer", "author", "official_reviewer", "author", "author", "author", "author", "author", "author", "author", "author", "official_reviewer", "official_reviewer" ]	[ "The paper discusses learning strategies for energy-based models, short sampling for image generation, midrun sampling for adversarial defense, and longrun sampling for density estimation. The paper claims these methods achieve significant performance gain across the three applications and achieved state-of-the-art...	[ 3, 3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 8, 5 ]	[ 4, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, 4 ]	[ "iclr_2022_psQ6wcNXjS1", "iclr_2022_psQ6wcNXjS1", "dW1pziE9UTm", "6LRmhlCZVe", "iclr_2022_psQ6wcNXjS1", "PZnZomD5GHp", "umQ-ZNKeETJ", "VDbyp85m-o", "ukZgSJNnh5h", "e24mo1Ton0_", "6N9PUwHWMzq", "iclr_2022_psQ6wcNXjS1", "iclr_2022_psQ6wcNXjS1", "iclr_2022_psQ6wcNXjS1" ]
iclr_2022_Lv-G9XqLRRy	Restricted Category Removal from Model Representations using Limited Data	Deep learning models are trained on multiple categories jointly to solve several real-world problems. However, there can be cases where some of the classes may become restricted in the future and need to be excluded after the model has already been trained on them (Class-level Privacy). It can be due to privacy, ethica...	Reject	The paper proposes a technique to efficiently retrain a model when a small number of classes are required to be removed. Reviewers in general like the paper, but the key issue is motivation for the problem. The motivating examples in the rebuttal are not very good because a. authors do not provide any evidence that su...	val	[ "lICAQuPHY7m", "QLjmJH0Xtls", "kh6m6WuwBXW", "vMAq7nhUSGW", "81Zb9ufHcvi", "0yyGbeCUSkL", "dWK2pKdofdN", "a0rU2DVE7u", "Jvh7qc98UAA", "gLsumLKw48D", "Oblyg6xXMkZ", "5XULAEVr1Ni", "jtrZKp5L2Sx", "gEqNLWFEnfa", "x_pZ8SaDIiH", "EW4aCEM4edT", "wWo5aDISV3M", "TocRM-rIOkO", "znXTuTLubU...	[ "author", "official_reviewer", "official_reviewer", "author", "author", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "official_reviewer", "official_reviewer", "author", "author", "author", "official_reviewer", "official_reviewer", "author", ...	[ " Dear Reviewer,\n\nThank you for increasing your rating and for your response. \n\nWe would like to point out that the knowledge-distillation based regularization loss (eq 5) used in our approach prevents the output logits corresponding to the non-excluded classes from changing for the images of the excluded class...	[ -1, 5, -1, -1, -1, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, 6, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 5 ]	[ -1, 4, -1, -1, -1, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 4 ]	[ "kh6m6WuwBXW", "iclr_2022_Lv-G9XqLRRy", "0yyGbeCUSkL", "iclr_2022_Lv-G9XqLRRy", "a0rU2DVE7u", "81Zb9ufHcvi", "iclr_2022_Lv-G9XqLRRy", "-P1bsd_emWX", "TocRM-rIOkO", "jtrZKp5L2Sx", "5XULAEVr1Ni", "EW4aCEM4edT", "gEqNLWFEnfa", "GoMKgEHDsTW", "-P1bsd_emWX", "HqOvh0FgSSG", "iclr_2022_Lv-G...
iclr_2022_sEIl_stzQyB	Greedy-based Value Representation for Efficient Coordination in Multi-agent Reinforcement Learning	Due to the representation limitation of the joint Q value function, multi-agent reinforcement learning (MARL) methods with linear or monotonic value decomposition can not ensure the optimal consistency (i.e. the correspondence between the individual greedy actions and the maximal true Q value), leading to instability a...	Reject	The paper studies the join-Q value decomposition problem in MARL. Some of the results are interesting, e.g., the True-Global-Max condition and several experiments. However, the majority of the reviews are negative due to the current presentation of the paper. We encourage the authors address all the reviewers' comments...	train	[ "m4fPoTIrKx2", "5stwl1I8KsU", "GFhMCkmIqc", "368D_7DwwGeA", "jqbsRWgDXj5", "-9bzk8bae9F", "cbqL0UsK0dB", "N6d0A8axSQp", "evImWrLdJAX", "9e3A3rYtvNP" ]	[ "official_reviewer", "official_reviewer", "official_reviewer", "author", "author", "author", "author", "official_reviewer", "official_reviewer", "official_reviewer" ]	[ " Thank you for the author's responses. \n\n (1) In author's response, author says the joint Q function is modeled by neural network. Then I think the utility funcitons in Eq(3) should also be modeled by neural network so we don't know the utility function. However, the author makes the utility function as expactat...	[ -1, 3, -1, -1, -1, -1, -1, 6, 3, 3 ]	[ -1, 3, -1, -1, -1, -1, -1, 2, 4, 3 ]	[ "-9bzk8bae9F", "iclr_2022_sEIl_stzQyB", "368D_7DwwGeA", "5stwl1I8KsU", "evImWrLdJAX", "9e3A3rYtvNP", "N6d0A8axSQp", "iclr_2022_sEIl_stzQyB", "iclr_2022_sEIl_stzQyB", "iclr_2022_sEIl_stzQyB" ]

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.