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_YpPiNigTzMT | Universalizing Weak Supervision | Weak supervision (WS) frameworks are a popular way to bypass hand-labeling large datasets for training data-hungry models.
These approaches synthesize multiple noisy but cheaply-acquired estimates of labels into a set of high-quality pseudo-labels for downstream training. However, the synthesis technique is specific to... | Accept (Poster) | The paper propose a universal technique that enables weak supervision over any label type while still offering desirable properties, including practical flexibility, computational efficiency, and theoretical guarantees.
Over the course of the rebuttal, the authors have made a substantial overhaul on writing and experi... | train | [
"CDJQYlHsLbm",
"bs8zAz6RDC_",
"indn7WDyAS5",
"7iI1GVHNvw",
"QkNkHTUBynD",
"-tfANKlaKt4",
"Dmh_wJoE2gq",
"oQa9hItnJg",
"Ni4z4Te65zj",
"MyYO-sEV1lw",
"stSjQuj4DFH"
] | [
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Dear Reviewers,\n\nWe want to thank you again for your feedback, questions, and suggestions. We believe we have answered all of your questions in our responses and the updated draft. \n\nIf you have further questions, we would love to answer them. ",
"This work studies a weakly supervised learning setup of aggr... | [
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
5,
3,
8
] | [
-1,
2,
-1,
-1,
-1,
-1,
-1,
-1,
3,
4,
3
] | [
"oQa9hItnJg",
"iclr_2022_YpPiNigTzMT",
"stSjQuj4DFH",
"MyYO-sEV1lw",
"bs8zAz6RDC_",
"Ni4z4Te65zj",
"Ni4z4Te65zj",
"iclr_2022_YpPiNigTzMT",
"iclr_2022_YpPiNigTzMT",
"iclr_2022_YpPiNigTzMT",
"iclr_2022_YpPiNigTzMT"
] |
iclr_2022_Vr_BTpw3wz | Hindsight: Posterior-guided training of retrievers for improved open-ended generation | Many text generation systems benefit from retrieving passages from a textual knowledge corpus (e.g., Wikipedia) and using them to generate the output. For open-ended generation tasks, like generating informative utterances in conversations, many varied passages $z$ are relevant to the context $x$ but few are relevant t... | Accept (Poster) | The authors study the problem of open-ended knowledge-grounded natural language generation, in the context of free-form QA or knowledge-grounded dialogue, focusing on improving the retrieval component of the retrieval-augmented system. By retrieving more relevant passages, the generations are more grounded in retrieved... | train | [
"gXhDidH6EvK",
"ebNAy8Z6EqD",
"9bJN9XR3NMI",
"HjWYfyW0VdJ",
"fY4UlphaKGN",
"-tCb9u1N4EZ",
"QTRVwrYWpK2",
"MW4Pbe5q33W",
"_2b2aA-cPCV",
"72rv134gFbW",
"xGV9e74i87J",
"ooN4GKhj2Ii",
"hF5ULAL6vVv",
"uklxj3juEQY"
] | [
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
" Thank you for the positive feedback and suggestions.\n\nThe main concern raised was regarding a lack of baselines. \n\n1 - We would like to point out that Marginalized Loss is essentially the same training objective used in \"Retrieval-augmented generation for knowledge-intensive nlp tasks\" by Lewis et al. Ther... | [
-1,
8,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
6
] | [
-1,
4,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3
] | [
"uklxj3juEQY",
"iclr_2022_Vr_BTpw3wz",
"xGV9e74i87J",
"MW4Pbe5q33W",
"iclr_2022_Vr_BTpw3wz",
"iclr_2022_Vr_BTpw3wz",
"hF5ULAL6vVv",
"fY4UlphaKGN",
"iclr_2022_Vr_BTpw3wz",
"ebNAy8Z6EqD",
"ebNAy8Z6EqD",
"fY4UlphaKGN",
"iclr_2022_Vr_BTpw3wz",
"iclr_2022_Vr_BTpw3wz"
] |
iclr_2022_qhkFX-HLuHV | Can an Image Classifier Suffice For Action Recognition? | We explore a new perspective on video understanding by casting the video recognition problem as an image recognition task. Our approach rearranges input video frames into super images, which allow for training an image classifier directly to fulfill the task of action recognition, in exactly the same way as image class... | Accept (Poster) | This paper regards video understanding as an image classification task, and reports promising performance against state of the arts on several standard benchmarks. Though the method is quite simple, it achieves good results. The visualization in this paper also provides good insight. All reviewers give positive recomme... | train | [
"wyhry-NsJ-X",
"Plaqb37P-XQ",
"Qpm3he-iGeI",
"6Cy3LMuOAMr",
"8ylf25Sp3Vc",
"Xe6EsYeNwXA",
"5BSzeaJCbHF",
"BZrHIwHXX63",
"_9VPq_275L0",
"w4Kq-AdQssc",
"cGE4HvsZ6j2",
"iK2zDfLA81n",
"RaOjVhd1Dsy"
] | [
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Thank you for the reply.\nYes I think if we can clearly specify which scheme we are comparing against (joint-space-time attention, and not divided-space-time) in the text, then I think it would help avoid confusion. Thanks!\n\nThe authors have addressed my concerns, and I vote for acceptance of the paper.",
" W... | [
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
8,
8,
8
] | [
-1,
-1,
-1,
-1,
5,
-1,
-1,
-1,
-1,
-1,
4,
3,
3
] | [
"Plaqb37P-XQ",
"6Cy3LMuOAMr",
"5BSzeaJCbHF",
"_9VPq_275L0",
"iclr_2022_qhkFX-HLuHV",
"8ylf25Sp3Vc",
"iK2zDfLA81n",
"iclr_2022_qhkFX-HLuHV",
"RaOjVhd1Dsy",
"cGE4HvsZ6j2",
"iclr_2022_qhkFX-HLuHV",
"iclr_2022_qhkFX-HLuHV",
"iclr_2022_qhkFX-HLuHV"
] |
iclr_2022_Kef8cKdHWpP | DiffSkill: Skill Abstraction from Differentiable Physics for Deformable Object Manipulations with Tools | We consider the problem of sequential robotic manipulation of deformable objects using tools.
Previous works have shown that differentiable physics simulators provide gradients to the environment state and help trajectory optimization to converge orders of magnitude faster than model-free reinforcement learning algorit... | Accept (Poster) | Manipulating deformable objects is an up-and-coming topic in robotics and machine learning, and it creates many interesting scientific and real-world challenges. The paper looks into long horizon tasks of manipulation of deformable objects, using an interesting mix of more local trajectory optimization and differentiab... | train | [
"6zHh_Q3SEJn",
"6xvNnDTRGD",
"HVYuZRh3nt",
"KeMcWYy1rkf",
"Q6CdPoGtVZO",
"p27nmmssp97",
"M0Gr5UJX2Hg",
"3rNFT6_8H_T",
"A36QUbDetJy",
"-1-DBWY8jpc",
"9LTxyMHGOLC",
"dw4K8F6_D0",
"oMMUyhRaJng",
"wTJRhXZNFeu",
"QK6NO6spYhy",
"jyvBuTQmnxF"
] | [
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
" Dear Authors,\n\nThank you for addressing the points I made for updating the website with the videos for the baselines. This is a strong work!\n\nBest, Reviewer H1Qf",
"This paper proposes a method to learn long-horizon control policies. It uses differentiable physics simulator to generate skills for the latte... | [
-1,
6,
-1,
-1,
-1,
10,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
6
] | [
-1,
3,
-1,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
3
] | [
"wTJRhXZNFeu",
"iclr_2022_Kef8cKdHWpP",
"3rNFT6_8H_T",
"dw4K8F6_D0",
"oMMUyhRaJng",
"iclr_2022_Kef8cKdHWpP",
"3rNFT6_8H_T",
"A36QUbDetJy",
"-1-DBWY8jpc",
"6xvNnDTRGD",
"iclr_2022_Kef8cKdHWpP",
"jyvBuTQmnxF",
"p27nmmssp97",
"QK6NO6spYhy",
"iclr_2022_Kef8cKdHWpP",
"iclr_2022_Kef8cKdHWpP"... |
iclr_2022_Fn7i_r5rR0q | Do deep networks transfer invariances across classes? | In order to generalize well, classifiers must learn to be invariant to nuisance transformations that do not alter an input's class. Many problems have "class-agnostic" nuisance transformations that apply similarly to all classes, such as lighting and background changes for image classification. Neural networks can lear... | Accept (Poster) | This paper investigates how well properties invariant to changes such as lightening and background learned in the major class can be transferred to the minor class. In this paper, the authors reveal that invariances do not transfer well to small classes, and suggest that resolving this phenomenon can help increase the ... | train | [
"CmfWv58ne1",
"EGtNb3YvFjU",
"tEs7vqKW3MG",
"Z364X2hGnXN",
"HYvBh4ALOT",
"IPPWQrrzs10",
"QRZMPtXl1Y2",
"fvwHoQRtSN7",
"fa0IJ3mYKWi",
"Wkac5TOcv8I",
"qD5MLmifErr",
"pkSTbJfKgsG",
"1pXhAg9eFg9",
"RlQHZK2wPBi",
"D_1ZdsJ8j4s",
"j3M9U3aN3s",
"ey5GMxiLofb",
"SliPIFMQAcu",
"jrSACmLdWfF"... | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
" Sorry for the late reply. After thoroughly reading all the reviews especially the Official Review of Paper879 by Reviewer PgZ2, I decided to further raise my score to 6 to encourage the novel perspective of the long-tailed problem proposed by this paper.",
"=======================\nSummary: \nThis paper studies... | [
-1,
6,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
5
] | [
-1,
2,
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
5
] | [
"Wkac5TOcv8I",
"iclr_2022_Fn7i_r5rR0q",
"D_1ZdsJ8j4s",
"HYvBh4ALOT",
"fvwHoQRtSN7",
"iclr_2022_Fn7i_r5rR0q",
"fa0IJ3mYKWi",
"jrSACmLdWfF",
"IPPWQrrzs10",
"D_1ZdsJ8j4s",
"SliPIFMQAcu",
"EGtNb3YvFjU",
"jrSACmLdWfF",
"IPPWQrrzs10",
"pkSTbJfKgsG",
"jrSACmLdWfF",
"IPPWQrrzs10",
"iclr_20... |
iclr_2022_hcMvApxGSzZ | Fixed Neural Network Steganography: Train the images, not the network | Recent attempts at image steganography make use of advances in deep learning to train an encoder-decoder network pair to hide and retrieve secret messages in images. These methods are able to hide large amounts of data, but they also incur high decoding error rates (around 20%). In this paper, we propose a novel algori... | Accept (Poster) | This submission proposes a method for steganography, i.e. hiding "secret messages" in images. Specifically, the proposed approach implements a procedure similar to adversarial example generation, where a perturbation is found that a) is imperceptible and b) can be decoded by a fixed decoder. This approach results in th... | train | [
"R4hi4-46K12",
"G3FepxSFZ2",
"kpdEqrLRrcE",
"RxwDV2mVY-L",
"9P8eIUVRZRZ",
"cUnFOrMS6X",
"KnbZSADsc1",
"JDkrJMAAL3j",
"dn8rrPg09GW",
"eKJxFtHhW37",
"ebndvSEkJCa",
"MrSQsA-cKp-",
"-A_m6G-WkcZ",
"Ue8HV0fzW4L"
] | [
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
" Thank you for the follow-up. Here are a few additional details:\n\n* For steganography, Uniward can hide up to 0.5 bpp and we will add this number to the quantitative comparison section of the paper. \n* For Steganalysis, we agree with the reviewer that for some applications, steganographic images being undetecta... | [
-1,
8,
-1,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
8,
5
] | [
-1,
4,
-1,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
3,
5
] | [
"RxwDV2mVY-L",
"iclr_2022_hcMvApxGSzZ",
"eKJxFtHhW37",
"KnbZSADsc1",
"JDkrJMAAL3j",
"iclr_2022_hcMvApxGSzZ",
"Ue8HV0fzW4L",
"dn8rrPg09GW",
"MrSQsA-cKp-",
"G3FepxSFZ2",
"-A_m6G-WkcZ",
"cUnFOrMS6X",
"iclr_2022_hcMvApxGSzZ",
"iclr_2022_hcMvApxGSzZ"
] |
iclr_2022_C1_esHN6AVn | Learning Synthetic Environments and Reward Networks for Reinforcement Learning | We introduce Synthetic Environments (SEs) and Reward Networks (RNs), represented by neural networks, as proxy environment models for training Reinforcement Learning (RL) agents. We show that an agent, after being trained exclusively on the SE, is able to solve the corresponding real environment. While an SE acts as a f... | Accept (Poster) | This paper proposes a new method for generating synthetic environments and reward networks for reinforcement learning tasks. This happens as a nested process: policies are learned in an inner loop, and environments are evolved in an outer loop. The environment representation is quite simple: the parameters of an MDP. S... | train | [
"itXOJMEA9A",
"TiihJaeGD_6",
"kG-za7eEsmv",
"6ihfnhFq4V7",
"525tkFi_X0j",
"F0hE26NdwPl",
"nxPf3dUIMoq",
"46HgNUPsMir",
"jL5syahnP0",
"iCiKEfax3BC",
"PP-DjeuDoir",
"DShwhBSh0A",
"pL-K7K2fr2",
"vYYb9BXEqnw",
"hATN8m66oBT",
"0OpUsbtxmz3",
"7LnUREB0BS",
"ErKRc2Qx2yG",
"3uZZLn5OXu",
... | [
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
... | [
" Please see my previous comment, which highlighted the portions of my original review that your response failed to address.",
" Dear reviewer,\n\nConsidering the breadth and depth of our initial response and your relatively short response in return, we would like to point out that only through *actionable* feedb... | [
-1,
-1,
6,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
3
] | [
-1,
-1,
4,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
3
] | [
"TiihJaeGD_6",
"F0hE26NdwPl",
"iclr_2022_C1_esHN6AVn",
"iclr_2022_C1_esHN6AVn",
"nxPf3dUIMoq",
"vYYb9BXEqnw",
"PP-DjeuDoir",
"mfINjla2QH",
"eVsuRF6YQTa",
"DShwhBSh0A",
"kG-za7eEsmv",
"kG-za7eEsmv",
"6ihfnhFq4V7",
"s4BAoDkpYWz",
"s4BAoDkpYWz",
"s4BAoDkpYWz",
"s4BAoDkpYWz",
"GbSZnNGl... |
iclr_2022_6ET9SzlgNX | Understanding Intrinsic Robustness Using Label Uncertainty | A fundamental question in adversarial machine learning is whether a robust classifier exists for a given task. A line of research has made some progress towards this goal by studying the concentration of measure, but we argue standard concentration fails to fully characterize the intrinsic robustness of a classificatio... | Accept (Poster) | The paper was praised for being clearly written, well-motivated, and for addressing an important problem: measuring intrinsic robustness.
It improves the previous results on intrinsic robustness based on concentration of data distribution, by incorporating the constraint on the label uncertainty of the models.
This req... | train | [
"it3PmTP-so",
"RUFBSbcfqRC",
"QnexAr3c1gy",
"6Dl-PfH4CHa",
"IZoXjtVazAB",
"YYG2IGliEJH",
"jXPk1y5xmTX",
"3E3haWzrQPr",
"fVo9d1oF4pi",
"8jKktUgdYVS",
"z7zHUGDgLeI"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"The paper proposes to include label uncertainty (LU) in the formulation of the concentration of measure problem which has been deemed the cause of adversarial vulnerability. It suggests that the current formulation of intrinsic robustness based on concentration measures is insufficient because of the exclusion of ... | [
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
8,
6
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
2,
3
] | [
"iclr_2022_6ET9SzlgNX",
"6Dl-PfH4CHa",
"iclr_2022_6ET9SzlgNX",
"z7zHUGDgLeI",
"jXPk1y5xmTX",
"fVo9d1oF4pi",
"3E3haWzrQPr",
"it3PmTP-so",
"iclr_2022_6ET9SzlgNX",
"iclr_2022_6ET9SzlgNX",
"iclr_2022_6ET9SzlgNX"
] |
iclr_2022_s03AQxehtd_ | ProtoRes: Proto-Residual Network for Pose Authoring via Learned Inverse Kinematics | Our work focuses on the development of a learnable neural representation of human pose for advanced AI assisted animation tooling. Specifically, we tackle the problem of constructing a full static human pose based on sparse and variable user inputs (e.g. locations and/or orientations of a subset of body joints). To sol... | Accept (Oral) | This paper proposes a novel representation for pose authoring, and was uniformly lauded by all reviewers. The AC concurs this paper is far above the threshold for acceptance at ICLR. | train | [
"E0uQprqPI2J",
"7GRFYc6YqS7",
"RemzOSX7NoB",
"gF90g3dhX4",
"PR7SHf9oa9",
"iW9VtQ6tfu1",
"xvSy7MKVV7i",
"cA7yifo2Ud8",
"w9FI5jqW1bG",
"eOmUmLkLGOQ",
"W3LbZn7cSzZ",
"1cWNZvvUjU8",
"1iqjK2T8WMS",
"rBRfIfeUW1m",
"afeHf5FQyol",
"up_eQiSPTF",
"Blb8QV65oLk",
"1pPRRKH4eqE",
"8IeHf25rO0P"... | [
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
... | [
" We would like to thank Reviewer J8Sg for the additional feedback! To address the point raised by the reviewer, we will additionally report the quantitative result for ProtoRes in the full constraints scenario (the positions and rotations of all joints are given) in the supplementary of the revised manuscript.",
... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
6
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
3
] | [
"RemzOSX7NoB",
"PR7SHf9oa9",
"iW9VtQ6tfu1",
"w9FI5jqW1bG",
"xvSy7MKVV7i",
"cA7yifo2Ud8",
"1iqjK2T8WMS",
"1cWNZvvUjU8",
"iclr_2022_s03AQxehtd_",
"ssGxfwILAgA",
"Nu0svYjWoZX",
"Nu0svYjWoZX",
"Nu0svYjWoZX",
"Nu0svYjWoZX",
"MiXUlTyakU",
"MiXUlTyakU",
"MiXUlTyakU",
"w9FI5jqW1bG",
"w9F... |
iclr_2022_V3C8p78sDa | Exploring the Limits of Large Scale Pre-training | Recent developments in large-scale machine learning suggest that by scaling up data, model size and training time properly, one might observe that improvements in pre-training would transfer favorably to most downstream tasks. In this work we systematically study this phenomena and establish that, as we increase the ... | Accept (Spotlight) | This paper provides a very large-scale study on the pretraining of image recognition models. Specifically, three scaling factors (model sizes, dataset sizes, and training time) are extensively investigated. One important phenomenon observed by this paper is that stronger upstream accuracy may not necessarily contribute... | train | [
"Kj0OkKot7z8",
"bZFnIaZkolW",
"PMsPhNkrBOq",
"GJqsPDAM9i",
"xkBbg9UEyO",
"cMD1ip2HX8f",
"wLl1Txg2l4D",
"PSzFan1eKp"
] | [
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Dear reviewer 3WYy,\n\nThank you so much for the insightful review and the interesting comments and questions. \n\nHereby, we respond to each of your concerns:\n\n1. We don’t argue that model size, US data size, and compute can not predict DS accuracy, we argue that all these scaling variables contribute toward p... | [
-1,
-1,
-1,
-1,
8,
6,
8,
8
] | [
-1,
-1,
-1,
-1,
3,
3,
3,
3
] | [
"PSzFan1eKp",
"wLl1Txg2l4D",
"cMD1ip2HX8f",
"xkBbg9UEyO",
"iclr_2022_V3C8p78sDa",
"iclr_2022_V3C8p78sDa",
"iclr_2022_V3C8p78sDa",
"iclr_2022_V3C8p78sDa"
] |
iclr_2022_8FhxBtXSl0 | CKConv: Continuous Kernel Convolution For Sequential Data | Conventional neural architectures for sequential data present important limitations. Recurrent neural networks suffer from exploding and vanishing gradients, small effective memory horizons, and must be trained sequentially. Convolutional neural networks cannot handle sequences of unknown size and their memory horizon ... | Accept (Poster) | This paper introduces a convolution where the kernel is parametrised continuously over time (in the context of recurrent networks) to address vanishing gradients issues, by using another neural network to generate the kernels.
This is a meaningful idea, addressing an important problem.
The paper is well written and cle... | train | [
"sMa5pQXsu2L",
"GXKklTlgmTG",
"D2Xrg7RSAJS",
"6Ig5Z1G4Wwn",
"p91R48lO0L",
"S2rDzq2tgM",
"dGDDOJZ-TVu",
"O_CSR0fUGtu",
"nQMth1ijVGo",
"BcYMxOfgbn",
"zrOxS9m7v8a",
"zuF5BNOgAf",
"2cn0JbChIQv"
] | [
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"The paper introduces convolutions with continuously parametrized kernels for sequential data. Continuous kernel convolutions, CKConv for short, parametrize the kernel associated to a convolutional layer as a continuous mapping, $\\psi: \\mathbb{R}^+ \\rightarrow \\mathbb{R}^{N_{out} \\times N_{in}}$, from the rela... | [
8,
-1,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
8,
6
] | [
4,
-1,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3
] | [
"iclr_2022_8FhxBtXSl0",
"D2Xrg7RSAJS",
"dGDDOJZ-TVu",
"BcYMxOfgbn",
"iclr_2022_8FhxBtXSl0",
"zrOxS9m7v8a",
"O_CSR0fUGtu",
"sMa5pQXsu2L",
"zuF5BNOgAf",
"p91R48lO0L",
"2cn0JbChIQv",
"iclr_2022_8FhxBtXSl0",
"iclr_2022_8FhxBtXSl0"
] |
iclr_2022_T0B9AoM_bFg | Improving Mutual Information Estimation with Annealed and Energy-Based Bounds | Mutual information (MI) is a fundamental quantity in information theory and machine learning. However, direct estimation of mutual information is intractable, even if the true joint probability density for the variables of interest is known, as it involves estimating a potentially high-dimensional log partition functio... | Accept (Poster) | This paper investigates a tighter bound for mutual information and proposes some novel estimators of MI from the importance sampling perspective. The proposed approach provides a unifying framework for mutual information bounds that can deduce many existing approaches. The theoretical and experimental analyses well jus... | train | [
"8QrojjuuPc",
"FjcVvyDBqh",
"gWlvqfGtmu",
"MHYijowP_jR",
"l7k2X3jVNR",
"0QW0fwUXeyg",
"qUpyFBTBzZ0"
] | [
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" First, we would like to thank the reviewer for the constructive feedback in both rounds for reviews, which have significantly improved our paper.\n\n> The derivation of GIWAE is one of the key contributions of this paper, but there is no intuition provided in the main text on how this is derived...\n> The current... | [
-1,
-1,
-1,
-1,
8,
6,
8
] | [
-1,
-1,
-1,
-1,
4,
4,
5
] | [
"qUpyFBTBzZ0",
"0QW0fwUXeyg",
"l7k2X3jVNR",
"iclr_2022_T0B9AoM_bFg",
"iclr_2022_T0B9AoM_bFg",
"iclr_2022_T0B9AoM_bFg",
"iclr_2022_T0B9AoM_bFg"
] |
iclr_2022_DNRADop4ksB | On the Importance of Firth Bias Reduction in Few-Shot Classification | Learning accurate classifiers for novel categories from very few examples, known as few-shot image classification, is a challenging task in statistical machine learning and computer vision. The performance in few-shot classification suffers from the bias in the estimation of classifier parameters; however, an effective... | Accept (Spotlight) | This work starts from the observation that maximum likelihood estimation, while consistent, has a bias on a finite sample which is likely to hurt for small sample sizes. From this, they apply Firth bias reduction to the few-shot learning setting and demonstrate its empirical benefits, notably relatively to L2 regulariz... | test | [
"5uVv-mHxanl",
"1CMVqWJswLh",
"D8iv1-6Ez5S",
"5REYpN91mn",
"OIZXujwANmy",
"dpDk7rlSRat",
"WPxemMOXO3",
"HBmS5voiEmG",
"2R0nMsKGOjw",
"sWF2a7XWuYI",
"_Sh1J8-gxGu",
"zaiMy6-4X78",
"0gafaMWEKT7"
] | [
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Thanks to the authors for these clarifications and complements. There are no more issues for me, so I keep my score unchanged as accept.",
" **Question 1:**\nAccording to the intro, which I quote here \"We achieve this by deriving a simplified yet effective Firth formulation that penalizes the KL-divergence bet... | [
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
8,
6
] | [
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
5,
4
] | [
"dpDk7rlSRat",
"_Sh1J8-gxGu",
"iclr_2022_DNRADop4ksB",
"HBmS5voiEmG",
"0gafaMWEKT7",
"zaiMy6-4X78",
"_Sh1J8-gxGu",
"D8iv1-6Ez5S",
"iclr_2022_DNRADop4ksB",
"zaiMy6-4X78",
"iclr_2022_DNRADop4ksB",
"iclr_2022_DNRADop4ksB",
"iclr_2022_DNRADop4ksB"
] |
iclr_2022_RftryyYyjiG | Exploring extreme parameter compression for pre-trained language models | Recent work explored the potential of large-scale Transformer-based pre-trained models, especially Pre-trained Language Models (PLMs) in natural language processing. This raises many concerns from various perspectives, e.g., financial costs and carbon emissions.
Compressing PLMs like BERT with negligible performance ... | Accept (Poster) | This paper reviews a number of parameter decomposition methods for BERT style contextual embedding models. The authors argue for the application of Tucker decomposition to the attention and feedforward layers of such models. Evaluation is performed for a range of models on the GLUE benchmark. Further ablation studies i... | train | [
"bCJsdvDnsYT",
"ibm1z6I30-H",
"r9h7B4eSnin",
"Hfe4DcQ7Rzt",
"NksV-fBAPaZ",
"sXgGHGMq2UX",
"iDJNqOg09PI",
"VzFum59j1e",
"t6ncf2YsVTE",
"f1EsE6v9cDq",
"zxJV22GypQo",
"O5hoJu3GJ-z",
"lF-O9eqsLzD",
"qxnQ5-J_egJ",
"NwdCFToDeqh",
"-67AtyFnFQl",
"fDs1N9uLH84"
] | [
"author",
"author",
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
" Thank you again for contributing time to provide comments and reviews in this paper, which have significantly improved this paper. Can we provide any further explanations for your concerns that might still exist?\n\nWe really appreciate it if you could provide any further feedback.",
" **R 2.3 why compress SAN... | [
-1,
-1,
-1,
6,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
5
] | [
-1,
-1,
-1,
4,
-1,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4
] | [
"fDs1N9uLH84",
"fDs1N9uLH84",
"iclr_2022_RftryyYyjiG",
"iclr_2022_RftryyYyjiG",
"iclr_2022_RftryyYyjiG",
"iclr_2022_RftryyYyjiG",
"sXgGHGMq2UX",
"Hfe4DcQ7Rzt",
"sXgGHGMq2UX",
"Hfe4DcQ7Rzt",
"-67AtyFnFQl",
"sXgGHGMq2UX",
"fDs1N9uLH84",
"fDs1N9uLH84",
"-67AtyFnFQl",
"iclr_2022_RftryyYyji... |
iclr_2022_KB5onONJIAU | Comparing Distributions by Measuring Differences that Affect Decision Making | Measuring the discrepancy between two probability distributions is a fundamental problem in machine learning and statistics. We propose a new class of discrepancies based on the optimal loss for a decision task -- two distributions are different if the optimal decision loss is higher on their mixture than on each indiv... | Accept (Oral) | Although by now there are several approaches for comparing probability distribution, the paper innovates by making their measure take into account the decision space and loss functions directly. The paper also frames its contribution within the literature at large. Reviewers were unanimous that the result is of major i... | train | [
"B72nG7umT4u",
"VoorCF82B0p",
"M3M_-WUphnf",
"HzWEhH716gX",
"yTr83cYb6Hx",
"QIZSXyvvm6t",
"ofIwg4cYIt"
] | [
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" I would like to thank the authors for their careful responses to my comments.\n\nI have read the responses and confirmed the changes which had been made to the paper. I am satisfied with the new additions to the paper based on my comments (f) and (g). I will keep my good score.",
" Thank you for your detailed r... | [
-1,
-1,
-1,
-1,
8,
8,
8
] | [
-1,
-1,
-1,
-1,
3,
3,
3
] | [
"HzWEhH716gX",
"QIZSXyvvm6t",
"ofIwg4cYIt",
"yTr83cYb6Hx",
"iclr_2022_KB5onONJIAU",
"iclr_2022_KB5onONJIAU",
"iclr_2022_KB5onONJIAU"
] |
iclr_2022_P07dq7iSAGr | Explaining Point Processes by Learning Interpretable Temporal Logic Rules | We propose a principled method to learn a set of human-readable logic rules to explain temporal point processes.
We assume that the generative mechanisms underlying the temporal point processes are governed by a set of first-order temporal logic rules, as a compact representation of domain knowledge. Our method formul... | Accept (Poster) | This paper has been independently evaluated by four expert reviewers. After discussion with authors, three of them set their recommendations at marginal acceptance, one at straight accept. Perhaps the key criticism involved limited rigor of theoretical justification for the proposed method, but it appears to be applica... | train | [
"nEetOEhncPK",
"cNMNiv7ta8",
"LW7p1izKzre",
"IqABgD8xPVs",
"tcPjsQTgIk",
"cBPVHZlutky",
"IE0Q-kZGYjc",
"xjYMEa2INQf",
"pJvFHAYdMnD",
"jPqa0p295L",
"N0TQVgV2QRg",
"BBW1N7AmNP2",
"_1N6UJ2S4U_",
"GT3Uv55WAL7"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"This paper introduces TELLER, a framework with novel algorithms based on the temporal point process model to discover interpretable temporal logic rules. To this end, the authors designed a “rule generation - evaluation” two-stage way to solve the problem efficiently. The proposed method is evaluated using one syn... | [
6,
-1,
-1,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
6,
6
] | [
4,
-1,
-1,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
2,
2
] | [
"iclr_2022_P07dq7iSAGr",
"IqABgD8xPVs",
"N0TQVgV2QRg",
"IE0Q-kZGYjc",
"pJvFHAYdMnD",
"iclr_2022_P07dq7iSAGr",
"xjYMEa2INQf",
"jPqa0p295L",
"cBPVHZlutky",
"nEetOEhncPK",
"GT3Uv55WAL7",
"_1N6UJ2S4U_",
"iclr_2022_P07dq7iSAGr",
"iclr_2022_P07dq7iSAGr"
] |
iclr_2022_FPCMqjI0jXN | Domino: Discovering Systematic Errors with Cross-Modal Embeddings | Machine learning models that achieve high overall accuracy often make systematic errors on important subsets (or slices) of data. Identifying underperforming slices is particularly challenging when working with high-dimensional inputs (e.g. images, audio), where important slices are often unlabeled. In order to address... | Accept (Oral) | Three experts reviewed this paper and all recommended acceptance. The reviewers liked that the work addressed a common problem in prior related work that it is hard to quantitatively evaluate slide discovery methods. Moreover, the proposed method achieves superior performance over prior arts. Based on the reviewers' fe... | train | [
"W_ojUruamIr",
"9ZzIa1QpNK1",
"76xln7LyC1A",
"I8AD_B35_GP",
"n6_Ebz3o0xW",
"v6a1Er6e2Mo",
"gn8nqolUReT",
"jumdS-eVtfS",
"OHn6vWX-30F",
"dX4Nuxsj_ND",
"rvibLfZLDQf",
"q8bkepp3Nne",
"TQcfVA-0aC"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author"
] | [
"The paper propose a framework for identifying on which subsets of data machine learning models make systematic errors. The problem is cast in two parts: (1) identify a model that can be identify a subset of data and predict degraded performance of the machine learning model for this subset and (2) ensure that the ... | [
6,
-1,
8,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
3,
-1,
2,
-1,
2,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
"iclr_2022_FPCMqjI0jXN",
"iclr_2022_FPCMqjI0jXN",
"iclr_2022_FPCMqjI0jXN",
"q8bkepp3Nne",
"iclr_2022_FPCMqjI0jXN",
"W_ojUruamIr",
"jumdS-eVtfS",
"W_ojUruamIr",
"dX4Nuxsj_ND",
"n6_Ebz3o0xW",
"iclr_2022_FPCMqjI0jXN",
"76xln7LyC1A",
"rvibLfZLDQf"
] |
iclr_2022_12RoR2o32T | Predictive Modeling in the Presence of Nuisance-Induced Spurious Correlations | In many prediction problems, spurious correlations are induced by a changing relationship between the label and a nuisance variable that is also correlated with the covariates. For example, in classifying animals in natural images, the background, which is a nuisance, can predict the type of animal. This nuisance-label... | Accept (Poster) | The paper studies how to build predictive models that are robust to nuisance-induced spurious correlations present in the data. It introduces nuisance-randomized distillation (NuRD), constructed by reweighting the observed data, to break the nuisance-label dependence and find the most informative representation to pre... | train | [
"hqfhFyqmPVF",
"YBAcirS3zs8",
"RnEw47tC_ED",
"khs7QkYENQ",
"0THK9yicEZ0"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"The paper studies the prediction problem under spurious association. The idea is to construct a \"nuisance randomized distribution\" based on the observed distribution. The nuisance randomized distribution is constructed by reweighting the observed data. Then the samples from the nuisance randomized distribution a... | [
5,
8,
8,
5,
6
] | [
4,
3,
3,
5,
3
] | [
"iclr_2022_12RoR2o32T",
"iclr_2022_12RoR2o32T",
"iclr_2022_12RoR2o32T",
"iclr_2022_12RoR2o32T",
"iclr_2022_12RoR2o32T"
] |
iclr_2022_fVu3o-YUGQK | Efficient Self-supervised Vision Transformers for Representation Learning | This paper investigates two techniques for developing efficient self-supervised vision transformers (EsViT) for visual representation learning. First, we show through a comprehensive empirical study that multi-stage architectures with sparse self-attentions can significantly reduce modeling complexity but with a cost o... | Accept (Poster) | This paper proposes two techniques for improving self-supervised learning with a vision transformer. The first improvement is using a multi-stage ViT, which is very similar to Swin transformer and authors recognized this is not a major contribution. The authors further found that using a multi-stage ViT does not produc... | train | [
"lqXz2Br67h7",
"qSjNSe8PIee",
"RZe2W7Nxq0N",
"enKs1I2TpWB",
"24mE_8qHpsA",
"VQpZ2EJmNlz",
"kIgxZIT55DB",
"GPtlPjHBbN1"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"The contribution of the paper comprises:\n* the observation that the Multi-Stage Vision Transformer (MSVT), as opposed to the \"monolithic\" Vision Transformer (VT), does not produce discriminative patch representation, and\n* a loss function for self-supervised pre-training of MSVTs, that encourages discriminativ... | [
8,
-1,
-1,
-1,
-1,
-1,
8,
6
] | [
4,
-1,
-1,
-1,
-1,
-1,
4,
5
] | [
"iclr_2022_fVu3o-YUGQK",
"24mE_8qHpsA",
"kIgxZIT55DB",
"GPtlPjHBbN1",
"lqXz2Br67h7",
"iclr_2022_fVu3o-YUGQK",
"iclr_2022_fVu3o-YUGQK",
"iclr_2022_fVu3o-YUGQK"
] |
iclr_2022_mmUA7_O9mjY | Contact Points Discovery for Soft-Body Manipulations with Differentiable Physics | Differentiable physics has recently been shown as a powerful tool for solving soft-body manipulation tasks. However, the differentiable physics solver often gets stuck when the initial contact points of the end effectors are sub-optimal or when performing multi-stage tasks that require contact point switching, which of... | Accept (Spotlight) | The reviewers are unanimous that this is a strong submission that deserves to be accepted. | train | [
"DyyK7tzw_e",
"bV3GYS9wSFH",
"WH47yCsXGD",
"zeguB29SlQr",
"ym8ee3gKNSK",
"DRNtWTodRR",
"OADJNtspscy",
"ME2z_Ohf_bq",
"5uE_DSQu-nA",
"eV58XVQxgR",
"SmY3mtH447k",
"zwJn5bmEq7B"
] | [
"author",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
" Thank you for your constructive comments and suggestions. We will follow your recommendation and clarify the detail about physical awareness in the final version of the paper. Thank you again for your time.\n\nBest, Authors",
"The paper proposed an algorithm to discover appropriate contact points for deformable... | [
-1,
8,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
8
] | [
-1,
3,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
3
] | [
"WH47yCsXGD",
"iclr_2022_mmUA7_O9mjY",
"ME2z_Ohf_bq",
"DRNtWTodRR",
"iclr_2022_mmUA7_O9mjY",
"eV58XVQxgR",
"iclr_2022_mmUA7_O9mjY",
"bV3GYS9wSFH",
"zwJn5bmEq7B",
"SmY3mtH447k",
"ym8ee3gKNSK",
"iclr_2022_mmUA7_O9mjY"
] |
iclr_2022_tUa4REjGjTf | On the Certified Robustness for Ensemble Models and Beyond | Recent studies show that deep neural networks (DNN) are vulnerable to adversarial examples, which aim to mislead DNNs by adding perturbations with small magnitude. To defend against such attacks, both empirical and theoretical defense approaches have been extensively studied for a single ML model. In this work, we aim ... | Accept (Poster) | The paper proposes Diversity-Regularized Training (DRT), a new training method for an ensemble classifier to improve its certified robustness when randomized smoothing is applied. Specifically, it trains a set of base classifiers to diversify their input gradients while maximizing the confidence margin of each. The met... | train | [
"sVm2GJvAZs-",
"pf2RQNzyGwy",
"bJV6ydlRuly",
"mpa_PM6tfDq",
"GwEYOMA3bWx",
"IPGEDQgWnTP",
"iFb_mxGyV-S",
"aATOBoOv-m9",
"KeSkt7GdbOX",
"QXOlGsGriK",
"ZXot_4tgVP2",
"6lCTYEBX5a9",
"1CY6vvl4I0h",
"ptQ79d_dAJo",
"MGQvzMFpuxR",
"Q4ToBZ7eWi",
"k2tQrGhd7a",
"DidFq7BKfBu",
"vzW5N9vhK9u"... | [
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Thanks for your feedback and adding this further clarification to the main paper.",
" > “The paper should include a discussion about the computational costs of the proposed method - e.g., the increase in training time of DRT compared to the Gaussian training (or standard ensemble). Also, I would like to see the... | [
-1,
-1,
-1,
8,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
8,
6
] | [
-1,
-1,
-1,
3,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
4
] | [
"QXOlGsGriK",
"vzW5N9vhK9u",
"GwEYOMA3bWx",
"iclr_2022_tUa4REjGjTf",
"1CY6vvl4I0h",
"aATOBoOv-m9",
"iclr_2022_tUa4REjGjTf",
"ZXot_4tgVP2",
"iclr_2022_tUa4REjGjTf",
"DidFq7BKfBu",
"iFb_mxGyV-S",
"mpa_PM6tfDq",
"mpa_PM6tfDq",
"vzW5N9vhK9u",
"vzW5N9vhK9u",
"k2tQrGhd7a",
"iclr_2022_tUa4R... |
iclr_2022_TYw3-OlrRm- | Network Augmentation for Tiny Deep Learning | We introduce Network Augmentation (NetAug), a new training method for improving the performance of tiny neural networks. Existing regularization techniques (e.g., data augmentation, dropout) have shown much success on large neural networks by adding noise to overcome over-fitting. However, we found these techniques hur... | Accept (Poster) | This paper studies the problem of training tiny networks, by proposing a new training method called Network Augmentation (NetAug). The main challenge for training tiny networks lies in underfitting, which data augmentation and dropout etc. regularizations may suffer from for tiny networks. To overcome this hurdle, the ... | val | [
"Aj8B9xoAJuf",
"6gtOKytZey",
"U03w_7BMlfE",
"o9YqBUvqBE2",
"3P4YmKrm47",
"_SHnmIPlImC",
"2NmbxXL4Snt",
"hoEQqWiAf74",
"qZZ8GFIXd_",
"Q9jfgAzsJ3g",
"xp-cca5pOJp",
"bZClZ_ge6Cg",
"2UifWvizbnK",
"4aFEG7pxzWb",
"5yYenFPZ4OJ",
"gaFxHaFg_94",
"kROI_O3lgp",
"zN8KQeB9Gp",
"HZge85-CL0",
... | [
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
... | [
" Dear Reviewer XSG2,\n\nIn addition to previous clarifications, we want to emphasize that\n- We already provided experimental results showing that our method is also effective when combined with network pruning (Table 4). It shows that selecting an equivalent sub-network from a pre-trained larger network and train... | [
-1,
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
8
] | [
-1,
-1,
-1,
-1,
-1,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4
] | [
"DopuMlxYb5g",
"DopuMlxYb5g",
"2NmbxXL4Snt",
"3P4YmKrm47",
"qZZ8GFIXd_",
"iclr_2022_TYw3-OlrRm-",
"bZClZ_ge6Cg",
"5yYenFPZ4OJ",
"2UifWvizbnK",
"2UifWvizbnK",
"2UifWvizbnK",
"2UifWvizbnK",
"I7KUHDzrC4w",
"iclr_2022_TYw3-OlrRm-",
"2wWoMy7SY6l",
"HZge85-CL0",
"DopuMlxYb5g",
"iclr_2022... |
iclr_2022_-w2oomO6qgc | GeneDisco: A Benchmark for Experimental Design in Drug Discovery | In vitro cellular experimentation with genetic interventions, using for example CRISPR technologies, is an essential step in early-stage drug discovery and target validation that serves to assess initial hypotheses about causal associations between biological mechanisms and disease pathologies. With billions of potenti... | Accept (Poster) | This paper introduces a benchmark for experimental design algorithms
for an important cellular biological question, causal discovery of
effective genetic knock-out interventions. It uses existing datasets.
The paper was discussed by the reviewers after the authors correctly
pointed out that methodological machine lear... | train | [
"z1sWYNn1_Qr",
"ZQ6qNmJW8u9",
"BiobDYSqFd",
"fXp1RiF_Q45",
"qPoufA45N-C",
"im2KwlsmS_4",
"79tQpHphePc"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"The authors introduce a benchmark suite for evaluating active learning algorithms for experimental design in drug discovery. Specifically, they introduce various genome-wide CRISPR screens within immunology that evaluate the causal effect of intervening on a large number of genes in model systems in order to ident... | [
6,
6,
-1,
-1,
-1,
-1,
6
] | [
3,
4,
-1,
-1,
-1,
-1,
4
] | [
"iclr_2022_-w2oomO6qgc",
"iclr_2022_-w2oomO6qgc",
"ZQ6qNmJW8u9",
"79tQpHphePc",
"z1sWYNn1_Qr",
"79tQpHphePc",
"iclr_2022_-w2oomO6qgc"
] |
iclr_2022_aBsCjcPu_tE | SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations | Guided image synthesis enables everyday users to create and edit photo-realistic images with minimum effort. The key challenge is balancing faithfulness to the user inputs (e.g., hand-drawn colored strokes) and realism of the synthesized images. Existing GAN-based methods attempt to achieve such balance using either co... | Accept (Poster) | Thank you for your submission to ICLR.
This paper presents a technique for image synthesis based on stochastic differential equations and a diffusion model. This looks to be a very nice idea with good results. After discussion, the reviewers converged and all agreed that the paper is ready for publication---the most... | val | [
"oAQ74LszGX",
"MEXe5sSz338",
"QSlk84uhMj",
"f__hN2sAwdW",
"3IU9bjDd0N9",
"k48T9gRoz_d",
"zSQRGwMba8y",
"GJLCf3_aX2z"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"The paper proposesed an image editing and synthesis system (SDEdit) with SDE. By injecting the guided image into the reverse process with appropriate $t_0$, the results can have a trade-off between realism and faithfulness. Strengths:\nThe whole idea is interesting, effective and simple. The SDEdit perturbs the in... | [
8,
-1,
6,
-1,
-1,
-1,
-1,
6
] | [
4,
-1,
2,
-1,
-1,
-1,
-1,
3
] | [
"iclr_2022_aBsCjcPu_tE",
"zSQRGwMba8y",
"iclr_2022_aBsCjcPu_tE",
"iclr_2022_aBsCjcPu_tE",
"GJLCf3_aX2z",
"QSlk84uhMj",
"oAQ74LszGX",
"iclr_2022_aBsCjcPu_tE"
] |
iclr_2022_dg79moSRqIo | One After Another: Learning Incremental Skills for a Changing World | Reward-free, unsupervised discovery of skills is an attractive alternative to the bottleneck of hand-designing rewards in environments where task supervision is scarce or expensive. However, current skill pre-training methods, like many RL techniques, make a fundamental assumption -- stationary environments during trai... | Accept (Poster) | This paper is about an unsupervised method to learn new skills in non-stationary environments by maximizing an intrinsic reward function. Experimental evaluations on OpenAI gym environments show that the proposed approach improves the diversity of the learned skills and is able to adapt to continuously changing environ... | train | [
"gtZz9AtW2b",
"ceWf5hSfvgp",
"zJy01EK74jG",
"oI_A9uONNrK",
"8eeJho4UitT",
"U3KaQlWD2YU",
"dKA600ZNyXz",
"2ibuGlp8s2",
"xtqBr88mINu",
"-nWF9HgcgNw",
"dJtwtwbzthB",
"HN4LUIsZ9m5",
"oUQPvieW6z",
"fO6AV9X6YBn",
"83dtreADhIz",
"xuQ-3bpFgqm",
"3gZiy3595pU",
"IF8VVob0HE4",
"FwSsUXSWjI",... | [
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"... | [
" Absolutely. I only mentioned Atari because it has been used in related works. I am not aware of any paper evaluating skill discovery methods on ProcGen, but I totally agree that it would be a great evaluation benchmark in which DiSK should work well as long as a good solution is found for overcoming the represent... | [
-1,
-1,
-1,
-1,
-1,
-1,
6,
6,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6
] | [
-1,
-1,
-1,
-1,
-1,
-1,
4,
3,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5
] | [
"zJy01EK74jG",
"U3KaQlWD2YU",
"83dtreADhIz",
"xtqBr88mINu",
"dJtwtwbzthB",
"LJjD57j1sMu",
"iclr_2022_dg79moSRqIo",
"iclr_2022_dg79moSRqIo",
"IF8VVob0HE4",
"iclr_2022_dg79moSRqIo",
"oUQPvieW6z",
"xuQ-3bpFgqm",
"fO6AV9X6YBn",
"ena1nbu9nHj",
"3gZiy3595pU",
"iclr_2022_dg79moSRqIo",
"hijV... |
iclr_2022_49A1Y6tRhaq | Linking Emergent and Natural Languages via Corpus Transfer | The study of language emergence aims to understand how human languages are shaped by perceptual grounding and communicative intent. Computational approaches to emergent communication (EC) predominantly consider referential games in limited domains and analyze the learned protocol within the game framework. As a result,... | Accept (Spotlight) | This paper explores ways in which *emergent communication* (EC) methods from representation learning can be evaluated extrinsically, by hooking them into downstream NLP tasks. Reviewers agree that the paper is thorough, and finds encouraging results.
This paper is borderline, and difficult to evaluate, even after very... | train | [
"e-0C_dmhi6X",
"Yb-tFli0PpA",
"4LfJsLMJpzq",
"dJ7VKQ-itY",
"gjdvYFm8yP",
"KpB7KerJ1m",
"1D08Cqgf0D",
"MfXAAtYOU0v",
"qfyxjMzOcCQ",
"g6bKwaksJP7",
"_C7JxZV71Qe",
"z80Jhp47nAy",
"Q28wHtPfv7z",
"5P-sxLOeaLG",
"-egT9GK4x1",
"rQSvZExWBq",
"gTb6-NY6YbO",
"qOWePq1jaTu",
"K5OQL5Dge2m",
... | [
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",... | [
" Dear Reviewer wvqW,\n\nHappy Thanksgiving! Hope you enjoyed a wonderful holiday.\n\nAs the deadline for discussion is approaching, we want to make sure if our second response addresses your remaining concerns. We would love to further engage in the limited remaining time if needed, to the best of our ability.\n\n... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
8,
6,
8
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
3,
4
] | [
"Yb-tFli0PpA",
"4LfJsLMJpzq",
"Rl8WUe07Ms_",
"qfyxjMzOcCQ",
"K5OQL5Dge2m",
"iclr_2022_49A1Y6tRhaq",
"3wZWFxBX_pz",
"bM1QjyIArTa",
"OBWsOt8Lyc5",
"Rl8WUe07Ms_",
"iclr_2022_49A1Y6tRhaq",
"3wZWFxBX_pz",
"bM1QjyIArTa",
"-egT9GK4x1",
"Rl8WUe07Ms_",
"iclr_2022_49A1Y6tRhaq",
"qOWePq1jaTu",
... |
iclr_2022_5XmLzdslFNN | Modular Lifelong Reinforcement Learning via Neural Composition | Humans commonly solve complex problems by decomposing them into easier subproblems and then combining the subproblem solutions. This type of compositional reasoning permits reuse of the subproblem solutions when tackling future tasks that share part of the underlying compositional structure. In a continual or lifelong ... | Accept (Poster) | The paper presents a method for compositional task learning in the continual RL setting, by composing and reconfiguring neural modules. The method is evaluated on mini-grids and simulated robot manipulation tasks.
The reviewers agree, and I concur, that the paper proposes an interesting solution to a difficult and imp... | train | [
"sf2yNlyqTz",
"Z-wAKc7QDnK",
"ex7C6D9G8Cc",
"Xhu2iMzL-X",
"iSxPXIs3c9q",
"fhGiO9sDKz",
"wJoBABoaMdJ",
"gmqo_maSdC7",
"p3Hjnwdyldu",
"b7rvgSIP5X3",
"mp1Fh65D9YO",
"SP9xcmv7TtJ",
"u_E2HER3NgQ",
"ACcZdfOi5Gs",
"k_v8b79UxNt",
"c3TvAoQVzky",
"hiHxIwLZRqM",
"MXBz1g0eBC0",
"pFF8rtQKia",... | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"au... | [
"The paper introduces an algorithm for lifelong reinforcement learning using functional neural composition. The algorithm first maps a new problem onto a composition of previously acquired modules; then, the agent trains/finetunes the selected module combination on the new task; and finally, the agent incorporates ... | [
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
6
] | [
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,
3
] | [
"iclr_2022_5XmLzdslFNN",
"SP9xcmv7TtJ",
"b7rvgSIP5X3",
"p3Hjnwdyldu",
"gmqo_maSdC7",
"wJoBABoaMdJ",
"MEePz71yIWN",
"HlhT5l_npS",
"u_E2HER3NgQ",
"ACcZdfOi5Gs",
"iclr_2022_5XmLzdslFNN",
"c3TvAoQVzky",
"hiHxIwLZRqM",
"MXBz1g0eBC0",
"pFF8rtQKia",
"urQKNg0Y_BB",
"MXBz1g0eBC0",
"pFF8rtQK... |
iclr_2022_LedObtLmCjS | Bi-linear Value Networks for Multi-goal Reinforcement Learning | Universal value functions are a core component of off-policy multi-goal reinforcement learning.
The de-facto paradigm is to approximate Q(s, a, g) using monolithic neural networks which lack inductive biases to produce complex interactions between the state s and the goal g. In this work, we propose a bilinear decompo... | Accept (Poster) | This paper proposes a new bilinear decomposition for universal value functions. The bilinear network has one component dependent on state and goal and another component that depends on state and action. The experiments with the DDPG algorithm in robot simulations show that the proposed architecture improves performan... | train | [
"T5_aUkc4jcR",
"E8Hy9GWs4ZM",
"uixuAMwGa3z",
"b1oQIwG2DkF",
"XFM2MGFac5",
"42VGUZdUkIB",
"1R27M-loHq",
"mvFHWBSakZg",
"BYXD5VvtDfG",
"mFv3Wipnd34",
"p3SJM6Viy-I",
"FckE_eXkXAV",
"3hu4nFVJKzL",
"-G9iztTziG_",
"IUgH8NmUp16"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"The paper proposes a modification to goal-oriented universal value functions that split the neural network into two parts. One network accepts states and actions and outputs a vector representation, while the other outputs a vector given a state and goal. Part of the contribution of the paper is the interpretation... | [
6,
6,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6
] | [
3,
4,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3
] | [
"iclr_2022_LedObtLmCjS",
"iclr_2022_LedObtLmCjS",
"mFv3Wipnd34",
"iclr_2022_LedObtLmCjS",
"b1oQIwG2DkF",
"b1oQIwG2DkF",
"b1oQIwG2DkF",
"T5_aUkc4jcR",
"T5_aUkc4jcR",
"E8Hy9GWs4ZM",
"E8Hy9GWs4ZM",
"E8Hy9GWs4ZM",
"IUgH8NmUp16",
"IUgH8NmUp16",
"iclr_2022_LedObtLmCjS"
] |
iclr_2022_HFPTzdwN39 | Measuring the Interpretability of Unsupervised Representations via Quantized Reversed Probing | Self-supervised visual representation learning has recently attracted significant research interest. While a common way to evaluate self-supervised representations is through transfer to various downstream tasks, we instead investigate the problem of measuring their interpretability, i.e. understanding the semantics en... | Accept (Poster) | This paper proposes a method for inspecting and interpreting the visual representations learned by self-supervised methods.
The method is conceptually simple and intuititive, the authors assume that concept labels for the images are available, and then go on to learn a mapping between the learned image vectors and the... | train | [
"j_ElWaOaApl",
"bBJD-DhSr-K",
"SbKJpUMgpee",
"x5StL3G9Mq8",
"89OY10YLOma",
"IOp8LjESf4b",
"Tgzww0rgA6_",
"rn2OWD4Lqqz",
"NDX4EHO8Pi",
"fSa84b8CZo3",
"R-r7IArc27",
"2ggVaAPswh2",
"WuEroT8Phb6",
"W0Oji1CuO-8",
"S1qsljRBQEl",
"v2_Um2cjW-",
"xBfxqhLgrDX",
"d5oGu_Z4QJd"
] | [
"author",
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Dear reviewer phRe, \n\nThank you again for your valuable feedback on our work. Based on your comments, we have added a discussion on topic models in the appendix, categorized methods under a taxonomy, clarified our findings, and included additional experiments with varying K to verify that this choice does not a... | [
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
5,
3
] | [
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3,
4
] | [
"2ggVaAPswh2",
"R-r7IArc27",
"iclr_2022_HFPTzdwN39",
"89OY10YLOma",
"NDX4EHO8Pi",
"W0Oji1CuO-8",
"iclr_2022_HFPTzdwN39",
"SbKJpUMgpee",
"SbKJpUMgpee",
"d5oGu_Z4QJd",
"d5oGu_Z4QJd",
"xBfxqhLgrDX",
"xBfxqhLgrDX",
"v2_Um2cjW-",
"iclr_2022_HFPTzdwN39",
"iclr_2022_HFPTzdwN39",
"iclr_2022_... |
iclr_2022_xnYACQquaGV | Neural Contextual Bandits with Deep Representation and Shallow Exploration | We study neural contextual bandits, a general class of contextual bandits, where each context-action pair is associated with a raw feature vector, but the specific reward generating function is unknown. We propose a novel learning algorithm that transforms the raw feature vector using the last hidden layer of a deep Re... | Accept (Poster) | This paper tackles the neural contextual bandit problem, for which existing approaches consists rely on bandit algorithms based on deep neural networks to learn reward functions. In these existing strategies, exploration takes place over the entire network parameter space, which can be inefficient for the large-size ne... | train | [
"2dNI7wa2wRp",
"EN5ggHMHhI7",
"uobTCkYlwpJ",
"5XLv3zX1PgT",
"bRXU4vyuJpC",
"hZ79BvR-W_Z",
"mdTQLR6SlP",
"l0wt6l_zVRv",
"1yY_x7Vx_1p",
"xizT0S4ggtq",
"BE8_CrnKDkL",
"ApsHkPlQZyI",
"Wrk9N8u4A8",
"Qhf_JuO5vDy",
"fYZNpvDwV7v",
"bXhJzpZxrN"
] | [
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"author"
] | [
"The paper presents a new neural-bandit algorithm with shallow exploration and provides a regret bound for the proposed method. The existing approaches have introduced deep neural networks based bandit algorithms to learn reward functions, in which exploration takes place over the entire network parameter space, wh... | [
6,
-1,
3,
-1,
8,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
4,
-1,
4,
-1,
4,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
"iclr_2022_xnYACQquaGV",
"5XLv3zX1PgT",
"iclr_2022_xnYACQquaGV",
"BE8_CrnKDkL",
"iclr_2022_xnYACQquaGV",
"iclr_2022_xnYACQquaGV",
"l0wt6l_zVRv",
"Wrk9N8u4A8",
"2dNI7wa2wRp",
"ApsHkPlQZyI",
"ApsHkPlQZyI",
"bXhJzpZxrN",
"bRXU4vyuJpC",
"hZ79BvR-W_Z",
"2dNI7wa2wRp",
"uobTCkYlwpJ"
] |
iclr_2022_L7wzpQttNO | BDDM: Bilateral Denoising Diffusion Models for Fast and High-Quality Speech Synthesis | Diffusion probabilistic models (DPMs) and their extensions have emerged as competitive generative models yet confront challenges of efficient sampling. We propose a new bilateral denoising diffusion model (BDDM) that parameterizes both the forward and reverse processes with a schedule network and a score network, which... | Accept (Poster) | This work suggests an extension of diffusion-based generative models, where both the forward and reverse process have learnable parameters (rather than just the reverse process). This is then applied to speech synthesis, with high-fidelity audio generated in very few sampling steps compared to what is typical for this ... | train | [
"vEpZTcJ9oUt",
"rmY0PdKVmY9",
"WYW9F50ihDs",
"BTAX43BtkGD",
"NS4d481Ef1u",
"fH60CGeTiB5",
"ot6orC2CRS",
"w3M950oS7W7",
"RPg6ldibti7",
"U1Rpnm0wK6F"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
" Thank you! I am keeping my score (6), leaning towards acceptance. In case the paper is accepted, please include A/B test results to the main text, this is quite informative for the reader.",
" Thank you for your answer improved version of the paper. After this, I think the paper deserves a chance and therefore ... | [
-1,
-1,
8,
6,
-1,
-1,
-1,
-1,
-1,
6
] | [
-1,
-1,
2,
4,
-1,
-1,
-1,
-1,
-1,
3
] | [
"RPg6ldibti7",
"fH60CGeTiB5",
"iclr_2022_L7wzpQttNO",
"iclr_2022_L7wzpQttNO",
"ot6orC2CRS",
"WYW9F50ihDs",
"BTAX43BtkGD",
"BTAX43BtkGD",
"U1Rpnm0wK6F",
"iclr_2022_L7wzpQttNO"
] |
iclr_2022_boJy41J-tnQ | Subspace Regularizers for Few-Shot Class Incremental Learning | Few-shot class incremental learning---the problem of updating a trained classifier to discriminate among an expanded set of classes with limited labeled data---is a key challenge for machine learning systems deployed in non-stationary environments. Existing approaches to the problem rely on complex model architectures ... | Accept (Poster) | The paper proposes a subspace regularization technique that encourages the new class weight vector to be in the subspace spanned by those of the base classes for few-shot class incremental learning. Even though similar techniques exist in few-shot learning literature, reviewers appreciate the simplicity of the method a... | train | [
"EFW_AmYr8Hd",
"3bY2eSgdkkb",
"3vadlXWevIN",
"FXuL-6PT4Rl",
"ceMtY4fkQ9O",
"vpyuOvPDBda",
"ZIM_FLKiqCR",
"XqxWpHLQkdnN",
"whPjXO-ltW_",
"dGe5pZY260a",
"ScukpCGYSr1",
"9r4c5-BgF5T"
] | [
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"The paper proposes subspace regularization technique for incremental few-shot learning. The high-level idea is to find the basis vectors of the subspace spanned by the base classes, and then project the new classes into the subspace. The regularizer encourages the new class weights to be similar to the projected v... | [
6,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
5,
8
] | [
4,
-1,
-1,
-1,
5,
-1,
-1,
-1,
-1,
-1,
4,
4
] | [
"iclr_2022_boJy41J-tnQ",
"EFW_AmYr8Hd",
"vpyuOvPDBda",
"ScukpCGYSr1",
"iclr_2022_boJy41J-tnQ",
"ZIM_FLKiqCR",
"ceMtY4fkQ9O",
"EFW_AmYr8Hd",
"9r4c5-BgF5T",
"ScukpCGYSr1",
"iclr_2022_boJy41J-tnQ",
"iclr_2022_boJy41J-tnQ"
] |
iclr_2022_SsPCtEY6yCl | On the Uncomputability of Partition Functions in Energy-Based Sequence Models | In this paper, we argue that energy-based sequence models backed by expressive parametric families can result in uncomputable and inapproximable partition functions. Among other things, this makes model selection--and therefore learning model parameters--not only difficult, but generally _undecidable_. The reason is th... | Accept (Spotlight) | This is a deep theoretical paper with results that I consider very interesting. I have *not* had time to check them myself, but I have background in these theoretical matters and the results seem reasonable to me - the hardness of even checking the quality of a solution is well known for partition functions (as well as... | train | [
"XGx4_sQPbeo",
"gUT6stLNUh",
"ldT4l4S_z2",
"Ce2_S28GE6i",
"yqOasAFsTw",
"d--wTzfbHlf",
"WcAvWY1uEs",
"7qqlsuobed",
"XeEuyIZHD3S",
"708GxwWppV",
"0m2vSGF4L9",
"cSf0OEn3lhV",
"zuKcGCA5nfR"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" I have read the response of the authors. The authors have answered all my questions in detail and clarified certain aspects that were not that obvious to me originally.\nAfter also reading the other reviews, I will stick to my decision that the paper makes a solid contribution and keep my score.",
" Apologies f... | [
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
8,
6,
8
] | [
-1,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3,
3
] | [
"yqOasAFsTw",
"XeEuyIZHD3S",
"708GxwWppV",
"iclr_2022_SsPCtEY6yCl",
"zuKcGCA5nfR",
"WcAvWY1uEs",
"0m2vSGF4L9",
"0m2vSGF4L9",
"cSf0OEn3lhV",
"Ce2_S28GE6i",
"iclr_2022_SsPCtEY6yCl",
"iclr_2022_SsPCtEY6yCl",
"iclr_2022_SsPCtEY6yCl"
] |
iclr_2022_Ow1C7s3UcY | Vitruvion: A Generative Model of Parametric CAD Sketches | Parametric computer-aided design (CAD) tools are the predominant way that engineers specify physical structures, from bicycle pedals to airplanes to printed circuit boards. The key characteristic of parametric CAD is that design intent is encoded not only via geometric primitives, but also by parameterized constraints ... | Accept (Poster) | The paper describes an approach for automatically generating CAD sketches, including both the primitives that describe the drawing, as well as the constraints that describe relationships between the primitives that need to be maintained even if the primitives are changed. This is an important problem that is starting t... | train | [
"SKYBlfMGed",
"28OqAs_aOF",
"y4kIze_fvz7",
"AHn1jWd8VFd",
"D_qTuNlPKnj",
"eMo183Rh_WG",
"Afa3R5MgJqe",
"dTLC2Cqb264",
"OuPI4q1hice",
"fRpK5f9u75S",
"D-LGDV4VUt",
"aur5w3aIY5C",
"mdIwJHbSxs",
"HX5a30JLKyZ"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"public",
"official_reviewer",
"official_reviewer"
] | [
" Thank you for the response. While I do still have some of the concerns in my initial review with respect to high-level takeaways, as the authors point out, the existing work with substantial overlap is concurrent as per ICLR guidelines, and so I raise my score.",
"This paper proposes a new deep generative model... | [
-1,
6,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
8
] | [
-1,
4,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4
] | [
"D-LGDV4VUt",
"iclr_2022_Ow1C7s3UcY",
"dTLC2Cqb264",
"Afa3R5MgJqe",
"iclr_2022_Ow1C7s3UcY",
"OuPI4q1hice",
"D_qTuNlPKnj",
"HX5a30JLKyZ",
"mdIwJHbSxs",
"iclr_2022_Ow1C7s3UcY",
"28OqAs_aOF",
"HX5a30JLKyZ",
"iclr_2022_Ow1C7s3UcY",
"iclr_2022_Ow1C7s3UcY"
] |
iclr_2022_FZoZ7a31GCW | Ancestral protein sequence reconstruction using a tree-structured Ornstein-Uhlenbeck variational autoencoder | We introduce a deep generative model for representation learning of biological sequences that, unlike existing models, explicitly represents the evolutionary process. The model makes use of a tree-structured Ornstein-Uhlenbeck process, obtained from a given phylogenetic tree, as an informative prior for a variational a... | Accept (Poster) | The paper describes a VAE model for individual protein families that uses phylogenetic trees through an Ornstein-Uhlenbeck process on latent space. They also use a sequence likelihood which does not factorize over positions. The paper claims these two advances represent a more expressive and efficient model of protein ... | train | [
"5z7gDya2Ix-",
"ZHSiCygcEKi",
"J4ib26MnnGJ",
"F_UCSmh9Bk",
"1SN6iVlmy-s",
"ytwnA-XMbu-",
"5jiVmrHjMIO",
"l7s0Ii9R3q",
"y12VJ2tcSRF"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"The authors introduce a VAE for modeling individual protein families that incorporates phylogenetic trees through an OU process on latent space. They also use a sequence likelihood which does not factorize over positions. The authors claim these two advances represent a more expressive and efficient model of prote... | [
8,
-1,
-1,
-1,
-1,
-1,
-1,
5,
8
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
4,
5
] | [
"iclr_2022_FZoZ7a31GCW",
"ytwnA-XMbu-",
"l7s0Ii9R3q",
"5z7gDya2Ix-",
"y12VJ2tcSRF",
"5z7gDya2Ix-",
"l7s0Ii9R3q",
"iclr_2022_FZoZ7a31GCW",
"iclr_2022_FZoZ7a31GCW"
] |
iclr_2022_25kzAhUB1lz | Direct then Diffuse: Incremental Unsupervised Skill Discovery for State Covering and Goal Reaching | Learning meaningful behaviors in the absence of reward is a difficult problem in reinforcement learning. A desirable and challenging unsupervised objective is to learn a set of diverse skills that provide a thorough coverage of the state space while being directed, i.e., reliably reaching distinct regions of the enviro... | Accept (Poster) | The authors propose UPSIDE, a method for improving state coverage in unsupervised skill learning. "Direct-and-diffuse" policies concatenate phases of directed behaviour followed by dithering to improve neighbourhood coverage in the region of state space visited, trained with a discriminability objective to ensure diver... | train | [
"mkjcq8kkznB",
"bxJusjPM-FV",
"4npc_RgMV2",
"OTyMAcPj10t",
"wwEos5Si69h",
"lNUj-Iw0Kvc",
"Xg-mmjSlxTF",
"c5PCQXvZA7o",
"4EEEm991UcU",
"0Abgzzrv5Jm",
"jM0JSeWetz0",
"GlGp2Qq7otY",
"TRYafnqef1F",
"rUd6E5USpLA"
] | [
"author",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
" Indeed, as articulated in Section 3.4, the discussed proof holds in the \"flat case\". This gives us a principled way to optimize the tree structure *locally* level by level, whereas the tree is expanded incrementally, which enables a tractable search in an otherwise combinatorial space. \n\nWe point out that thi... | [
-1,
-1,
-1,
-1,
6,
-1,
6,
-1,
-1,
-1,
-1,
-1,
8,
8
] | [
-1,
-1,
-1,
-1,
4,
-1,
4,
-1,
-1,
-1,
-1,
-1,
3,
4
] | [
"bxJusjPM-FV",
"4npc_RgMV2",
"OTyMAcPj10t",
"GlGp2Qq7otY",
"iclr_2022_25kzAhUB1lz",
"jM0JSeWetz0",
"iclr_2022_25kzAhUB1lz",
"4EEEm991UcU",
"Xg-mmjSlxTF",
"TRYafnqef1F",
"rUd6E5USpLA",
"wwEos5Si69h",
"iclr_2022_25kzAhUB1lz",
"iclr_2022_25kzAhUB1lz"
] |
iclr_2022_jeLW-Fh9bV | Skill-based Meta-Reinforcement Learning | While deep reinforcement learning methods have shown impressive results in robot learning, their sample inefficiency makes the learning of complex, long-horizon behaviors with real robot systems infeasible. To mitigate this issue, meta-reinforcement learning methods aim to enable fast learning on novel tasks by learnin... | Accept (Poster) | The main contribution of this paper lies in the novel setting that is being considered: offline data without rewards is combined with meta-training tasks to quickly adapt to new long-horizon tasks at meta-test time. Within this setting, it is shown that the combination of SPiRL and PEARL outperforms the individual algo... | train | [
"50tJvb3Dmpq",
"EB4-C22D7TS",
"T2bmFVEaC7",
"d-97GUNrwZS",
"tjJCIHBQczi",
"g1QSYggsfh",
"4QH0V4E-4XL",
"A9Ze6T_TNRl",
"ptw5fR80gns",
"_8_i7NFeA74",
"IMPJrpp3z4",
"HL22MS0iVRV",
"hQg-vHUjkBO",
"R3NajCuoiRG",
"ZKKZdqBwbYx",
"oWkAyHpudTe",
"16TI5DXHyF_",
"uvigiLdRUIM",
"XlPpJS874KH"... | [
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
... | [
" We thank the reviewer for the constructive feedback. We will compare our method and SPiRL by sampling the tasks spreading across the entire maze and include the result in the revised paper.",
" Great, thanks for the clarification. Overall, I still find the empirical contribution to be compelling, even if the no... | [
-1,
-1,
8,
-1,
6,
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6
] | [
-1,
-1,
4,
-1,
4,
-1,
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4
] | [
"d-97GUNrwZS",
"g1QSYggsfh",
"iclr_2022_jeLW-Fh9bV",
"ptw5fR80gns",
"iclr_2022_jeLW-Fh9bV",
"4QH0V4E-4XL",
"R3NajCuoiRG",
"IMPJrpp3z4",
"HL22MS0iVRV",
"iclr_2022_jeLW-Fh9bV",
"ZKKZdqBwbYx",
"uvigiLdRUIM",
"iclr_2022_jeLW-Fh9bV",
"bzmxh_3R17K",
"_8_i7NFeA74",
"_8_i7NFeA74",
"T2bmFVEaC... |
iclr_2022_tBtoZYKd9n | Evaluation Metrics for Graph Generative Models: Problems, Pitfalls, and Practical Solutions | Graph generative models are a highly active branch of machine learning. Given the steady development of new models of ever-increasing complexity, it is necessary to provide a principled way to evaluate and compare them. In this paper, we enumerate the desirable criteria for such a comparison metric and provide an overv... | Accept (Spotlight) | This paper provides an overview of evaluating graph generative models (GGMs). It systematically evaluates one of the more popular metrics, maximum mean discrepancy (MMD). It highlights some challenges and pitfalls for practitioners and suggests some ways to mitigate them. The reviewers found the paper practically relev... | train | [
"Lr23PPj02_R",
"2yN1tHVUPdt",
"t6RHtDm2FXm",
"KJn6LiFSPjN",
"gt-bu4Hc58",
"hSxoMxIZ5Ou",
"jet2DuPEaqj",
"-XurEw2KHH",
"AzpbhOJAld9",
"6QMQXO6hYs",
"4l4ruJ4844"
] | [
"official_reviewer",
"author",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"In this work, the authors aim to provide a principled way to evaluate and compare graph generative models. The authors initially list desirable criteria an evaluation metric should possess and subsequently discuss the usage of maximum mean discrepancy (MMD) for model comparison. Subsequently they highlight issues ... | [
8,
-1,
8,
-1,
8,
-1,
-1,
-1,
-1,
-1,
5
] | [
4,
-1,
5,
-1,
4,
-1,
-1,
-1,
-1,
-1,
3
] | [
"iclr_2022_tBtoZYKd9n",
"iclr_2022_tBtoZYKd9n",
"iclr_2022_tBtoZYKd9n",
"hSxoMxIZ5Ou",
"iclr_2022_tBtoZYKd9n",
"-XurEw2KHH",
"4l4ruJ4844",
"gt-bu4Hc58",
"Lr23PPj02_R",
"t6RHtDm2FXm",
"iclr_2022_tBtoZYKd9n"
] |
iclr_2022_o_HsiMPYh_x | Leveraging unlabeled data to predict out-of-distribution performance | Real-world machine learning deployments are characterized by mismatches between the source (training) and target (test) distributions
that may cause performance drops. In this work, we investigate methods for predicting the target domain accuracy using only labeled source data and unlabeled target data. We propose Aver... | Accept (Poster) | The authors propose a simple method to estimate the accuracy of a classifier on an unlabeled dataset given an in-distribution validation set. In extensive experiments the authors show that the proposed method is significantly more accurate than previous methods and other baselines.
The reviewers are quite consistent ... | val | [
"NHVNO3KdgYo",
"JrBFN1SNp_4",
"Z9OMv6Rj1cb",
"mFuSgfI-qIa",
"kGGTgk07wzF",
"T5K7nvABlx",
"Pca9MUqOfwJ",
"6hfASwqICA-",
"HHg1wFDPhL",
"Hs_8TGbtL7Q",
"tdvgYdTKlOS",
"CApaJ4DigET",
"rXxivCSjgkL",
"8JSP1zXA_tc",
"Q7AS9MsFIzK",
"P9AsdC0mPe5",
"RlqWWFDUDlk",
"9DJP-Z8Sz3g",
"DGsQbtSW1Ti... | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_re... | [
" Hi Authors, \nthank you for your response. My concerns are properly addressed at this point and I will raise my score accordingly (see update to my review above). ",
"The paper proposes a new scoring mechanism, Average Thresholded Confidence (ATC), to estimate out-of-distribution performance (i.e. accuracy) of ... | [
-1,
8,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
8,
6
] | [
-1,
3,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
3
] | [
"JrBFN1SNp_4",
"iclr_2022_o_HsiMPYh_x",
"CApaJ4DigET",
"iclr_2022_o_HsiMPYh_x",
"T5K7nvABlx",
"Pca9MUqOfwJ",
"Q7AS9MsFIzK",
"HHg1wFDPhL",
"JrBFN1SNp_4",
"8JSP1zXA_tc",
"mFuSgfI-qIa",
"9DJP-Z8Sz3g",
"DGsQbtSW1Ti",
"RlqWWFDUDlk",
"tdvgYdTKlOS",
"iclr_2022_o_HsiMPYh_x",
"iclr_2022_o_Hsi... |
iclr_2022_vA7doMdgi75 | Implicit Bias of Projected Subgradient Method Gives Provable Robust Recovery of Subspaces of Unknown Codimension | Robust subspace recovery (RSR) is the problem of learning a subspace from sample data points corrupted by outliers. Dual Principal Component Pursuit (DPCP) is a robust subspace recovery method that aims to find a basis for the orthogonal complement of the subspace by minimizing the sum of the distances of the points to... | Accept (Spotlight) | The paper looks at subspace recovery in the presence of outliers, of which there have been many formulations. They study a recent formulation, DPCP, but relax the requirement that the dimension of the subspace is known -- obviously very important in practice. The approach is quite clever: they exploit the fact that for... | train | [
"_NunGFAaUUU",
"9s9z5KKyhHE",
"bYIR4SQgY8C",
"ZndcnvUKnWC",
"m0npmcd1XZ",
"60RkumkhT-u",
"1x-ZtGAKx1i",
"rxNghDR0Fxs",
"s8FyU09_0WE"
] | [
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" We would like to thank the reviewer for the positive evaluation of our work. \nWe agree with the reviewer that condition in (9) and the overestimate of the subspace codimension are critical for the proposed PSGM algorithm to converge to a matrix $\\hat{\\mathbf{B}}$ that will span the correct orthogonal compleme... | [
-1,
-1,
-1,
-1,
-1,
5,
8,
6,
8
] | [
-1,
-1,
-1,
-1,
-1,
4,
3,
4,
2
] | [
"s8FyU09_0WE",
"rxNghDR0Fxs",
"60RkumkhT-u",
"1x-ZtGAKx1i",
"iclr_2022_vA7doMdgi75",
"iclr_2022_vA7doMdgi75",
"iclr_2022_vA7doMdgi75",
"iclr_2022_vA7doMdgi75",
"iclr_2022_vA7doMdgi75"
] |
iclr_2022_6NePxZwfae | Goal-Directed Planning via Hindsight Experience Replay | We consider the problem of goal-directed planning under a deterministic transition model. Monte Carlo Tree Search has shown remarkable performance in solving deterministic control problems. It has been extended from complex continuous domains through function approximators to bias the search of the planning tree in Alp... | Accept (Poster) | The main detractor of this paper feels that the paper makes a relatively small technical and empirical contribution given existing results on HER (Andrychowicz et al., NeurIPS 2017). However, several other reviewers, who had more engagement in the discussion, were strong supporters. Having looked at the paper myself I... | train | [
"Y7XZrw8AsZQ",
"nkyJZv7CG_O",
"pKbNelKLsgd",
"9xwYlk2XXPc",
"JihENUCs5l1",
"eL4ugdZgYZp",
"fIWM6kTnFc",
"UW17JHcoT7r",
"gWWhYlBtwas",
"JQ01uX3TDca",
"xraB6de1IgB",
"d8AALDrDMzj",
"54e0N4tSAN",
"SEru52jBRNQ"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"The paper addresses multi-goal reinforcement learning. It presents empirical results with a variation of an existing algorithm, Hindsight Experience Replay (HER). Specifically, the paper uses HER alongside AlphaZero. My main comment is that the technical and empirical novelty of the paper is relatively low. Andry... | [
3,
6,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8
] | [
4,
4,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5
] | [
"iclr_2022_6NePxZwfae",
"iclr_2022_6NePxZwfae",
"eL4ugdZgYZp",
"iclr_2022_6NePxZwfae",
"iclr_2022_6NePxZwfae",
"JQ01uX3TDca",
"xraB6de1IgB",
"d8AALDrDMzj",
"54e0N4tSAN",
"nkyJZv7CG_O",
"SEru52jBRNQ",
"Y7XZrw8AsZQ",
"JihENUCs5l1",
"iclr_2022_6NePxZwfae"
] |
iclr_2022_2_vhkAMARk | Escaping limit cycles: Global convergence for constrained nonconvex-nonconcave minimax problems | This paper introduces a new extragradient-type algorithm for a class of nonconvex-nonconcave minimax problems. It is well-known that finding a local solution for general minimax problems is computationally intractable. This observation has recently motivated the study of structures sufficient for convergence of first o... | Accept (Spotlight) | The paper considers the saddle point problem of finding non-convex/non-concave minimax solutions. Building onEG+ of Diakonikolas et al., 2021 that works under weak MVI conditions, the work presents a new algorithm CurvatureEG+ that works for a larger range of weak MVI condition compared to previous work and also works ... | train | [
"Pfmtc8PMwvA",
"gQqVtOaj9q0",
"P5IrVbwjVBD",
"UbzT3M2dp0E",
"dIFOpI5jBVI",
"5eIXjVZIPNC",
"xbldYpNoaD",
"1KjwP63M2DG",
"UPlP6LwlC4Y",
"NHw5iVEeX-",
"89WEs3A_ZBY",
"hn0Bm3AJnnw",
"AOun3sOPiE"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" I also keep my score 8. \nThe questions proposed by the reviewer wk1L who gave 5 points are solved well by the authors from my perspective. ",
" I have gone through other reviewers' comments and authors' detailed response, and I decided to keep my score. The clarity of the paper has been improved after revision... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
5,
8,
8
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3,
3,
3
] | [
"gQqVtOaj9q0",
"UPlP6LwlC4Y",
"xbldYpNoaD",
"dIFOpI5jBVI",
"5eIXjVZIPNC",
"AOun3sOPiE",
"hn0Bm3AJnnw",
"89WEs3A_ZBY",
"NHw5iVEeX-",
"iclr_2022_2_vhkAMARk",
"iclr_2022_2_vhkAMARk",
"iclr_2022_2_vhkAMARk",
"iclr_2022_2_vhkAMARk"
] |
iclr_2022_wqD6TfbYkrn | A Conditional Point Diffusion-Refinement Paradigm for 3D Point Cloud Completion | 3D point clouds are an important data format that captures 3D information for real world objects. Since 3D point clouds scanned in the real world are often incomplete, it is important to recover the complete point cloud for many downstreaming applications. Most existing point cloud completion methods use the Chamfer D... | Accept (Poster) | This paper presents an approach based on conditional denoising diffusion models for point cloud completion. The reviewers have recognized the significance of contributions, the clarity of presentation, and the comprehensivity of experiments. I am happy to recommend this paper for presentation at ICLR. | train | [
"Ugsi0PDYafl",
"XLMdjfVhrZY",
"kZxgHbKuOM7",
"TnryXfllJMC",
"qeDqAlxQ1NJ",
"60DBNpJT8ZU"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer"
] | [
"The paper tackles the problem of shape-level point cloud completion in a fully supervised setting. The big picture of the proposed method is to use a conditional DDPM to generate a noisy but complete point cloud given incomplete input, and then use another refinement network, also conditional on the incomplete inp... | [
8,
6,
-1,
-1,
-1,
8
] | [
4,
5,
-1,
-1,
-1,
5
] | [
"iclr_2022_wqD6TfbYkrn",
"iclr_2022_wqD6TfbYkrn",
"XLMdjfVhrZY",
"Ugsi0PDYafl",
"60DBNpJT8ZU",
"iclr_2022_wqD6TfbYkrn"
] |
iclr_2022_ljxWpdBl4V | Closed-form Sample Probing for Learning Generative Models in Zero-shot Learning | Generative model based approaches have led to significant advances in zero-shot learning (ZSL) over the past few years. These approaches typically aim to learn a conditional generator that synthesizes training samples of classes conditioned on class definitions. The final zero-shot learning model is then obtained by tr... | Accept (Poster) | The paper proposes to improve (generalized) zero-shot learning, by training a generator jointly with the classification task, such that it generates samples that reduce the classification loss. To achieve this, they use a zero shot model that has a (differentiable) closed form solution (ESZSL), so the full model can b... | train | [
"vLs5a7WVBOc",
"lLE1yDjpze",
"qDIhyGivD2G",
"o5yv4imD0p3",
"2R1YGaGoYrJ",
"u9gTN4_wznK",
"98Z98hWEbeG",
"amnOH9Y-_gE",
"EF5CqIPlS6J",
"R69FXAKcdtd",
"fkQ_ibRlUsr",
"Xp9LMULne80",
"6QySZcIYyRl",
"ZTOCCIxrJTy",
"jzMLX_iCxh",
"eToE-dS7Dwd"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"This paper aims to address the problem of generalized zero-shot classification based on generative models. The main contribution is that it considers training and evaluating a generative model in synthesizing training examples that are helpful to improve classification performance. To this end, it leverages the ze... | [
6,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
5,
6
] | [
4,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
2,
5,
4
] | [
"iclr_2022_ljxWpdBl4V",
"iclr_2022_ljxWpdBl4V",
"98Z98hWEbeG",
"vLs5a7WVBOc",
"eToE-dS7Dwd",
"lLE1yDjpze",
"lLE1yDjpze",
"ZTOCCIxrJTy",
"ZTOCCIxrJTy",
"vLs5a7WVBOc",
"jzMLX_iCxh",
"jzMLX_iCxh",
"iclr_2022_ljxWpdBl4V",
"iclr_2022_ljxWpdBl4V",
"iclr_2022_ljxWpdBl4V",
"iclr_2022_ljxWpdBl4... |
iclr_2022_cGDAkQo1C0p | Reversible Instance Normalization for Accurate Time-Series Forecasting against Distribution Shift | Statistical properties such as mean and variance often change over time in time series, i.e., time-series data suffer from a distribution shift problem. This change in temporal distribution is one of the main challenges that prevent accurate time-series forecasting. To address this issue, we propose a simple yet effect... | Accept (Poster) | This paper introduces the "reversible instance normalization" (RevIN), a method for addressing temporal distribution shift in time-series forecasting. RevIN consists in normalizing (subtracting the mean and dividing by the standard deviation) each layer of of deep neural network in a given temporal window for a given i... | train | [
"MlvogQVnOYA",
"4sXWNIJwF6z",
"Rcp0CAWvdbg",
"F_6HQI0UEIM",
"OtfTrY9Sxr",
"XkylkqLxrOl",
"Vt0Rbo6jlE1",
"M9nKv_Zzp5I",
"ugldmCn9TB2",
"o5xltoGCZP",
"l3RsZvPUJl",
"Uf1aUPZH_V",
"Esn5VK0sGj",
"hzc2um10xE1"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"This work proposes to use a normalization method to address temporal distribution shift in time-series forecasting. The proposed approach, *RevIN*, consists of two steps: instance normalization on input sequences and \"de-normalization\" of output sequences by re-using statistics (mean and variance) computed durin... | [
6,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5
] | [
3,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4
] | [
"iclr_2022_cGDAkQo1C0p",
"iclr_2022_cGDAkQo1C0p",
"hzc2um10xE1",
"4sXWNIJwF6z",
"4sXWNIJwF6z",
"ugldmCn9TB2",
"MlvogQVnOYA",
"Esn5VK0sGj",
"o5xltoGCZP",
"l3RsZvPUJl",
"4sXWNIJwF6z",
"4sXWNIJwF6z",
"iclr_2022_cGDAkQo1C0p",
"iclr_2022_cGDAkQo1C0p"
] |
iclr_2022_U4uFaLyg7PV | T-WaveNet: A Tree-Structured Wavelet Neural Network for Time Series Signal Analysis | Time series signal analysis plays an essential role in many applications, e.g., activity recognition and healthcare monitoring.
Recently, features extracted with deep neural networks (DNNs) have shown to be more effective than conventional hand-crafted ones.
However, most existing solutions rely solely on the network t... | Accept (Poster) | This paper introduces a tree-structured wavelet deep neural network to effectively extract more discriminative and expressive feature representations in time series signals. Based on a frequency spectrum energy analysis, the approach decomposes input signals into multiple subbands and builds a tree structure with dat... | train | [
"1VZeE3EOtNu",
"P5A62I5MWx",
"GhHxYLoHOL2",
"PcyN6z9nqp"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"This paper introduces T-Wavenet, a fused approach including both classic feature engineering and recent advance of deep learning, in an attempt to improve the state-of-the-art analysis of time series. Empirical study is mainly conducted on human activity or health monitoring data to support the design decisions. ... | [
6,
8,
6,
6
] | [
4,
3,
2,
3
] | [
"iclr_2022_U4uFaLyg7PV",
"iclr_2022_U4uFaLyg7PV",
"iclr_2022_U4uFaLyg7PV",
"iclr_2022_U4uFaLyg7PV"
] |
iclr_2022_figzpGMrdD | Pretrained Language Model in Continual Learning: A Comparative Study | Continual learning (CL) is a setting in which a model learns from a stream of incoming data while avoiding to forget previously learned knowledge. Pre-trained language models (PLMs) have been successfully employed in continual learning of different natural language problems. With the rapid development of many continua... | Accept (Poster) | This paper presents a comparison and analysis of continual learning methods for pretrained language models. The authors categorise continual learning methods into three categories, those that use cross task regularisation, those that employ some form of experience replay of previous training examples, and those that dy... | train | [
"eUpJaZpzdsM",
"0fEFzpOP1na",
"f3FRIcTRJVF",
"Uq7Rr1QTrH",
"8WyiDUNHss",
"T3-rjALZF8w",
"THfsDi8skUi",
"H-fyceAnFJc",
"fklgqgq6wpz",
"jba1rokxQO1",
"Bj8K6YPwt9J",
"fKlyAPB6drG",
"v17qRS6HB4h",
"zhqKTyU59e"
] | [
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" | MAVEN | Method | task1 | task2 | task3 | task4 | task5 | task6 | task7 | task8 | task9 | task10 | task11 | task12 | task13 | task14 | task15 | task16 |\n|---------|---------|--------|-------|-------|-------|-------|-------|-------|-------|-------|--------|--------|--------|--------|--------|--------|-------... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
5,
6,
8
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
4,
5
] | [
"0fEFzpOP1na",
"Uq7Rr1QTrH",
"iclr_2022_figzpGMrdD",
"fklgqgq6wpz",
"Bj8K6YPwt9J",
"fKlyAPB6drG",
"zhqKTyU59e",
"v17qRS6HB4h",
"f3FRIcTRJVF",
"H-fyceAnFJc",
"iclr_2022_figzpGMrdD",
"iclr_2022_figzpGMrdD",
"iclr_2022_figzpGMrdD",
"iclr_2022_figzpGMrdD"
] |
iclr_2022_f9D-5WNG4Nv | Online Coreset Selection for Rehearsal-based Continual Learning | A dataset is a shred of crucial evidence to describe a task. However, each data point in the dataset does not have the same potential, as some of the data points can be more representative or informative than others. This unequal importance among the data points may have a large impact in rehearsal-based continual lear... | Accept (Poster) | The authors propose three strategies for coreset selection in the context of continual learning. In particular, the authors consider class-imbalance and noisy scenarios. The authors run extensive benchmarks and ablation showing that the approach can be effective in practice. All reviewers were positive about this work,... | test | [
"1SJRu1OTxXh",
"l-WjbJ9dfje",
"va1s4AyWOB",
"BMVFFFBO46M",
"WIZyEE-8xp_",
"NdJLwZHWFeq",
"3d0l6u-SB4h",
"cCIUZcwZJLc",
"YAie8P0GaWX",
"0bZxjmbM008",
"KXiL4gHQyav",
"lEcjtphGJu2",
"BVeeReAVu4W",
"VcdCPLbgH9t",
"jXgmixJ_egK",
"owXfW30tGlB",
"TWN3vIZFa-v",
"xBBPgASgRxX",
"gjXl-44fT_... | [
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
... | [
" Dear Reviewer iADy, \n\nWe sincerely appreciate your effort for constructive feedback. One thing to note is that **there is no additional phase for the reviewers' final decision**, and now ACs are writing meta-reviews and recommending the submissions. \n\n**Could you finalize your decision reflecting our respon... | [
-1,
-1,
-1,
-1,
-1,
-1,
6,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
5
] | [
-1,
-1,
-1,
-1,
-1,
-1,
4,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4
] | [
"BMVFFFBO46M",
"WIZyEE-8xp_",
"BMVFFFBO46M",
"BVeeReAVu4W",
"NdJLwZHWFeq",
"cCIUZcwZJLc",
"iclr_2022_f9D-5WNG4Nv",
"VcdCPLbgH9t",
"KXiL4gHQyav",
"iclr_2022_f9D-5WNG4Nv",
"0bZxjmbM008",
"3d0l6u-SB4h",
"pcppp5NMh15",
"jMAKn33nVHO",
"0bZxjmbM008",
"jMAKn33nVHO",
"0bZxjmbM008",
"pcppp5... |
iclr_2022_USIgIY6TNDe | Graph-based Nearest Neighbor Search in Hyperbolic Spaces | The nearest neighbor search (NNS) problem is widely studied in Euclidean space, and graph-based algorithms are known to outperform other approaches for this task. However, hyperbolic geometry often allows for better data representation in various domains, including graphs, words, and images. In this paper, we show that... | Accept (Poster) | The paper studies the nearest-neighbor search problem for objects embedded in hyperbolic space. It develops a graph-based approach to NNS in hyperbolic space, showing (interestingly) that the time complexity of graph-based NNS can be lower in hyperbolic space than in Euclidean space under some assumptions. This nice th... | train | [
"xRNa4Qv63P9",
"K2RgW9KSiS",
"pSa-09oLAki",
"KtYiskhF_o",
"ZJPYJXYYmIR",
"L0jf7yXvT4",
"v_utPTJBMq1",
"MZLbpTNR0K5",
"T-VRscP1WCw",
"1m9xRQN52-5",
"AUTidq-02TR",
"khrbu5DUtaA",
"SYPb-C6Yfyl",
"NDwMrqpTUW",
"1v35cwnYK9f",
"LDpTAoG3Ucs",
"gSvs7oE45b7",
"lJXMuYL4Rcy",
"ktbA5DlzOgD",... | [
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
... | [
" Dear Reviewer, We would like to ask whether your additional concerns raised during the discussion period have been addressed in the revised version of the paper. Your feedback and comments would help us to improve the paper. Thank you! ",
"The paper considers the setting of nearest neighbor search (NNS) and dev... | [
-1,
5,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
6,
-1,
-1,
-1,
6
] | [
-1,
4,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
-1,
-1,
-1,
3,
-1,
-1,
-1,
3
] | [
"gSvs7oE45b7",
"iclr_2022_USIgIY6TNDe",
"khrbu5DUtaA",
"NDwMrqpTUW",
"iclr_2022_USIgIY6TNDe",
"MZLbpTNR0K5",
"zVBCNrr4Fqv",
"3HUttATisWk",
"iclr_2022_USIgIY6TNDe",
"Pdpwqp8R8Kc",
"K2RgW9KSiS",
"K2RgW9KSiS",
"3Ajz3w7Bd5g",
"ZJPYJXYYmIR",
"Pdpwqp8R8Kc",
"gSvs7oE45b7",
"iclr_2022_USIgIY... |
iclr_2022_OjPmfr9GkVv | Enhancing Cross-lingual Transfer by Manifold Mixup | Based on large-scale pre-trained multilingual representations, recent cross-lingual transfer methods have achieved impressive transfer performances. However, the performance of target languages still lags far behind the source language. In this paper, our analyses indicate such a performance gap is strongly associated ... | Accept (Poster) | This paper proposes X-Mixup, a model that considers the source languages and target languages together for cross-lingual transfer. The designed model takes a pair of sentences (or the translated sentences) in a source language and a target language as the input and computes the cross-attention between them.
The empi... | train | [
"Ih9rF9U0i9",
"hpETAIWt0EO",
"m2KVxQaKm4K",
"N9630hYUMJ2",
"qrA0D5f4cqZ",
"79br-cnMW4",
"CTKigbFy0_c",
"IrfNt8lBBvB",
"mOjI2VEIRH",
"KW33HBh-1Yr",
"t_z2zTO2OU",
"KnpZ8VjLdut",
"030Yqlexgp7j",
"ryCrehDKfep",
"qgxzcob69E",
"sR8r9Pi9-8R"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"The paper proposes cross-lingual mixup, a technique which performs manifold mixup of source and target sequences. This technique also includes mixup ratio which factors in MT quality and scheduled sampling which deals with exposure bias. Experiments on 3 task types from the XTREME benchmark show that x-mixup leads... | [
8,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
6
] | [
4,
-1,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
4
] | [
"iclr_2022_OjPmfr9GkVv",
"CTKigbFy0_c",
"iclr_2022_OjPmfr9GkVv",
"qrA0D5f4cqZ",
"KW33HBh-1Yr",
"KnpZ8VjLdut",
"030Yqlexgp7j",
"mOjI2VEIRH",
"t_z2zTO2OU",
"m2KVxQaKm4K",
"sR8r9Pi9-8R",
"qgxzcob69E",
"Ih9rF9U0i9",
"iclr_2022_OjPmfr9GkVv",
"iclr_2022_OjPmfr9GkVv",
"iclr_2022_OjPmfr9GkVv"
... |
iclr_2022_UMfhoMtIaP5 | Provably Robust Adversarial Examples | We introduce the concept of provably robust adversarial examples for deep neural networks – connected input regions constructed from standard adversarial examples which are guaranteed to be robust to a set of real-world perturbations (such as changes in pixel intensity and geometric transformations). We present a novel... | Accept (Poster) | In this paper, authors introduce and study provably robust adversarial examples. Reviewers had mixed thoughts on the work. One reviewer mentioned that the "provable" robustness is somehow overstated in the work: looking at the title and abstract, it sounds like the paper develops a new algorithm that is guaranteed to b... | train | [
"R6LtOybiofH",
"PwBB9Ut0JAi",
"1dorgGkelPk",
"LZXEZbxL1WA",
"_fAuCG6CMN2",
"uZh9qtaVA7x",
"C2zIpXMrdhU",
"HJU87UTHIi",
"coMO6L592jI",
"e2e27GvySrX",
"FMwC5Mw0Q08",
"f6dPpK3He26",
"bU8CVOWQ8ev",
"4eiAPG5gWRR",
"RT-t_9ttxPm",
"rJn13BAcZq3",
"Mv9_J2JZUsu",
"LPND9f3q3En"
] | [
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author"
] | [
"The main contribution of the paper is a framework that will output a large batch of adversarial examples, assuming access to a few black-box mechanisms. The term \"provably robust\" appears misleading; there is no theory showing that the examples must be adversarial.\n\nWhile authors highlight that there are massi... | [
6,
-1,
-1,
8,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
4,
-1,
-1,
5,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
"iclr_2022_UMfhoMtIaP5",
"uZh9qtaVA7x",
"coMO6L592jI",
"iclr_2022_UMfhoMtIaP5",
"iclr_2022_UMfhoMtIaP5",
"4eiAPG5gWRR",
"f6dPpK3He26",
"e2e27GvySrX",
"FMwC5Mw0Q08",
"Mv9_J2JZUsu",
"LPND9f3q3En",
"bU8CVOWQ8ev",
"R6LtOybiofH",
"_fAuCG6CMN2",
"_fAuCG6CMN2",
"LZXEZbxL1WA",
"LZXEZbxL1WA",... |
iclr_2022_BrFIKuxrZE | Fair Normalizing Flows | Fair representation learning is an attractive approach that promises fairness of downstream predictors by encoding sensitive data. Unfortunately, recent work has shown that strong adversarial predictors can still exhibit unfairness by recovering sensitive attributes from these representations. In this work, we present ... | Accept (Poster) | This paper addresses fair representation learning, with the aim of obstructing the recovery of sensitive features from the learned representation, hence enforcing the fairness of subsequent prediction tasks. In the setting where probability density can be estimated for sensitive groups, Fair Normalizing Flows (FNF) tr... | train | [
"TDAgia1tIvf",
"cXVzs8SCH_K",
"p0d6OTHgyzY",
"wRm38QdtcJr",
"8W-MzvRSbFX",
"Fw2n2dPeZC",
"Ym3zatRJDG_",
"VehmklfGQhV",
"R16WDEiX72B",
"0q_RzNIf-Uj",
"u7YnZdGrOc3",
"bmJLeQFTaQ0",
"m7pY1wjMTcG",
"Ywy6sx2CXDT",
"A6-8ZYVtX2g",
"9uVjNMo2Avh",
"fxCsFLzDPho",
"GWd0gLlnnb4",
"OsACfnEwY-... | [
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"of... | [
" Thank you for your response. We address your concerns below.\n\n**Alright, to each their own.**\n\nWe would appreciate it if you could provide reasons for why you disagree with our response and a more constructive feedback would certainly help us improve our paper.\n\n**Can I solve the problem by assuming that so... | [
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
6
] | [
-1,
-1,
-1,
-1,
3,
-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,
2,
4
] | [
"p0d6OTHgyzY",
"wRm38QdtcJr",
"R16WDEiX72B",
"Ym3zatRJDG_",
"iclr_2022_BrFIKuxrZE",
"oGGh9-ETcn",
"0q_RzNIf-Uj",
"iclr_2022_BrFIKuxrZE",
"0q_RzNIf-Uj",
"u7YnZdGrOc3",
"9uVjNMo2Avh",
"Ywy6sx2CXDT",
"p9LHD7IKOvB",
"zq3v_u9Haa",
"iclr_2022_BrFIKuxrZE",
"pySAjjhDacC",
"iclr_2022_BrFIKuxr... |
iclr_2022_jJOjjiZHy3h | Defending Against Image Corruptions Through Adversarial Augmentations | Modern neural networks excel at image classification, yet they remain vulnerable to common image corruptions such as blur, speckle noise or fog. Recent methods that focus on this problem, such as AugMix and DeepAugment, introduce defenses that operate in expectation over a distribution of image corruptions. In contrast... | Accept (Poster) | The paper proposes an adversarial data augmentation technique searching for adversarial weight perturbations of a corruption network (e.g. a pretrained image-to-image model). The goal is to achieve common corruption robustness as well as a non-trivial level of adversarial robustness. The authors claim state-of-the-art-... | train | [
"PLnI1FotrO",
"6iB9Jfl6Xx",
"aXAWOMGgTF",
"HoSS4X-5nY3",
"0ZfDEy85Ua9",
"rmkql4WXkHo",
"nD8qsfq7zOE",
"SUDj8IdxZI",
"1J1LeTE43gU",
"eX0h8ZbmE6b",
"gy2d4bzrNJd",
"_XVUEUHhiGM",
"wPHu6iN-fss",
"T1MSR9SAIJf",
"TXgmJW8OmRY",
"CWEhlC8CfQo",
"_tgKEfbFmLj",
"DnQl1reKPZ3",
"IqO9ljLZpUY",... | [
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"a... | [
" I thank for the detailed responses. The paper now clearly discusses the limitations and compares the proposed method with other baselines with similar computational requirements. Based on this, I revised the correctness score as well as the recommendation score.",
"The authors propose a new data augmentation te... | [
-1,
8,
-1,
-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,
-1,
6
] | [
-1,
3,
-1,
-1,
-1,
4,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5
] | [
"eX0h8ZbmE6b",
"iclr_2022_jJOjjiZHy3h",
"rmkql4WXkHo",
"nD8qsfq7zOE",
"CWEhlC8CfQo",
"iclr_2022_jJOjjiZHy3h",
"iclr_2022_jJOjjiZHy3h",
"XRlwXCszbT",
"XRlwXCszbT",
"4Deaawc3Vau",
"_tgKEfbFmLj",
"TXgmJW8OmRY",
"T1MSR9SAIJf",
"p7P10rIihow",
"OHtYT3F3dC",
"SFK2tRp90p",
"Td_7CHiBB_s",
"... |
iclr_2022_q7n2RngwOM | $\beta$-Intact-VAE: Identifying and Estimating Causal Effects under Limited Overlap | As an important problem in causal inference, we discuss the identification and estimation of treatment effects (TEs) under limited overlap; that is, when subjects with certain features belong to a single treatment group. We use a latent variable to model a prognostic score which is widely used in biostatistics and suff... | Accept (Poster) | In this paper, the authors proposed a method for causal inference under limited overlap -- an important and understudied complication. The authors propose to recover a prognostic score using a variational autoencoder, and thereby map a higher dimensional set of covariates with limited overlap to a lower dimensional se... | val | [
"WvODp4PCjoH",
"9aDzmsN-Qj",
"6j_QRI5Veop",
"5b8QAyl0TvV",
"Msgx1RtAip",
"oDRTj2zz2Ea",
"9u6kFNs_2uc",
"bNFmvbkS666",
"qCG3zHGys77",
"svzDBG44RMM",
"fnyslN136UX",
"YJYDpC2JPh1",
"LNZSOWqqMeL",
"fimPzOgPr-",
"Ua8JMg__fFk",
"ScUv-6qLfnE",
"g2j634Fmwi_",
"p58NZYd9COq",
"FTdq1u0bOCt"... | [
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" And thanks for the affirmations of our replies and revisions!\n\nOn related work, we will update the paper and add more pointers to the Appendix, and possibly move some important points to the main text.",
" (Real-world scenario). \n\nWe believe the real-world examples in Appendix C.2 are possible examples that... | [
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
6,
6
] | [
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
3,
3
] | [
"6j_QRI5Veop",
"5b8QAyl0TvV",
"iclr_2022_q7n2RngwOM",
"YJYDpC2JPh1",
"6j_QRI5Veop",
"fnyslN136UX",
"iclr_2022_q7n2RngwOM",
"T-7GKkGxr5E",
"svzDBG44RMM",
"ScUv-6qLfnE",
"g2j634Fmwi_",
"FTdq1u0bOCt",
"6j_QRI5Veop",
"iclr_2022_q7n2RngwOM",
"p58NZYd9COq",
"T-7GKkGxr5E",
"6j_QRI5Veop",
... |
iclr_2022_f2lrIbGx3x7 | Bayesian Framework for Gradient Leakage | Federated learning is an established method for training machine learning models without sharing training data. However, recent work has shown that it cannot guarantee data privacy as shared gradients can still leak sensitive information. To formalize the problem of gradient leakage, we propose a theoretical framework ... | Accept (Poster) | The paper formalizes the problem of gradient leakage through a Bayesian framework. They show that existing attacks can be viewed as approximations of a Bayesian optimal adversary. The empirical results show that heuristic defences are not good against stronger attacks and that the early part of the training is particul... | train | [
"bRkXbIWIDUE",
"4x9asivrDrX",
"fvfqJ1NzMf8",
"FZ1kPsj0lBe",
"j7HAUfNZOEO",
"R6n4r3CsPdE",
"baHFcBxRW6d",
"rx22kmNxpp9",
"m1PohuuzLMG",
"fvBvbxv3OVM",
"a3Mqj9zikTH",
"UUdNASSY7bR",
"p8EyW1vdxTF",
"XDsGSvFy5P4",
"nzmDjrVDyV",
"9SeJ5D4DvXp",
"as7ZXjVmbGG"
] | [
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
" Thank you for the response. We address your concerns below.\n\n**Is the framework not practical without some kind of mathematical approximation?**\n\nNote that attacks in Section 4 are obtained using a mathematical approximation: we first apply Jensen’s inequality, and then estimate the integral using Monte Carlo... | [
-1,
-1,
-1,
8,
6,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6
] | [
-1,
-1,
-1,
5,
5,
-1,
-1,
-1,
2,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4
] | [
"4x9asivrDrX",
"UUdNASSY7bR",
"baHFcBxRW6d",
"iclr_2022_f2lrIbGx3x7",
"iclr_2022_f2lrIbGx3x7",
"rx22kmNxpp9",
"fvBvbxv3OVM",
"a3Mqj9zikTH",
"iclr_2022_f2lrIbGx3x7",
"XDsGSvFy5P4",
"nzmDjrVDyV",
"as7ZXjVmbGG",
"m1PohuuzLMG",
"FZ1kPsj0lBe",
"j7HAUfNZOEO",
"iclr_2022_f2lrIbGx3x7",
"iclr... |
iclr_2022_wQfgfb8VKTn | Context-Aware Sparse Deep Coordination Graphs | Learning sparse coordination graphs adaptive to the coordination dynamics among agents is a long-standing problem in cooperative multi-agent learning. This paper studies this problem and proposes a novel method using the variance of payoff functions to construct context-aware sparse coordination topologies. We theoreti... | Accept (Spotlight) | All reviewers found that the paper offers interesting contributions for multi-agent RL and favour acceptance of the paper. The strengths of the paper are summarized below:
- Good algorithmic contribution
- Offers a new set of benchmark tasks for coordination in MARL settings
- Exhaustive experiments on complex tasks wi... | train | [
"rj1DAkkfQDk",
"7XTBaG-3_aJ",
"MR5Mp3TSTni",
"Jxjwg5Vz7sr",
"6NEDI2mhBp",
"N3r01o5R6vc",
"4wxVhrJliwW",
"y3A-ofpo_zs",
"qyREzNdDgit",
"dttgr0aNZNY",
"a9lC7iYmK8u",
"KLk-zlMVid2",
"0NneMPMSRTK",
"FZ0OWNXnTsm",
"we8MfEGpIgV",
"UFpva-ko5mv",
"MIgRFihvlQ"
] | [
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
" Thanks for the helpful clarifications. All my questions have been addressed.",
" Dear reviewer 7Rh6,\n\nThank you for the valuable suggestions. To address your concerns, we make the following revisions to our submission:\n\n(1) In Appendix G, we comprehensively study whether the influence of cycles is severe. T... | [
-1,
-1,
-1,
8,
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
8,
6
] | [
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3
] | [
"MIgRFihvlQ",
"N3r01o5R6vc",
"6NEDI2mhBp",
"iclr_2022_wQfgfb8VKTn",
"a9lC7iYmK8u",
"MIgRFihvlQ",
"dttgr0aNZNY",
"UFpva-ko5mv",
"iclr_2022_wQfgfb8VKTn",
"FZ0OWNXnTsm",
"KLk-zlMVid2",
"0NneMPMSRTK",
"Jxjwg5Vz7sr",
"qyREzNdDgit",
"MIgRFihvlQ",
"iclr_2022_wQfgfb8VKTn",
"iclr_2022_wQfgfb8... |
iclr_2022_Ug-bgjgSlKV | Finding an Unsupervised Image Segmenter in each of your Deep Generative Models | Recent research has shown that numerous human-interpretable directions exist in the latent space of GANs. In this paper, we develop an automatic procedure for finding directions that lead to foreground-background image separation, and we use these directions to train an image segmentation model without human supervisio... | Accept (Poster) | The reviewers all agree that this paper proposes a very interesting approach of finding useful information encoded inside a generative model. They show how foreground/background semantics learnt in a generative model are useful for tasks like segmentation.
This is a general approach that can be applied to other models ... | train | [
"mM0y8GZnP97",
"KCgz3UPpA2K",
"PrOGLfDoMOj",
"gxZADuNGqst",
"xbxKUioGVbJ",
"7RwurDsXzu0",
"TQ23Z5cesTM",
"m9Bwq1xVFu",
"4M7UtJiBTxC",
"MSP0HLAGkL3",
"4lerEoB_-ux",
"jgPrI6FTy81"
] | [
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"public",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Dear Reviewers, \n\nThank you for your time and the constructive reviews. We appreciate your thorough readings of the paper and your insightful questions. Based on these questions, we have added additional discussions to our paper and conducted five new experiments to further demonstrate the usefulness of our met... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
6,
5,
6
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3,
3,
3
] | [
"iclr_2022_Ug-bgjgSlKV",
"m9Bwq1xVFu",
"jgPrI6FTy81",
"jgPrI6FTy81",
"4lerEoB_-ux",
"MSP0HLAGkL3",
"4M7UtJiBTxC",
"iclr_2022_Ug-bgjgSlKV",
"iclr_2022_Ug-bgjgSlKV",
"iclr_2022_Ug-bgjgSlKV",
"iclr_2022_Ug-bgjgSlKV",
"iclr_2022_Ug-bgjgSlKV"
] |
iclr_2022_xCVJMsPv3RT | Dropout Q-Functions for Doubly Efficient Reinforcement Learning | Randomized ensembled double Q-learning (REDQ) (Chen et al., 2021b) has recently achieved state-of-the-art sample efficiency on continuous-action reinforcement learning benchmarks. This superior sample efficiency is made possible by using a large Q-function ensemble. However, REDQ is much less computationally efficient ... | Accept (Poster) | The manuscript describes a method for improving the computational efficiency of randomized ensemble double Q-learning for continuous action RL, by using a small ensemble of Q-functions equipped with dropout and layer normalization, achieving matched sample efficiency at considerably less computational cost.
Reviewers ... | train | [
"CLXa2gjfYhs",
"nN3WnYnJtSK",
"Gx-bEyntJsg",
"vtewHB7bX_O",
"uzmfmfAydXk",
"RHS_fHkHIzY",
"Z9JhgIJNzSb",
"8D5FT_TKJSZ",
"6IdZIJnQBk",
"bFlMw-WX25r",
"EYRtCcBdBfl",
"N85q-mlwwf",
"dfH5EmKz6I5",
"CqZ2oQujMuo",
"NvTVKMecTAS",
"1HyHtweyyv"
] | [
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"public",
"official_reviewer"
] | [
" Thank you so much for your suggestions and feedback on our revision. \nWe will reflect your suggestions in the next revision.\n\n**Q1:** For Hopper, not sure what happened, perhaps it got stuck in a local optima? How many data do you use for initial exploration?\n\n**A1:** As in the REDQ paper, the number of ... | [
-1,
-1,
-1,
-1,
6,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6
] | [
-1,
-1,
-1,
-1,
3,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4
] | [
"8D5FT_TKJSZ",
"vtewHB7bX_O",
"RHS_fHkHIzY",
"CqZ2oQujMuo",
"iclr_2022_xCVJMsPv3RT",
"dfH5EmKz6I5",
"iclr_2022_xCVJMsPv3RT",
"EYRtCcBdBfl",
"Z9JhgIJNzSb",
"NvTVKMecTAS",
"Z9JhgIJNzSb",
"uzmfmfAydXk",
"uzmfmfAydXk",
"1HyHtweyyv",
"iclr_2022_xCVJMsPv3RT",
"iclr_2022_xCVJMsPv3RT"
] |
iclr_2022_7TZeCsNOUB_ | Collapse by Conditioning: Training Class-conditional GANs with Limited Data | Class-conditioning offers a direct means to control a Generative Adversarial Network (GAN) based on a discrete input variable. While necessary in many applications, the additional information provided by the class labels could even be expected to benefit the training of the GAN itself. On the contrary, we observe that ... | Accept (Poster) | This paper examines conditional GANs, which are found to lead to model collapse in low data settings. The paper proposes what appears to be a simple but effective method that addresses the issue. Reviewers were generally happy with the experiments and the utility of the observations and analysis. Code for the method wa... | train | [
"_T-razGQNZT",
"1DRSHG4I942",
"4Ca61fpanKM",
"n-VQsIUEur",
"HrJCjtbyv5h",
"p6S6UUHx3ZT",
"14Wgamjx8Bz",
"TZazW28DD5j",
"I-c83w6WdFF",
"sJWofuxoOqN",
"OLHv7gpOwd",
"PIAnX2iHwA9",
"JGgbGGdJK0n",
"kR1MNnL6fQd",
"0curQshYD19",
"-mrfY8zYKfK"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Thanks for the authors' response and most of my concerns have been addressed.",
"In this paper, the authors work towards training conditional GANs with limited data. Based on the observation that conditional GAN training suffers worse mode collapse than unconditional training, the authors proposed a training st... | [
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
8,
6
] | [
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
4
] | [
"TZazW28DD5j",
"iclr_2022_7TZeCsNOUB_",
"HrJCjtbyv5h",
"I-c83w6WdFF",
"14Wgamjx8Bz",
"sJWofuxoOqN",
"OLHv7gpOwd",
"kR1MNnL6fQd",
"PIAnX2iHwA9",
"0curQshYD19",
"1DRSHG4I942",
"-mrfY8zYKfK",
"iclr_2022_7TZeCsNOUB_",
"iclr_2022_7TZeCsNOUB_",
"iclr_2022_7TZeCsNOUB_",
"iclr_2022_7TZeCsNOUB_... |
iclr_2022_zNR43c03lRy | Learning to Annotate Part Segmentation with Gradient Matching | The success of state-of-the-art deep neural networks heavily relies on the presence of large-scale labelled datasets, which are extremely expensive and time-consuming to annotate. This paper focuses on tackling semi-supervised part segmentation tasks by generating high-quality images with a pre-trained GAN and labellin... | Accept (Poster) | This work deals with training generators of aligned pairs of images and segmentation maps. It is based on the recent DatasetGAN approach, which generates images and maps, but requires human annotations on a handful of generated images. This paper is addressing this problem by learning the annotation model over annotate... | train | [
"g1hhvDiMHVn",
"SFvekxpsqVA",
"RCm_Led0Gfs",
"HI5dVMRG868",
"ondKf58yxXh",
"_VXGf3EX4zm",
"m8bo_W6XCz",
"cxHXAtfVsSu",
"K1hwLF4A5cS",
"_ndiRK9euQM",
"wuL_-_5itz",
"JoTqOK6ZP4D",
"vBK4OV9c0L_",
"FjyHhEQnrPD",
"Ffs1U2Q2C-V",
"rjSz-bbkP7A",
"2cwCXU8MK2F",
"dkV0IaWyGJW"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author"
] | [
"In this paper, the authors mainly propose a gradient-matching-based method for part segmentation to reduce the annotation cost. Based on the DatasetGAN, the proposed model also used the Style GAN family to generate high-quality images and remove the human annotations on a handful of synthesized images. Compared wi... | [
6,
-1,
-1,
8,
-1,
6,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
3,
-1,
-1,
3,
-1,
4,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
"iclr_2022_zNR43c03lRy",
"Ffs1U2Q2C-V",
"vBK4OV9c0L_",
"iclr_2022_zNR43c03lRy",
"FjyHhEQnrPD",
"iclr_2022_zNR43c03lRy",
"cxHXAtfVsSu",
"_ndiRK9euQM",
"iclr_2022_zNR43c03lRy",
"2cwCXU8MK2F",
"_VXGf3EX4zm",
"iclr_2022_zNR43c03lRy",
"HI5dVMRG868",
"_VXGf3EX4zm",
"g1hhvDiMHVn",
"K1hwLF4A5c... |
iclr_2022_mHu2vIds_-b | Boosting Randomized Smoothing with Variance Reduced Classifiers | Randomized Smoothing (RS) is a promising method for obtaining robustness certificates by evaluating a base model under noise. In this work, we: (i) theoretically motivate why ensembles are a particularly suitable choice as base models for RS, and (ii) empirically confirm this choice, obtaining state-of-the-art results in... | Accept (Spotlight) | This paper integrates model ensembles with randomized smoothing to improve the certified accuracy. The methodology is motivated theoretically by showing the effect of model ensemble on reducing the variance of smooth classifiers. Moreover, it proposes an adaptive sampling algorithm to reduce the computation required fo... | val | [
"a2i0kEwiuUh",
"csaZ4n6sNPI",
"z_XscCMNetZ",
"8ktQUAbd8-U",
"68gj_svF1jc",
"Ecg1XNAJqT",
"REswunzFryH",
"_PZAlQt8xkL",
"FfwvYZXwCNX",
"enlF8rj8X3G",
"-P5d9gq-HyY"
] | [
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
" We thank the reviewer for their feedback and now explicitly emphasize the requirement of predetermined radii also in the abstract and the paragraphs on “Computational Overhead Reduction” and “Adaptive Sampling Ablation” in the experimental evaluation. Please see the updated version of the abstract in the pdf as t... | [
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
6,
8
] | [
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
4,
3
] | [
"csaZ4n6sNPI",
"Ecg1XNAJqT",
"iclr_2022_mHu2vIds_-b",
"REswunzFryH",
"-P5d9gq-HyY",
"enlF8rj8X3G",
"z_XscCMNetZ",
"iclr_2022_mHu2vIds_-b",
"REswunzFryH",
"iclr_2022_mHu2vIds_-b",
"iclr_2022_mHu2vIds_-b"
] |
iclr_2022_UYneFzXSJWh | Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution | When transferring a pretrained model to a downstream task, two popular methods are full fine-tuning (updating all the model parameters) and linear probing (updating only the last linear layer---the "head"). It is well known that fine-tuning leads to better accuracy in-distribution (ID). However, in this paper, we find ... | Accept (Oral) | The paper provides a solid and thorough analysis to the two basic methods of fine-tuning, linear probing (LP) and fine-tuning (FT). The authors provide an important and highly interesting observation about the performance of both in and out of domain (OOD) setting. They validate the known phenomena that FT outperforms ... | test | [
"To9Q8g8Nhrc",
"0RWPm6jwQjg",
"YX8sLU74N8b",
"q238Exw6CFS",
"Sxw9PS5pUIC",
"N4BWRziPsv2",
"lQ45kqaQBwm",
"H4ZGiZVZtvs",
"VjMCUUT9rZb",
"xpCZbP6G4HJ",
"sJLKID3ua5T",
"Yp7sx-Qg6oI",
"zmN6QoJpzE",
"vMJlWEMbTh",
"TkSngeYJfZ",
"UodZ1bm1nrc",
"qJmPwRdaKeE",
"lj49Uq9EJAM",
"eO4XEvdbuKS"... | [
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"This paper explores how different strategies for fine-tuning affect in- and out-of-distribution performance. The authors contrast linear probing (updating only the parameters of the final linear layer), end-to-end fine-tuning (updating all parameters of the model) and a two-stage approach, where linear probing is ... | [
8,
-1,
8,
-1,
-1,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8
] | [
4,
-1,
4,
-1,
-1,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3
] | [
"iclr_2022_UYneFzXSJWh",
"Sxw9PS5pUIC",
"iclr_2022_UYneFzXSJWh",
"To9Q8g8Nhrc",
"iclr_2022_UYneFzXSJWh",
"To9Q8g8Nhrc",
"VjMCUUT9rZb",
"iclr_2022_UYneFzXSJWh",
"xpCZbP6G4HJ",
"sJLKID3ua5T",
"vMJlWEMbTh",
"To9Q8g8Nhrc",
"To9Q8g8Nhrc",
"UodZ1bm1nrc",
"To9Q8g8Nhrc",
"H4ZGiZVZtvs",
"H4ZG... |
iclr_2022_NMEceG4v69Y | CycleMLP: A MLP-like Architecture for Dense Prediction | This paper presents a simple MLP-like architecture, CycleMLP, which is a versatile backbone for visual recognition and dense predictions. As compared to modern MLP architectures, e.g. , MLP-Mixer, ResMLP, and gMLP, whose architectures are correlated to image size and thus are infeasible in object detection and segmenta... | Accept (Oral) | The authors propose a new MLP-Mixer-like architecture called Cycle MLP which has two main advantages with respect to MLP-Mixer: (i) it’s applicable to varying input image sizes, and (ii) linear computational complexity. The authors present competitive results on image classification, object detection and segmentation.
... | train | [
"Q11by1Dmi8m",
"wy-kmTDcSqk",
"IwLKiPCV4RX",
"RJKfiDbE-S0"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"The paper proposes Cycle MLP architecture, the idea is to bring spatial context into Channel FC and increase its receptive field. The main objective of the paper is to address the challenges faced by the current MLP-Architectures. Cycle MLP allows flexible image resolution and avoids quadratic computational compl... | [
8,
6,
8,
8
] | [
4,
4,
3,
3
] | [
"iclr_2022_NMEceG4v69Y",
"iclr_2022_NMEceG4v69Y",
"iclr_2022_NMEceG4v69Y",
"iclr_2022_NMEceG4v69Y"
] |
iclr_2022_wRODLDHaAiW | iLQR-VAE : control-based learning of input-driven dynamics with applications to neural data | Understanding how neural dynamics give rise to behaviour is one of the most fundamental questions in systems neuroscience. To achieve this, a common approach is to record neural populations in behaving animals, and model these data as emanating from a latent dynamical system whose state trajectories can then be related... | Accept (Oral) | The paper introduces a novel control-based variational inference approach that learns latent dynamics in an *input-driven* state-space model. An optimal control solution (iLQR) is implicitly used as the recognition model which is fast and compact. Reviewers unanimously agree on the high quality writing and high signifi... | train | [
"bZSyAe4_hug",
"Xfk41Zsi_Po",
"0GSrgb07970"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"This paper presents a new approach for inference in a model that simultaneously provides latent dynamics, initial conditions, and - importantly - external inputs. This approach is enabled by using the outcome of an optimization algorithm (iLQR) in the recognition model, recently enabled by other work in the field.... | [
8,
8,
8
] | [
3,
4,
4
] | [
"iclr_2022_wRODLDHaAiW",
"iclr_2022_wRODLDHaAiW",
"iclr_2022_wRODLDHaAiW"
] |
iclr_2022_SYB4WrJql1n | On the Existence of Universal Lottery Tickets | The lottery ticket hypothesis conjectures the existence of sparse subnetworks of large randomly initialized deep neural networks that can be successfully trained in isolation. Recent work has experimentally observed that some of these tickets can be practically reused across a variety of tasks, hinting at some form of ... | Accept (Poster) | Dear Authors,
The paper was received nicely and discussed during the rebuttal period. The current consensus suggests the paper be accepted, but could have another round of revisions before it gets published
- Definition of sparsity within theoretical results + clarity of results. This seems to be the main concern by ... | train | [
"BOhUTBn2NPV",
"zHl4Z2ZF0cy",
"JiCb7JA604Z",
"UDsXdJW2wzv",
"eUY5dWJLM-",
"9_2yPgaE6R",
"pWrBT7scFUS",
"rUYD_u1qWkq",
"4JqzKv5tvo0",
"gzo4R7uEuJa",
"3KfWFltV0AA",
"tOLF8wFsvxA"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"This paper develops a theory to explain the lottery ticket hypothesis based on extensions of subset-sum results and a strategy to leverage higher amounts of depth. The developed proof could be of independent interest since it leverages the power of depth and improves the existing bounds. Overall this paper is tech... | [
6,
-1,
6,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
8
] | [
3,
-1,
2,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
4
] | [
"iclr_2022_SYB4WrJql1n",
"4JqzKv5tvo0",
"iclr_2022_SYB4WrJql1n",
"3KfWFltV0AA",
"gzo4R7uEuJa",
"iclr_2022_SYB4WrJql1n",
"rUYD_u1qWkq",
"9_2yPgaE6R",
"BOhUTBn2NPV",
"tOLF8wFsvxA",
"JiCb7JA604Z",
"iclr_2022_SYB4WrJql1n"
] |
iclr_2022_XEW8CQgArno | Training invariances and the low-rank phenomenon: beyond linear networks | The implicit bias induced by the training of neural networks has become a topic of rigorous study. In the limit of gradient flow and gradient descent with appropriate step size, it has been shown that when one trains a deep linear network with logistic or exponential loss on linearly separable data, the weights converg... | Accept (Poster) | *Summary:* Low-rank bias in nonlinear architectures.
*Strengths:*
- Significant theoretical contribution.
- Well written; detailed sketch of proofs.
*Weaknesses:*
- More intuitions desired.
- Restrictive assumptions.
*Discussion:*
Authors made efforts to improve the discussion in response to 6P7z. Authors ... | train | [
"wWEvImuSAqX",
"9F0Mex2Vy1a",
"1WiWMitLghh",
"qzYnqApkPwB",
"gsbFIplfSkF",
"8xXqxtvFAp",
"5La3sUD9tBz",
"NGVNOVcbpVH",
"KoVbW4rI2TL",
"9yZfV0623TM",
"mCiYpzTLatZ"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Thank you for your answer and the edits you made on the revision. I maintain my score. I think this paper should be accepted. ",
" The authors' response has addressed my concerns. I recommend accepting this submission.",
"This paper studies the invariance in training nonlinear networks under gradient flow or ... | [
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
8,
8,
6
] | [
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
3,
4,
3
] | [
"5La3sUD9tBz",
"mCiYpzTLatZ",
"iclr_2022_XEW8CQgArno",
"1WiWMitLghh",
"mCiYpzTLatZ",
"9yZfV0623TM",
"KoVbW4rI2TL",
"iclr_2022_XEW8CQgArno",
"iclr_2022_XEW8CQgArno",
"iclr_2022_XEW8CQgArno",
"iclr_2022_XEW8CQgArno"
] |
iclr_2022_wIzUeM3TAU | Expressiveness and Approximation Properties of Graph Neural Networks | Characterizing the separation power of graph neural networks (GNNs) provides an understanding of their limitations for graph learning tasks. Results regarding separation power are, however, usually geared at specific GNNs architectures, and tools for understanding arbitrary GNN architectures are generally lacking. We p... | Accept (Oral) | This paper gives a new theoretical framework to characterize the expressive power of graph neural networks that describes GNN by tensor language (TL) and then makes it possible to analyze its expressive power through the lens of TL. The authors connect the expressive ability of TL to the color refinement algorithms and... | train | [
"iCCdllnwbxL",
"Vh_qX3Hyrxe",
"fD4Jm071vJ_",
"xZBuc9dbeNY",
"AOHt6gYbwpC",
"fkGmmCazB58",
"jOkgQQzyH7H",
"984h6pu_cnf",
"O5Dd5qfcOEt",
"Rq2gIDsaabt",
"fDtiNO8OlR",
"-B6qcevcriD",
"pqrmFnkST1",
"e6OX9NGMxaV",
"Po6BUm7MUlV",
"WUcKgv1wOXB",
"Z7m2F0fav8T",
"qBPgj2-EJ51",
"q7BFQGBxEe"... | [
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
... | [
" We thank the reviewer for the very positive assessment of the original submission and the revision.",
" I raised my score to 10.",
"This paper introduces a new approach to study the separation power and approximation properties of graph neural networks (GNN). The authors introduce the Tensor Languages (TL) an... | [
-1,
-1,
10,
8,
-1,
-1,
-1,
8,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
-1,
-1,
4,
2,
-1,
-1,
-1,
4,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
"Vh_qX3Hyrxe",
"K53Ak0aNZ9u",
"iclr_2022_wIzUeM3TAU",
"iclr_2022_wIzUeM3TAU",
"O5Dd5qfcOEt",
"Rq2gIDsaabt",
"-B6qcevcriD",
"iclr_2022_wIzUeM3TAU",
"984h6pu_cnf",
"nWcXIcvmRR",
"iclr_2022_wIzUeM3TAU",
"fDtiNO8OlR",
"fDtiNO8OlR",
"fDtiNO8OlR",
"fDtiNO8OlR",
"984h6pu_cnf",
"fDtiNO8OlR",... |
iclr_2022__F9xpOrqyX9 | Spread Spurious Attribute: Improving Worst-group Accuracy with Spurious Attribute Estimation | The paradigm of worst-group loss minimization has shown its promise in avoiding to learn spurious correlations, but requires costly additional supervision on spurious attributes. To resolve this, recent works focus on developing weaker forms of supervision---e.g., hyperparameters discovered with a small number of valid... | Accept (Poster) | This paper presents a new method to decrease the supervision cost for learning spurious attributes using worst-group loss minimization. Their method uses samples both with and without spurious attribute annotations to train a model to predict the spurious attribute, then use the pseudo-attribute predicted by the traine... | train | [
"jKva2H9L2z",
"S-UWMBPCAyN",
"GYrJwMbkgqW",
"wD5DaN3Xb6q",
"nkvXs5aTEKm",
"DHlKtjXZ8j8",
"XyapwapD-KF",
"XtbHRvZxx-B",
"--e-tb0GzW8",
"ZzuyviJwIb",
"lD8BPt6Jpxi",
"_OR6J8tyEzs",
"DGAIsIZTbuI",
"jbs3cmIXmMG",
"svi4Rpw3d7i",
"nFW32Nenan0"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"This paper presents a technique, spread spurious attribute, for inferring the group annotation for training samples in a dataset. The inferred group information is then used as part of a group DRO minimization scheme or some other worst case scheme. The key insight in this work is to use a small validation set (1-... | [
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
6,
6
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
5,
4
] | [
"iclr_2022__F9xpOrqyX9",
"XtbHRvZxx-B",
"_OR6J8tyEzs",
"iclr_2022__F9xpOrqyX9",
"XyapwapD-KF",
"jbs3cmIXmMG",
"jbs3cmIXmMG",
"jKva2H9L2z",
"jKva2H9L2z",
"svi4Rpw3d7i",
"svi4Rpw3d7i",
"nFW32Nenan0",
"iclr_2022__F9xpOrqyX9",
"iclr_2022__F9xpOrqyX9",
"iclr_2022__F9xpOrqyX9",
"iclr_2022__F... |
iclr_2022_v6s3HVjPerv | Do Users Benefit From Interpretable Vision? A User Study, Baseline, And Dataset | A variety of methods exist to explain image classification models. However, whether they provide any benefit to users over simply comparing various inputs and the model’s respective predictions remains unclear. We conducted a user study (N=240) to test how such a baseline explanation technique performs against concept-... | Accept (Poster) | This work presents a novel and clever experiment for interpretable vision. Reviewers all agreed that it tackles an important and interesting research question via a user study design. There are some concerns around the generalization and transfer to large-scale real-world settings, as well as dataset construction. Wi... | test | [
"HosDef9JO-s",
"L70wCKj_UC",
"aFnLeIh9spk",
"xpcPXx_Sc8m",
"Pm_APhp2OU",
"h0cFbwuzLWI",
"nu8Kh4_ziy7",
"qZUpUbSEsp",
"JaF-G-EHKip",
"v3pTsY1wZCp",
"-lTYuHC5TTV",
"MA3QU0ZPAjW",
"N2Wgud73KSh",
"NDxiW2thZ_9",
"H4_NVNinCb",
"ZoTNJtyd0LD",
"r3ryRnwNAj8"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"The paper introduces an experiment design and an approach to synthesizing the dataset for the experiment. The experiment asked the systems to classify whether the shape of the \"animal\" is Peaky or Stretchy. The advantages are that authors can controllably generate trainable examples under arbitrary biases of the... | [
6,
-1,
-1,
-1,
-1,
-1,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
8
] | [
3,
-1,
-1,
-1,
-1,
-1,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
3
] | [
"iclr_2022_v6s3HVjPerv",
"ZoTNJtyd0LD",
"nu8Kh4_ziy7",
"JaF-G-EHKip",
"v3pTsY1wZCp",
"qZUpUbSEsp",
"iclr_2022_v6s3HVjPerv",
"-lTYuHC5TTV",
"HosDef9JO-s",
"iclr_2022_v6s3HVjPerv",
"r3ryRnwNAj8",
"ZoTNJtyd0LD",
"ZoTNJtyd0LD",
"nu8Kh4_ziy7",
"nu8Kh4_ziy7",
"iclr_2022_v6s3HVjPerv",
"iclr... |
iclr_2022_bVvMOtLMiw | DIVA: Dataset Derivative of a Learning Task | We present a method to compute the derivative of a learning task with respect to a dataset. A learning task is a function from a training set to the validation error, which can be represented by a trained deep neural network (DNN). The ``dataset derivative'' is a linear operator, computed around the trained model, that... | Accept (Poster) | This paper has been independently reviewed by four expert reviewers. Two of them recommended straight acceptance, one of them assesses this work as marginally acceptable after increasing their score as a result of the author's rebuttal, and the last reviewer considers this paper marginally below the acceptance threshol... | train | [
"eAjhu7SFulc",
"lJIIk-yh-h",
"Zj0qeIQGQMr",
"X2wdbZRNaSG",
"erDXvLy_6fh",
"OP75bo87wEs",
"OfMgg1_LxW",
"auwuiOXxMdF",
"WASIgR9NjKn",
"PeWmDlrrG_0"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"This paper proposes a method for dataset optimization that learns sample weights without a separate validation dataset.\nBy using squared loss and linearization around a pre-trained model, derivative w.r.t. the sample weights can be obtained as closed-form, which realizes efficient end-to-end learning. In addition... | [
6,
-1,
-1,
-1,
-1,
-1,
-1,
5,
8,
8
] | [
3,
-1,
-1,
-1,
-1,
-1,
-1,
5,
3,
4
] | [
"iclr_2022_bVvMOtLMiw",
"eAjhu7SFulc",
"X2wdbZRNaSG",
"PeWmDlrrG_0",
"WASIgR9NjKn",
"OfMgg1_LxW",
"auwuiOXxMdF",
"iclr_2022_bVvMOtLMiw",
"iclr_2022_bVvMOtLMiw",
"iclr_2022_bVvMOtLMiw"
] |
iclr_2022_5hLP5JY9S2d | Open-Set Recognition: A Good Closed-Set Classifier is All You Need | The ability to identify whether or not a test sample belongs to one of the semantic classes in a classifier's training set is critical to practical deployment of the model. This task is termed open-set recognition (OSR) and has received significant attention in recent years. In this paper, we first demonstrate that the... | Accept (Oral) | This paper provides well-written and thorough analysis demonstrating that closed-set recognition performance correlates with open-set recognition performance, and that simply making the close-set model strong via augmentation, label smoothing, etc. along with small scoring changes (using logits rather than softmax prob... | train | [
"8chiiD0bPLj",
"Gdiik6ZH7qJ",
"YVyQ-6ARiZF",
"SawE1zw8SIa",
"hYZxL0E65s",
"AD_lXh80jCn",
"0WBVx_fJu8k",
"Soqu0We-zw1",
"nk2YOUSrfDu",
"ycNTayHghfR",
"sohU-Ri_CS_",
"-ehYDB8ZFT5"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
" The updates by the authors addressed my concerns well, also consider the comments from other reviewers, I will keep my previous rating of this paper and suggest to accept this paper. ",
" I read the other reviews and I think there is a clear agreement that the paper is well written, clarify some important point... | [
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
8
] | [
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4
] | [
"0WBVx_fJu8k",
"Soqu0We-zw1",
"iclr_2022_5hLP5JY9S2d",
"hYZxL0E65s",
"YVyQ-6ARiZF",
"-ehYDB8ZFT5",
"sohU-Ri_CS_",
"-ehYDB8ZFT5",
"YVyQ-6ARiZF",
"iclr_2022_5hLP5JY9S2d",
"iclr_2022_5hLP5JY9S2d",
"iclr_2022_5hLP5JY9S2d"
] |
iclr_2022_iulEMLYh1uR | The Efficiency Misnomer | Model efficiency is a critical aspect of developing and deploying machine learning models.
Inference time and latency directly affect the user experience, and some applications have hard requirements. In addition to inference costs, model training also have direct financial and environmental impacts.
Although there ar... | Accept (Poster) | Overall the reviewers like the ideas in this paper. It calls out some of the issues with the current line of thinking in the ML/AI community. There were some concerns, but overall this paper offers a new way to think about, present, and question efficiency results. This could be quite infulential. I think this is i... | train | [
"6-BY7hiGIge",
"2W9DubYl98K",
"aZcHcLgtV6T",
"Iep_2fRgsU2",
"5lLDIyRnRTG",
"qK-NZMG3GXR",
"B49wGw261Bf",
"TnaaupxcPK",
"15theIgx6d0",
"MmyfpOydOuF",
"0ZsGq3qhCjK",
"TVVUZKl2io",
"sy583h23H5y",
"_iRHg_HY-zG",
"Z5TSR8TKuBX",
"KHmLhAgjJ2T",
"zNQFsGx7dw9"
] | [
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
" We thank the reviewer for the additional comment and for the discussion.\n\nWe agree with the reviewer that the existence of such guidelines can be helpful. We described the difficulty of having such guidelines in [our response to reviewer sinW](https://openreview.net/forum?id=iulEMLYh1uR¬eId=2W9DubYl98K) and ... | [
-1,
-1,
-1,
-1,
5,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5,
6
] | [
-1,
-1,
-1,
-1,
4,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4
] | [
"aZcHcLgtV6T",
"Iep_2fRgsU2",
"sy583h23H5y",
"0ZsGq3qhCjK",
"iclr_2022_iulEMLYh1uR",
"B49wGw261Bf",
"15theIgx6d0",
"iclr_2022_iulEMLYh1uR",
"_iRHg_HY-zG",
"TVVUZKl2io",
"5lLDIyRnRTG",
"zNQFsGx7dw9",
"KHmLhAgjJ2T",
"TnaaupxcPK",
"iclr_2022_iulEMLYh1uR",
"iclr_2022_iulEMLYh1uR",
"iclr_... |
iclr_2022_KJztlfGPdwW | Rethinking Goal-Conditioned Supervised Learning and Its Connection to Offline RL | Solving goal-conditioned tasks with sparse rewards using self-supervised learning is promising because of its simplicity and stability over current reinforcement learning (RL) algorithms. A recent work, called Goal-Conditioned Supervised Learning (GCSL), provides a new learning framework by iteratively relabeling and i... | Accept (Poster) | The authors introduce a method that improves goal-conditioned supervised learning (GCSL) by iteratively re-weighting the experience by a variable that correlates with the number of steps till the desired goal. The reviewers mention that the authors focus on an important problem, their method is simple and the empirical... | train | [
"9TSKcg-NUyw",
"tqiOWlHkebF",
"r1EqFF41ZbI",
"deQ7JHGt2pJ",
"e2-ajw-cqEP",
"ShnRwImEmq",
"ihqSminl_l",
"RcI21r2wN_L",
"zYGXlh6SgCF",
"5so9iqYpQEM",
"dnkVe09kyW9",
"ZR8_9IgO-Zj",
"GGKGbHoDID",
"4ux7q5Hv_Ff",
"nBAvOQ-LGNE",
"VszUrSfPPR",
"Ix5R-31SLvQ",
"DSCs6ZswUVx",
"1nFMDIAmDDt",... | [
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_re... | [
" Thanks for your feedback. We are pleased that our response has addressed some of your concerns. Here we provide further clarification. \n\n\n**Q1:** A1 -- The response here doesn't change my mind, as it simply states information that was already discussed in other threads. (To be fair, the only way I can think I... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
5,
8
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
3,
3
] | [
"tqiOWlHkebF",
"deQ7JHGt2pJ",
"e2-ajw-cqEP",
"e2-ajw-cqEP",
"ShnRwImEmq",
"DlWKGI7UkV6",
"RcI21r2wN_L",
"GGKGbHoDID",
"5so9iqYpQEM",
"VszUrSfPPR",
"4ux7q5Hv_Ff",
"1nFMDIAmDDt",
"Ix5R-31SLvQ",
"nBAvOQ-LGNE",
"DSCs6ZswUVx",
"CmZ2ILYYHiv",
"F_VjoHYy3-o",
"axN-0LaUcfi",
"F_VjoHYy3-o"... |
iclr_2022_31d5RLCUuXC | A Tale of Two Flows: Cooperative Learning of Langevin Flow and Normalizing Flow Toward Energy-Based Model | This paper studies the cooperative learning of two generative flow models, in which the two models are iteratively updated based on the jointly synthesized examples. The first flow model is a normalizing flow that transforms an initial simple density into a target density by applying a sequence of invertible transforma... | Accept (Poster) | The paper proposes a methodological improvement in the Langevin-based training of energy-based models. The idea is to initialize the Langevin flow used to train an energy-based model with a normalizing flow which learns to mimic the Langevin flow as the energy-based model is being trained. The method is empirically eva... | train | [
"FECLXtpcHd7",
"jdiB-P1SlcE",
"J6vJVrneIrK",
"4GgUkv5FLJt",
"WzLHxJ5x7d1",
"szRoStDVSE",
"cjerQr3Yk9I",
"nPyEjI4vWh",
"fI83kijNCZ4",
"OZjWAjHAfqO",
"dn-lCy_-LA6",
"7_qz9B7J957",
"Rux8xCH9Cu",
"ecS02bSrKpM",
"jlKMxWjVNj1",
"phN2khkgq5k",
"HtS2CsOChip",
"NSUNMvvY5lt",
"zeYUQa6v793"... | [
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
... | [
" Dear bAHo,\n\nPlease kindly let us know if you still need further clarification. We will be more than happy to answer.\n\nThe major contribution of our paper is to study the interaction between a short-run stochastic Langevin flow and a deterministic normalizing flow in the cooperative learning framework (existin... | [
-1,
-1,
-1,
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,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
6,
8
] | [
-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,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
4,
3
] | [
"zeYUQa6v793",
"fI83kijNCZ4",
"jdiB-P1SlcE",
"iclr_2022_31d5RLCUuXC",
"szRoStDVSE",
"jlKMxWjVNj1",
"ecS02bSrKpM",
"zeYUQa6v793",
"7_qz9B7J957",
"7_qz9B7J957",
"7_qz9B7J957",
"HtS2CsOChip",
"ecS02bSrKpM",
"7b5UVrwcHys",
"xR0BwH8hG0",
"7b5UVrwcHys",
"4GgUkv5FLJt",
"zeYUQa6v793",
"V... |
iclr_2022_avgclFZ221l | Asymmetry Learning for Counterfactually-invariant Classification in OOD Tasks | Generalizing from observed to new related environments (out-of-distribution) is central to the reliability of classifiers. However, most classifiers fail to predict label $Y$ from input $X$ when the change in environment is due a (stochastic) input transformation $T^\text{te} \circ X'$ not observed in training, as in t... | Accept (Oral) | This paper proposes asymmetry learning for learning counterfactual classifiers, i.e. classifiers which are invariant to certain symmetry transformations w.r.t. hidden variables that differ between the training and test sets.
The reviewers universally agreed that the proposed setting, and theoretical contribution, were... | train | [
"dYOi24Cyt9w",
"tSgDs6DEBo7",
"0PqK1WvzhJ",
"f_ZH3wrMwNR",
"f7h8v_NE9yy",
"b8yNWLT3LiL",
"kA9upaJn2z",
"Y0AoGFAz0Ms",
"ONZXQh6tHY2",
"R93g3P56co",
"BFKZZcCQb9c",
"B0deQn6Gtj"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author"
] | [
"The authors propose an approach for constructing classifiers that achieve out-of-distribution (OOD) generalization using a new learning paradigm they call _asymmetry learning_. They consider OOD tasks where the test input is obtained from the training input by applying a sequence of (random) input transformations.... | [
8,
8,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
3,
2,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
"iclr_2022_avgclFZ221l",
"iclr_2022_avgclFZ221l",
"iclr_2022_avgclFZ221l",
"b8yNWLT3LiL",
"ONZXQh6tHY2",
"R93g3P56co",
"B0deQn6Gtj",
"tSgDs6DEBo7",
"tSgDs6DEBo7",
"dYOi24Cyt9w",
"0PqK1WvzhJ",
"0PqK1WvzhJ"
] |
iclr_2022_dwg5rXg1WS_ | ViTGAN: Training GANs with Vision Transformers | Recently, Vision Transformers (ViTs) have shown competitive performance on image recognition while requiring less vision-specific inductive biases. In this paper, we investigate if such performance can be extended to image generation. To this end, we integrate the ViT architecture into generative adversarial networks (... | Accept (Spotlight) | The paper proposes a GAN architecture with a ViT-based discriminator and a ViT-based generator. The paper initially received a mixed rating with two "slightly above the acceptance threshold" ratings and "three slightly below the acceptance threshold" ratings. Several concerns were raised in the reviews, including wheth... | train | [
"nSybb1CFO7R",
"okMyeo-Yz0",
"j_nlRxVP67d",
"6F625tyW4i",
"rWKqDonp24"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
"In this paper, the authors study employing ViTs for GAN-based image generation. For discriminators, they propose improved methods to enforce Lipschitz constraints and spectral norm regularization, and modified architecture; for generators, they adopted self-modulated LayerNorm and Fourier features to improve the p... | [
6,
8,
6,
6,
6
] | [
4,
5,
4,
4,
4
] | [
"iclr_2022_dwg5rXg1WS_",
"iclr_2022_dwg5rXg1WS_",
"iclr_2022_dwg5rXg1WS_",
"iclr_2022_dwg5rXg1WS_",
"iclr_2022_dwg5rXg1WS_"
] |
iclr_2022_2t7CkQXNpuq | ToM2C: Target-oriented Multi-agent Communication and Cooperation with Theory of Mind | Being able to predict the mental states of others is a key factor to effective social interaction. It is also crucial for distributed multi-agent systems, where agents are required to communicate and cooperate. In this paper, we introduce such an important social-cognitive skill, i.e. Theory of Mind (ToM), to build soc... | Accept (Poster) | The current paper presents a new method for communication and cooperation in multi-agent settings. Specifically, the authors propose to model other agents' intentions and internal states using ToM nets and using these predictions to then decide how to communicate/coordinate. The authors present experiments in two multi... | test | [
"dkGNNvnFPp",
"h5Z2wmfHV5",
"BqIzW-Mpzg7",
"STDfl40i4iZ",
"ByljMoI51jV",
"Pevq1utFTMX",
"pPrW4w51bN6",
"YtESwrN7n2O",
"YN1bn6j2dv1",
"NCFhJto22qh",
"HiN7na2VKtE",
"N03NH422AzR"
] | [
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
" Thanks for your reply and suggestions! To address your concern about the environment setting, we conducted additional validation by adding a restriction on the access of agent poses, i.e, agents can only access the poses of others within an observable distance. In this case, the agent is also partially observable... | [
-1,
-1,
6,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
6
] | [
-1,
-1,
3,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
4
] | [
"h5Z2wmfHV5",
"HiN7na2VKtE",
"iclr_2022_2t7CkQXNpuq",
"YtESwrN7n2O",
"iclr_2022_2t7CkQXNpuq",
"ByljMoI51jV",
"iclr_2022_2t7CkQXNpuq",
"BqIzW-Mpzg7",
"ByljMoI51jV",
"ByljMoI51jV",
"N03NH422AzR",
"iclr_2022_2t7CkQXNpuq"
] |
iclr_2022_Qycd9j5Qp9J | Understanding the Variance Collapse of SVGD in High Dimensions | Stein variational gradient descent (SVGD) is a deterministic inference algorithm that evolves a set of particles to fit a target distribution. Despite its computational efficiency, SVGD often underestimates the variance of the target distribution in high dimensions. In this work we attempt to explain the variance colla... | Accept (Poster) | This paper proposes new analysis on Variance Collapse of SVGD in High Dimensions. The analysis provides some new insights despite of some limitations. | train | [
"BOfDs3Hk9V",
"HCbubF9pqEy",
"uk7X6hrfGYn",
"Uaa-_wH0-sb",
"u4cYljWc8CU",
"JIwVx1aMufK",
"59OjK2PANdO",
"njLNB1ud_pe",
"JjU9VLqM2iu"
] | [
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Thank you for the valuable comments, which helped us improve the manuscript. We address the technical points below. \n\n**1. A more thorough investigation of the \"fixed SVGD\".**\n\nThank you for the suggestion. We have made the following updates to the manuscript. \n- In Appendix A.3 we included an additional B... | [
-1,
-1,
-1,
-1,
-1,
8,
6,
6,
6
] | [
-1,
-1,
-1,
-1,
-1,
4,
3,
3,
4
] | [
"59OjK2PANdO",
"JIwVx1aMufK",
"iclr_2022_Qycd9j5Qp9J",
"JjU9VLqM2iu",
"njLNB1ud_pe",
"iclr_2022_Qycd9j5Qp9J",
"iclr_2022_Qycd9j5Qp9J",
"iclr_2022_Qycd9j5Qp9J",
"iclr_2022_Qycd9j5Qp9J"
] |
iclr_2022_PLDOnFoVm4 | Reinforcement Learning under a Multi-agent Predictive State Representation Model: Method and Theory | This paper proposes a new algorithm for learning the optimal policies under a novel multi-agent predictive state representation reinforcement learning model. Compared to the state-of-the-art methods, the most striking feature of our approach is the introduction of a dynamic interaction graph to the model, which allows ... | Accept (Spotlight) | This paper presents an extension of the Predictive State Representation (PSRs) to multi-agent systems, with a dynamic interaction graph represents each agent’s predictive state based on its “neighborhood” agents. Three types of agent networks are considered: static complete graphs (all agents affect all others experien... | train | [
"4T8S7MUbbFn",
"X2ktntlzBYG",
"aHoNZUHn5c",
"bUKaRDLMgBh",
"Q5FhWNMPSMa",
"c0zoFBZ-sty",
"BmTHj8FUyVS",
"fO-InGptG9H",
"As0eDmLVSd_",
"LK491cyaR4r",
"eHZI07HDbvx"
] | [
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" As I stated in the original review, I believe the paper to be of high quality. The changes made by the authors in light of the reviewer comments seem sensible and strengthen the paper further. I continue to champion this paper for acceptance.",
" We thank all reviewers for their recognition of our work. All the... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
8,
8
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
3,
3
] | [
"BmTHj8FUyVS",
"iclr_2022_PLDOnFoVm4",
"As0eDmLVSd_",
"aHoNZUHn5c",
"eHZI07HDbvx",
"LK491cyaR4r",
"fO-InGptG9H",
"c0zoFBZ-sty",
"iclr_2022_PLDOnFoVm4",
"iclr_2022_PLDOnFoVm4",
"iclr_2022_PLDOnFoVm4"
] |
iclr_2022_64trBbOhdGU | HyAR: Addressing Discrete-Continuous Action Reinforcement Learning via Hybrid Action Representation | Discrete-continuous hybrid action space is a natural setting in many practical problems, such as robot control and game AI. However, most previous Reinforcement Learning (RL) works only demonstrate the success in controlling with either discrete or continuous action space, while seldom take into account the hybrid acti... | Accept (Poster) | This paper proposes a new approach to solve mixed discrete-continuous action RL problems, based on embedding actions into a latent space so that standard continuous control algorithms (like TD3) can be applied. Experiments over standard discrete-continuous benchmarks demonstrate the superiority of the proposed approach... | val | [
"ZBlSnNH0WN3",
"KoWVKbkfBZ9",
"w9tzUXNXB5A",
"xd2tIOPOqBn",
"ftl7NB2cIcx",
"VlhN3uDQXF",
"UqitAYm64Pk",
"ypMKKTeGIgo",
"-44lEdcQrU-",
"MO1jAOTalfy",
"4kC5EixIH42",
"BmIgJJ2MmFk",
"YBB3LQ_W11",
"voNvOVsdb3",
"t05EYjC_0LM",
"DZrCxzx9wq"
] | [
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
"The authors proposed a novel framework for hybrid (discrete and continuous) action RL, called Hybrid Action Representation (HyAR). The main idea is to take advantage of representation learning in Deep RL to encode hybrid action in a conbtinuous latent vector space. Then any RL algorithm for continuous action space... | [
8,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
8
] | [
4,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
1
] | [
"iclr_2022_64trBbOhdGU",
"iclr_2022_64trBbOhdGU",
"iclr_2022_64trBbOhdGU",
"ypMKKTeGIgo",
"UqitAYm64Pk",
"4kC5EixIH42",
"ZBlSnNH0WN3",
"w9tzUXNXB5A",
"DZrCxzx9wq",
"YBB3LQ_W11",
"iclr_2022_64trBbOhdGU",
"iclr_2022_64trBbOhdGU",
"4kC5EixIH42",
"4kC5EixIH42",
"4kC5EixIH42",
"iclr_2022_64... |
iclr_2022_MljXVdp4A3N | Know Your Action Set: Learning Action Relations for Reinforcement Learning | Intelligent agents can solve tasks in various ways depending on their available set of actions. However, conventional reinforcement learning (RL) assumes a fixed action set. This work asserts that tasks with varying action sets require reasoning of the relations between the available actions. For instance, taking a nai... | Accept (Poster) | This paper studies the problem of how to train an agent to understand relationships and dependencies among available (and potentially changing) actions in an RL environment to more efficiently solve a task. For instance, in the absence of a hammer for the task of putting up a painting on a wall, the agent could use an ... | train | [
"OiwdreT1qU",
"U7KKjXtW94x",
"uCVWdn2rpd7",
"bJBjaPGL4g",
"Zcb6toqcHzI",
"GK0pDT47Y84",
"XSy0NiAbDan",
"IsD9Cq20vAT",
"IqDi5kc9_6n"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author"
] | [
"This paper considers the reinforcement learning problem where the action space is fixed. Having a variable action space makes it necessary to learn interdependence between different actions (use of some actions might depend on the existence of other complementary actions), which is modelled via a graph attention n... | [
6,
8,
8,
8,
-1,
-1,
-1,
-1,
-1
] | [
3,
3,
4,
5,
-1,
-1,
-1,
-1,
-1
] | [
"iclr_2022_MljXVdp4A3N",
"iclr_2022_MljXVdp4A3N",
"iclr_2022_MljXVdp4A3N",
"iclr_2022_MljXVdp4A3N",
"bJBjaPGL4g",
"OiwdreT1qU",
"uCVWdn2rpd7",
"U7KKjXtW94x",
"iclr_2022_MljXVdp4A3N"
] |
iclr_2022_kj0_45Y4r9i | Discriminative Similarity for Data Clustering | Similarity-based clustering methods separate data into clusters according to the pairwise similarity between the data, and the pairwise similarity is crucial for their performance. In this paper, we propose {\em Clustering by Discriminative Similarity (CDS)}, a novel method which learns discriminative similarity for d... | Accept (Poster) | The authors provide a framework for unsupervised clarification based on minimizing a between-cluster discriminative similarity. It is more flexible than existing methods whose kernel similarity implicitly assumes uniform weights, and the authors connect to ideas such as max-margin and weighted kernel approaches. This y... | train | [
"lOZQyoh1E_r",
"l0G3XaajwDZ",
"tI35xP6JpN3",
"A2iCNXwhvKE",
"-UzAGTUwLu",
"YfKT7y3h0N",
"mB8_iPxD16_",
"qa2MRyTpOi3",
"GH2Ku5SDxwe",
"R2Q7iNjN7zd"
] | [
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" **Significance** (Cont'd) The discriminative Bayesian nonparametric clustering (ref. [C]) and BMMC (ref. [B]) require extra efforts of sampling hidden variables and tuning hyperparameters to generate the desirable number of clusters (or model selection), which could reduce the effect of discriminative measures us... | [
-1,
-1,
-1,
-1,
-1,
-1,
5,
8,
6,
6
] | [
-1,
-1,
-1,
-1,
-1,
-1,
3,
5,
4,
3
] | [
"l0G3XaajwDZ",
"GH2Ku5SDxwe",
"R2Q7iNjN7zd",
"qa2MRyTpOi3",
"lOZQyoh1E_r",
"-UzAGTUwLu",
"iclr_2022_kj0_45Y4r9i",
"iclr_2022_kj0_45Y4r9i",
"iclr_2022_kj0_45Y4r9i",
"iclr_2022_kj0_45Y4r9i"
] |
iclr_2022_vJZ7dPIjip3 | Generalization of Neural Combinatorial Solvers Through the Lens of Adversarial Robustness | End-to-end (geometric) deep learning has seen first successes in approximating the solution of combinatorial optimization problems. However, generating data in the realm of NP-hard/-complete tasks brings practical and theoretical challenges, resulting in evaluation protocols that are too optimistic. Specifically, most ... | Accept (Poster) | The paper studies how neural combinatorial solvers can be susceptible to adversarial examples and what implications does this susceptibility have on the evaluation of neural solvers. Besides proposing some successful adversarial attacks, the authors provide a method for adversarial training and show its effectiveness o... | train | [
"gEZ7BIOoz6B",
"r4AEW5ojA-9",
"yPtrTpm8qzD",
"8zH9VFYk1c",
"qRAIa9DeVf",
"ien1Syp-gT4"
] | [
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"In this paper, the authors propose to evaluate and improve the robustness and generalization of neural combinatorial solvers with adversarial examples, that is, perturbed inputs that fool the neural network to generate outputs with high loss. The authors claim that their proposal reconciles the tension between the... | [
6,
-1,
-1,
-1,
8,
8
] | [
3,
-1,
-1,
-1,
2,
4
] | [
"iclr_2022_vJZ7dPIjip3",
"gEZ7BIOoz6B",
"qRAIa9DeVf",
"ien1Syp-gT4",
"iclr_2022_vJZ7dPIjip3",
"iclr_2022_vJZ7dPIjip3"
] |
iclr_2022_9wOQOgNe-w | Differentiable DAG Sampling | We propose a new differentiable probabilistic model over DAGs (DP-DAG). DP-DAG allows fast and differentiable DAG sampling suited to continuous optimization. To this end, DP-DAG samples a DAG by successively (1) sampling a linear ordering of the node and (2) sampling edges consistent with the sampled linear ordering. W... | Accept (Poster) | The authors proposed an algorithm for sampling DAGs that is suited for continuous optimization. The sampling algorithm has two main steps: In the first step, a causal order over the variables is selected. In the second step, edges are sampled based on the selected order. Moreover, based on this algorithm, they proposed... | train | [
"ZFt9djNbI1N",
"_-R5Qr8mNIq",
"ptiWZIjMhpY",
"iN-REaBE7yE",
"OCwsLmO0ffw",
"l-DV1eMFihU",
"Om9sLct9ZyQ",
"oU6su8I5uPv",
"fwXej6MiRT_",
"bfCiPUiR-8"
] | [
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Thanks a lot for your comment.\n\nWe would like to clarify the contributions of the work which is not limited to a sampling procedure for DAGs (see Sec. 1 paragraph 'contribution'). The first contribution is a **fast** and **differentiable** DAG sampling procedure (DP-DAG) which can be used during optimization. T... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
5,
5
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
4,
4
] | [
"_-R5Qr8mNIq",
"OCwsLmO0ffw",
"iN-REaBE7yE",
"Om9sLct9ZyQ",
"bfCiPUiR-8",
"oU6su8I5uPv",
"fwXej6MiRT_",
"iclr_2022_9wOQOgNe-w",
"iclr_2022_9wOQOgNe-w",
"iclr_2022_9wOQOgNe-w"
] |
iclr_2022_tV3N0DWMxCg | Natural Posterior Network: Deep Bayesian Predictive Uncertainty for Exponential Family Distributions | Uncertainty awareness is crucial to develop reliable machine learning models. In this work, we propose the Natural Posterior Network (NatPN) for fast and high-quality uncertainty estimation for any task where the target distribution belongs to the exponential family. Thus, NatPN finds application for both classificatio... | Accept (Spotlight) | This paper presents a method for producing higher quality uncertainty estimates by mapping the predictions from an arbitrary (e.g. deep learning) model to an exponential family distribution. This is achieved by using the model to map from the inputs to a low-dimensional latent space and then using a normalizing flow t... | train | [
"W6fMvBmWe-",
"sFMXznifODv",
"IdogXXdNkyW",
"hC8HeZ0XTMu",
"cISzmlNDhgM",
"U-c3KYV6QSy",
"6Pb8LspAzV",
"1f5EhwkRK21",
"pGtJ4zxnRln",
"KPnDRgX_5Gx",
"yWowMdkXTsB",
"HsMnFw0WA2P",
"gGPPkZ-IYBZ",
"FHtdmL5G8cU",
"z3J1QRmUd0A",
"2cPNSH3SXX",
"2WATqkLKOD_",
"sbrEmqbKgEB",
"6aroPW-Q7jr"... | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
... | [
"The paper targets the task of getting useful predictive uncertainty, as measured by in-distribution calibration and out-of-distribution detection capability. Towards this, the authors focus on the distributions from the exponential family, which allow for a closed-form posterior form whose parameters are predicted... | [
6,
8,
-1,
-1,
-1,
8,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6
] | [
4,
3,
-1,
-1,
-1,
3,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4
] | [
"iclr_2022_tV3N0DWMxCg",
"iclr_2022_tV3N0DWMxCg",
"FHtdmL5G8cU",
"nZBgYM_eyfO",
"6aroPW-Q7jr",
"iclr_2022_tV3N0DWMxCg",
"2WATqkLKOD_",
"iclr_2022_tV3N0DWMxCg",
"KPnDRgX_5Gx",
"W6fMvBmWe-",
"1f5EhwkRK21",
"U-c3KYV6QSy",
"MJPWTRIKHk",
"z3J1QRmUd0A",
"2cPNSH3SXX",
"sFMXznifODv",
"sbrEmq... |
iclr_2022_RxplU3vmBx | Looking Back on Learned Experiences For Class/task Incremental Learning | Classical deep neural networks are limited in their ability to learn from emerging streams of training data. When trained sequentially on new or evolving tasks, their performance degrades sharply, making them inappropriate in real-world use cases. Existing methods tackle it by either storing old data samples or only up... | Accept (Spotlight) | This paper presents a zero-shot incremental learning approach that does not store past samples for experience replay. The idea is novel and well motivated, and the paper is well written. Reviewers' comments were mainly about missing baselines, missing ablation studies, and clarifications about the proposed method. In ... | train | [
"YI5mRodxZPO",
"8NiJp8sCUL5",
"g4GX7sVUsFz",
"yhZ3Ipf_KUa",
"I2XFWctXwI",
"12Z97qVVmVz",
"toJptkP_Cnr",
"eBWBvt_mHZX",
"7Oxyxvp1K34",
"DnJExidUGWe",
"e-HjviMdUV9",
"vx3mU73OqA",
"TW0zAeY1cY3",
"GKC9Ljs2Eqg",
"dOznxrOVKFi",
"JGD4Zg-_BZp",
"_5WWPKus14C",
"yczwuj5DZZA",
"akc_H0kXUuA... | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"a... | [
"This paper proposes a zero-shot incremental learning approach that doesn't store past examples or metadata for experience reply. It synthesizes past experiences from the network parameters through a memory recovery paradigm. The proposed method doesn't rely on external memory or parallel networks and achieves ver... | [
6,
6,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
4,
3,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
"iclr_2022_RxplU3vmBx",
"iclr_2022_RxplU3vmBx",
"Biy6IRIOnP4",
"nQPckhsnSBy",
"iclr_2022_RxplU3vmBx",
"toJptkP_Cnr",
"eBWBvt_mHZX",
"7Oxyxvp1K34",
"DnJExidUGWe",
"e-HjviMdUV9",
"vx3mU73OqA",
"TW0zAeY1cY3",
"GKC9Ljs2Eqg",
"dOznxrOVKFi",
"_5WWPKus14C",
"I2XFWctXwI",
"JGD4Zg-_BZp",
"8... |
iclr_2022_Zr5W2LSRhD | Constructing Orthogonal Convolutions in an Explicit Manner | Convolutions with orthogonal input-output Jacobian matrix, i.e., orthogonal convolution, have recently attracted substantial attention. A convolution layer with an orthogonal Jacobian matrix is 1-Lipschitz in the 2-norm, making the output robust to the perturbation in input. Meanwhile, an orthogonal Jacobian matrix... | Accept (Poster) | The paper studies the problem of how to construct orthogonal convolutional layers. It is known that a convolution layer is orthogonal if and only if its filters are obtained by certain Fourier operations on an orthogonal matrix. Previous work proposes to learn this orthogonal matrix, parameterized either through Cayley... | train | [
"yY7AbYTEIva",
"ENmbTsrbmsj",
"C3HjUPS435I",
"GgetyQ4gXK",
"uf5jBdOmuri",
"Ssig1g39Fej",
"3Xn3gVei-06",
"WJiLkd4P44G",
"k97fbaksdut",
"4Yqiz6ZsJDV",
"dEocI10m2B",
"6dvFViAX58d",
"kCcgPkt2oev",
"VEWmZA7rY-",
"enk5Fngf8e",
"9FFZbZpBFnN",
"PEWH-vC1lJy",
"IZLibyJtGNP",
"rmBrF545Qoz",... | [
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
... | [
" Dear Reviewer mniG,\n\nThanks for further checking.\nAs we mentioned previously, if the feature size $n$ is not divisible by the kernel size $k$, we can add zero paddings to make the feature size divisible by $k$. To be specific, we need to pad $\\lceil n/k \\rceil *k - n $ rows/columns and the size of the featu... | [
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6
] | [
-1,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3
] | [
"C3HjUPS435I",
"iclr_2022_Zr5W2LSRhD",
"uf5jBdOmuri",
"3Xn3gVei-06",
"3Xn3gVei-06",
"3Xn3gVei-06",
"WJiLkd4P44G",
"dEocI10m2B",
"4Yqiz6ZsJDV",
"iclr_2022_Zr5W2LSRhD",
"6dvFViAX58d",
"eF1dNBiPhWn",
"ENmbTsrbmsj",
"9FFZbZpBFnN",
"iclr_2022_Zr5W2LSRhD",
"RGHLJfhqb02",
"9FFZbZpBFnN",
"... |
iclr_2022_4C93Qvn-tz | MCMC Should Mix: Learning Energy-Based Model with Flow-Based Backbone | Learning energy-based model (EBM) requires MCMC sampling of the learned model as an inner loop of the learning algorithm. However, MCMC sampling of EBMs in high-dimensional data space is generally not mixing, because the energy function, which is usually parametrized by deep network, is highly multi-modal in the data s... | Accept (Poster) | The authors set up a simple combination of an energy based model and a flow based model that corrects the flow based model with an energy based term. The merits of this relative only an energy based model is improved sampling to compute the gradient. The advantage over a only flow based model is that the kinds of trans... | test | [
"7fHO17ADV2S",
"bS2O58raxR",
"89g5IiB82E",
"ygi7yVk0yNk",
"cEOygtS-sRb",
"mDrmZwAPpL2",
"NgXSr2BK8H",
"CX7EUs2kttA",
"lnc63IrkXhF",
"paNGHvQdFmr",
"X1lndlZDSDW",
"RoIZ4pd43jr",
"syYq37SbhR",
"fWJX4wVcPl0",
"IyJ2BHTrhGi",
"nXMmXSRJei1",
"_Gncq44bXf0",
"lLvqLyUI5bC"
] | [
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"I find that this paper is a re-submission from NeurIPS2020, for which I acted as one of the reviewers. The content almost remains the same.\n\nThis paper studies the learning of a special class of EBMs, which is a correction or an exponential tilting of a flow-based model.\nAn interesting observation that the resu... | [
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
8,
8
] | [
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
4,
2
] | [
"iclr_2022_4C93Qvn-tz",
"89g5IiB82E",
"cEOygtS-sRb",
"mDrmZwAPpL2",
"lLvqLyUI5bC",
"nXMmXSRJei1",
"CX7EUs2kttA",
"lnc63IrkXhF",
"paNGHvQdFmr",
"fWJX4wVcPl0",
"syYq37SbhR",
"iclr_2022_4C93Qvn-tz",
"IyJ2BHTrhGi",
"7fHO17ADV2S",
"RoIZ4pd43jr",
"_Gncq44bXf0",
"iclr_2022_4C93Qvn-tz",
"i... |
iclr_2022_rw1mZl_ss3L | Concurrent Adversarial Learning for Large-Batch Training | Large-batch training has become a commonly used technique when training neural networks with a large number of GPU/TPU processors. As batch size increases, stochastic optimizers tend to converge to sharp local minima, leading to degraded test performance. Current methods usually use extensive data augmentation to incre... | Accept (Poster) | Thank you for your submission to ICLR. This paper presents a straightforward but reasonable approach to (slightly) improving the performance of large-batch training via adversarial training. The basic approach is to apply (small epsilon) adversarial training, shown to help performance in small-batch settings, but acc... | val | [
"U9RY_mxfaH",
"mMB0TzTfu2p",
"6hFCk4fJFMe",
"5IHzyVyfNGY",
"-fcwQyb-v9w",
"XiT8Ds2q-X9",
"ytYmG-4q5uq",
"O_WF9i-BzeT"
] | [
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Dear Reviewer M9ZZ,\n\nThanks for your valuable review and comments again.\n\nIf you have any other questions, please feel free to ask us.",
" Dear Reviewer nTet,\n\nThanks for your constructive review and comments again.\n\nIf you have any other questions, please feel free to ask us.",
" Thanks for your cons... | [
-1,
-1,
-1,
-1,
-1,
8,
5,
6
] | [
-1,
-1,
-1,
-1,
-1,
3,
4,
5
] | [
"6hFCk4fJFMe",
"5IHzyVyfNGY",
"O_WF9i-BzeT",
"ytYmG-4q5uq",
"XiT8Ds2q-X9",
"iclr_2022_rw1mZl_ss3L",
"iclr_2022_rw1mZl_ss3L",
"iclr_2022_rw1mZl_ss3L"
] |
iclr_2022_lpkGn3k2YdD | Learning Long-Term Reward Redistribution via Randomized Return Decomposition | Many practical applications of reinforcement learning require agents to learn from sparse and delayed rewards. It challenges the ability of agents to attribute their actions to future outcomes. In this paper, we consider the problem formulation of episodic reinforcement learning with trajectory feedback. It refers to a... | Accept (Spotlight) | Description of paper content:
The paper addresses the problem of credit assignment for delayed reward problems. Their method, Randomized Return Decomposition, learns a reward function that provides immediate reward. The algorithm works by randomly subsampling trajectories and predicting the empirical return by regress... | train | [
"h6LASJG6tdS",
"xbiqAbYZ8Fu",
"N_GuPK4gblE",
"vQLjTanBvI",
"HDeWnkujV2T",
"9b4JHj7P92h",
"lmy7_k1h_Br",
"AD5Q55qSQcu",
"9T1ZrIvXO-7",
"m9UsWDSIb_6",
"cDsLqlOVSVn",
"za0UQ8Te4u2",
"I1QG3e_P5C",
"gLQhT33DY3P"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
" This looks great and is convincing. Since you have promised to do this for all curves, I have updated my score to an 8. Thank you for your hard work.",
"The authors propose a method of constructing a proxy reward function (and the loss function for learning that reward function), which generalizes two prior m... | [
-1,
8,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
5
] | [
-1,
3,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
2,
4
] | [
"N_GuPK4gblE",
"iclr_2022_lpkGn3k2YdD",
"lmy7_k1h_Br",
"9b4JHj7P92h",
"iclr_2022_lpkGn3k2YdD",
"m9UsWDSIb_6",
"AD5Q55qSQcu",
"9T1ZrIvXO-7",
"xbiqAbYZ8Fu",
"gLQhT33DY3P",
"HDeWnkujV2T",
"I1QG3e_P5C",
"iclr_2022_lpkGn3k2YdD",
"iclr_2022_lpkGn3k2YdD"
] |
iclr_2022_FQOC5u-1egI | Handling Distribution Shifts on Graphs: An Invariance Perspective | There is increasing evidence suggesting neural networks' sensitivity to distribution shifts, so that research on out-of-distribution (OOD) generalization comes into the spotlight. Nonetheless, current endeavors mostly focus on Euclidean data, and its formulation for graph-structured data is not clear and remains under-... | Accept (Poster) | The reviewers have improved their scores after the rebuttal, and I agree that the work has value. It proposes a model-driven data augmentation approach to environment-invariant graph representation. Just like most data augmentation works in graph representation learning, the approach relies on graph proposal generator.... | train | [
"Jk8PeT5zsFy",
"muy6rtAJeuC",
"vRFW9bcs4dA",
"0Y0oYkIB1GW",
"ltJm42iJHBb",
"sm_QIWCQexx",
"vNgOoHUPsIW",
"RvwSKGtS0mo",
"w4axxmL5dUP",
"SUSsXbIRrpU",
"H7Fp2VPzk8t",
"0QDGeUIKlY8",
"FW3pWN7jJn",
"fFIppzxdRtW",
"0XUK4soyskS",
"MIjqoKw3pAq",
"n4xPtg0tj9",
"CRYlrVuK1tc",
"hrOzT49Or-U... | [
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_r... | [
" Thank you for the time and nice feedback. We agree that OoD generalization and adversarial robustness have some overlaps in terms of formulation. Especially, the latter can be seen as OoD data from a pertubation set. \n\nYet, apart from the difference in problem-solving (as we discussed in previous response), the... | [
-1,
-1,
5,
6,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6
] | [
-1,
-1,
4,
4,
-1,
3,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
2
] | [
"muy6rtAJeuC",
"H7Fp2VPzk8t",
"iclr_2022_FQOC5u-1egI",
"iclr_2022_FQOC5u-1egI",
"hrOzT49Or-U",
"iclr_2022_FQOC5u-1egI",
"n4xPtg0tj9",
"w4axxmL5dUP",
"CRYlrVuK1tc",
"iclr_2022_FQOC5u-1egI",
"0QDGeUIKlY8",
"vRFW9bcs4dA",
"MGVbVURaf0k",
"0XUK4soyskS",
"0Y0oYkIB1GW",
"sm_QIWCQexx",
"MIjq... |
iclr_2022_Rty5g9imm7H | Transformer Embeddings of Irregularly Spaced Events and Their Participants | The neural Hawkes process (Mei & Eisner, 2017) is a generative model of irregularly spaced sequences of discrete events. To handle complex domains with many event types, Mei et al. (2020a) further consider a setting in which each event in the sequence updates a deductive database of facts (via domain-specific pattern-m... | Accept (Poster) | The paper builds upon previous work on neural temporal point processes. It mainly proposes the replacement of the LSTMs with Transformers as transformers are widely considered as a more powerful sequence modeling tool and the three advantages listed in the end of section 1 in this paper.
However, on the modeling side,... | train | [
"YTKeA3xpXok",
"G67aASNIjKL",
"IKPbErtd7Ua",
"y1HR5j2oQ_N",
"qDbcpyn0v6",
"tybsaGJGhJd",
"E3SZCLNyMA_",
"_923ja-GSEO",
"twuVgFVrj3u",
"ETaX6Rjuv0h",
"DO7XsOESE-C",
"xZiV3sh3gBV",
"mb4kB-fHx2s",
"Ikh5jxjzk46",
"eY-DbDPQEbF",
"3XU9GuJ8nPp",
"r5Sa5zOdCmZ",
"9FsnXt5gwhb",
"urnTZzsGwT... | [
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
... | [
" Thanks a lot for the clarification. Please see the updated review for more details.",
"The main contribution of this work is a new transformer-based temporal point process (TPP) model.\nThe proposed model defines the conditional intensity at each time $t$ as a function of all past events using the attention mec... | [
-1,
5,
-1,
-1,
-1,
-1,
-1,
-1,
5,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
6,
3
] | [
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
2,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
3
] | [
"IKPbErtd7Ua",
"iclr_2022_Rty5g9imm7H",
"tybsaGJGhJd",
"tybsaGJGhJd",
"mb4kB-fHx2s",
"ETaX6Rjuv0h",
"xZiV3sh3gBV",
"DO7XsOESE-C",
"iclr_2022_Rty5g9imm7H",
"jO1nDIfgtmO",
"Ikh5jxjzk46",
"Q-ssNi24iFA",
"urnTZzsGwTu",
"G67aASNIjKL",
"twuVgFVrj3u",
"twuVgFVrj3u",
"iclr_2022_Rty5g9imm7H",... |
iclr_2022_oDFvtxzPOx | Self-Supervision Enhanced Feature Selection with Correlated Gates | Discovering relevant input features for predicting a target variable is a key scientific question. However, in many domains, such as medicine and biology, feature selection is confounded by a scarcity of labeled samples coupled with significant correlations among features. In this paper, we propose a novel deep learnin... | Accept (Spotlight) | This paper proposes a feature selection method to identify features for downstream supervised tasks, focused on addressing challenges with sample scarcity and feature correlations. The proposed approach is highly motivating in biological and medical applications. Reviewers pointed out various strengths including pote... | train | [
"cwG5NT2WYOC",
"ggnSIBoea5f",
"PFof3iW6T7-",
"gxGLFgPg-lH",
"xa17npJ0W4K",
"zg85FDH__Cl",
"-wLgY1vBI4S"
] | [
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer"
] | [
"This work aims to use self-supervision to identify the most useful features for downstream tasks particularly in the context of correlated features. This is an important aspect that is lacking from many feature selection approaches commonly used within the highly correlated datasets of healthcare. This approach - ... | [
10,
-1,
-1,
-1,
-1,
8,
8
] | [
4,
-1,
-1,
-1,
-1,
2,
3
] | [
"iclr_2022_oDFvtxzPOx",
"xa17npJ0W4K",
"-wLgY1vBI4S",
"zg85FDH__Cl",
"cwG5NT2WYOC",
"iclr_2022_oDFvtxzPOx",
"iclr_2022_oDFvtxzPOx"
] |
iclr_2022_YRq0ZUnzKoZ | A Relational Intervention Approach for Unsupervised Dynamics Generalization in Model-Based Reinforcement Learning | The generalization of model-based reinforcement learning (MBRL) methods to environments with unseen transition dynamics is an important yet challenging problem.
Existing methods try to extract environment-specified information $Z$ from past transition segments to make the dynamics prediction model generalizable to diff... | Accept (Poster) | Description of paper content:
The authors propose a dynamics model that can generalize to novel environments. The train and test MDPs have the same state and action spaces but different dynamics. Environment specific inference is achieved by estimating latent vectors Z that describe the non-stationary or variable part... | train | [
"8EhKHHYOvxk",
"f_DGQFWfKf",
"3HoQTrhl7sn",
"Z4vZhK_vGpm",
"A4DfVTRnqn",
"_DPi6MGDM3i",
"Ss_dwQWY0Wv",
"8z9Iz4eisnP",
"1gC3xX2wruk",
"G2Vk5HyJJco",
"TBI6k7qgkVI",
"GFRjWIGgQfJ",
"94kpjMB2N6e",
"7orHQ9Q6G-V",
"e858gdRzcN"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"author",
"official_reviewer"
] | [
" Dear Authors,\n\nThank you for addressing my concerns. I have raised my score to a 6.",
"This paper considers the unsupervised dynamics generalization problem. In this problem there are a set of train MDPs and a set of test MDPs, all with the same state and action spaces, but with different dynamics functions. ... | [
-1,
6,
6,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8
] | [
-1,
4,
4,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3
] | [
"GFRjWIGgQfJ",
"iclr_2022_YRq0ZUnzKoZ",
"iclr_2022_YRq0ZUnzKoZ",
"Ss_dwQWY0Wv",
"iclr_2022_YRq0ZUnzKoZ",
"1gC3xX2wruk",
"7orHQ9Q6G-V",
"iclr_2022_YRq0ZUnzKoZ",
"G2Vk5HyJJco",
"A4DfVTRnqn",
"e858gdRzcN",
"94kpjMB2N6e",
"f_DGQFWfKf",
"3HoQTrhl7sn",
"iclr_2022_YRq0ZUnzKoZ"
] |
iclr_2022_zuqcmNVK4c2 | Self-Joint Supervised Learning | Supervised learning is a fundamental framework used to train machine learning systems. A supervised learning problem is often formulated using an i.i.d. assumption that restricts model attention to a single relevant signal at a time when predicting. This contrasts with the human ability to actively use related samples ... | Accept (Poster) | This paper explores a classification approach based on labeling pairs of inputs concurrently using a single network, rather than singletons. The authors test the approach on adversarial robustness (towards norm bounded perturbations), OOD detection next to basic standard accuracy calculations.
While the key idea is po... | train | [
"0mn3CuGOr1J",
"79tJyc2F8Dg",
"bgq04xhGww",
"4l7GAqeKj4I"
] | [
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Reviewer Ky9n,\n\nThe Authors did not send any response, unfortunately, but that doesn't mean the reviewers cannot have a discussion between themselves :)\n\nI wanted to briefly comment about the weaknesses you have pointed out.\n\n1. Lack of theoretical analysis: I agree it would be better to have it, but the em... | [
-1,
8,
5,
8
] | [
-1,
5,
3,
4
] | [
"bgq04xhGww",
"iclr_2022_zuqcmNVK4c2",
"iclr_2022_zuqcmNVK4c2",
"iclr_2022_zuqcmNVK4c2"
] |
iclr_2022_UdxJ2fJx7N0 | Minimax Optimization with Smooth Algorithmic Adversaries | This paper considers minimax optimization $\min_x \max_y f(x, y)$ in the challenging setting where $f$ can be both nonconvex in $x$ and nonconcave in $y$. Though such optimization problems arise in many machine learning paradigms including training generative adversarial networks (GANs) and adversarially robust models,... | Accept (Poster) | The paper addresses the problem of non-convex non-concave min-max optimization under the perspective of application of smoothed algorithms between two opponents.
The paper examines a model where the max-player applied a zero-memory smooth (from differential perspective) algorithm and min-player SGD/SNAG or proximal met... | test | [
"jklHOC6--r4",
"3e-sQ81Z9PC",
"XXMN5v3_iMB",
"_2xzC2ysVW",
"2rFIF-Nji2",
"TKYnWKIumXl",
"ROd_O_KDGhE",
"JukexQm5UHa",
"VEwsLemjvhF"
] | [
"author",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Dear reviewers, a gentle reminder to please look at our responses and let us know if you have any further questions/concerns.",
" - **Issue 1**: \nProposition A.22 from Bertsekas’ Thesis:\nLet $f:\\mathbb{R}^n\\times \\mathbb{R}^m\\to (-\\infty,\\infty]$ be a function and let $Y$ be a compact subset of $\\mathb... | [
-1,
-1,
-1,
-1,
-1,
8,
8,
6,
6
] | [
-1,
-1,
-1,
-1,
-1,
3,
4,
4,
5
] | [
"iclr_2022_UdxJ2fJx7N0",
"VEwsLemjvhF",
"JukexQm5UHa",
"ROd_O_KDGhE",
"TKYnWKIumXl",
"iclr_2022_UdxJ2fJx7N0",
"iclr_2022_UdxJ2fJx7N0",
"iclr_2022_UdxJ2fJx7N0",
"iclr_2022_UdxJ2fJx7N0"
] |
iclr_2022_4AZz9osqrar | Self-supervised Learning is More Robust to Dataset Imbalance | Self-supervised learning (SSL) is a scalable way to learn general visual representations since it learns without labels. However, large-scale unlabeled datasets in the wild often have long-tailed label distributions, where we know little about the behavior of SSL. In this work, we systematically investigate self-superv... | Accept (Spotlight) | This paper is proposed to investigate the robustness of self-supervised learning (SSL) and supervised learning (SL) in both balanced (in domain) and imbalanced (out of domain) settings. It can be concluded that SL can regularly learn better representations than SSL, and representations are better from balanced than fro... | train | [
"Jn8l4scSxjY",
"KtTmpCysbJl",
"s16tpaAt0XX",
"CEG4fHomZ88",
"PYQYeoo2V41",
"M1KDVxpajcr",
"71iO5oopWcV",
"o6M-FjJcbNf",
"fIbo1RGHOE",
"7NKD4SjYKV_",
"DQ5q96cJ5jg",
"u2fIj1hvf-m"
] | [
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" Thanks for the added experiments and insights.\n\nOn Q3, that's a good call and you may want to add that to your paper as NLP pre-training datasets are often not the same dataset that you do the final task fine-tuning.",
" After reading the other reviews and rebuttal, my opinion is unchanged and I will keep my ... | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
8,
8,
5,
8
] | [
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
3,
4,
2,
4
] | [
"PYQYeoo2V41",
"71iO5oopWcV",
"CEG4fHomZ88",
"o6M-FjJcbNf",
"u2fIj1hvf-m",
"DQ5q96cJ5jg",
"7NKD4SjYKV_",
"fIbo1RGHOE",
"iclr_2022_4AZz9osqrar",
"iclr_2022_4AZz9osqrar",
"iclr_2022_4AZz9osqrar",
"iclr_2022_4AZz9osqrar"
] |
iclr_2022_jT1EwXu-4hj | From Intervention to Domain Transportation: A Novel Perspective to Optimize Recommendation | The interventional nature of recommendation has attracted increasing attention in recent years. It particularly motivates researchers to formulate learning and evaluating recommendation as causal inference and data missing-not-at-random problems. However, few take seriously the consequence of violating the critical ass... | Accept (Poster) | This paper presented a domain transportation perspective on optimizing recommender systems. The basic motivation is to view recommendation as applying some form of intervention, implying a distributional shift after the recommendation/intervention. Distribution shift brings tremendous difficulty to traditional causal i... | train | [
"tQMLqxfcSAi",
"mP3PDJ1m8gM",
"w677to7Cj8D",
"fNveKDxIjPD",
"KP4jK-_7am3",
"oB6GgZJ_BOR",
"hg0fdY41_aj",
"YAn9tTBsAXZ",
"chyjwblC5aS",
"td5FGU78LE"
] | [
"author",
"official_reviewer",
"author",
"author",
"author",
"author",
"official_reviewer",
"official_reviewer",
"official_reviewer",
"official_reviewer"
] | [
" We thank the reviewer for the further response that helps to clarify the questions. We believe there are several misconceptions that cause the reviewer's confusion.\n\n1. Our work aims only at the IW- and DA-based IR methods where sufficient overlapping is a crucial presumption. We have made our scope very clear ... | [
-1,
-1,
-1,
-1,
-1,
-1,
6,
6,
6,
5
] | [
-1,
-1,
-1,
-1,
-1,
-1,
4,
3,
3,
4
] | [
"mP3PDJ1m8gM",
"fNveKDxIjPD",
"YAn9tTBsAXZ",
"td5FGU78LE",
"chyjwblC5aS",
"hg0fdY41_aj",
"iclr_2022_jT1EwXu-4hj",
"iclr_2022_jT1EwXu-4hj",
"iclr_2022_jT1EwXu-4hj",
"iclr_2022_jT1EwXu-4hj"
] |
iclr_2022_fOsN52jn25l | Dual Lottery Ticket Hypothesis | Fully exploiting the learning capacity of neural networks requires overparameterized dense networks. On the other side, directly training sparse neural networks typically results in unsatisfactory performance. Lottery Ticket Hypothesis (LTH) provides a novel view to investigate sparse network training and maintain its ... | Accept (Poster) | I recommend this paper for acceptance but I do so with significant reservations. Since this metareview will be public for all time, I direct this metareview to future readers of this paper so that they can weigh its merits and drawbacks in a clear-minded way.
This paper proposes a "dual lottery ticket hypothesis." For... | train | [
"x2Kop_APU__",
"lWfAa_KeCIv",
"d715oiRBXC4",
"WjejlmhoqQ-",
"kF0hPb079vR",
"wrJkqn7BiO",
"zTzCzUrjwA",
"GFAVi8KFUUp",
"zDbK-BqLcUE",
"nvkz0PpG1h-",
"jIGtLw-2Ewc",
"-LehyWamjl",
"HK8OlSydXIS",
"vmIWy1WMEg",
"oYEYRvqS4cY",
"jviIAIi2oa",
"APJsvJCNlud",
"s1AmwI_PpCH",
"bxqKywVEoUY",
... | [
"author",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_re... | [
" We further supplement the experiments for the \"ablation of delaying the pruning\" on the CIFAR100 dataset using ResNet56. The results are shown below.\n\n|Method|pr0.5|pr0.7|pr0.9|pr0.95|pr0.98|\n|----|----|----|----|----|----|\n|L1|71.96| 71.59| 68.29| 64.74| 50.04|\n|LTH Iter-5|70.57| 69.54| 64.84| 60.45| 53.8... | [
-1,
8,
-1,
-1,
6,
-1,
-1,
-1,
8,
-1,
-1,
-1,
6,
-1,
-1,
-1,
-1,
-1,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
-1,
4,
-1,
-1,
5,
-1,
-1,
-1,
4,
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1
] | [
"wrJkqn7BiO",
"iclr_2022_fOsN52jn25l",
"wrJkqn7BiO",
"wrJkqn7BiO",
"iclr_2022_fOsN52jn25l",
"GFAVi8KFUUp",
"nvkz0PpG1h-",
"kF0hPb079vR",
"iclr_2022_fOsN52jn25l",
"jviIAIi2oa",
"-LehyWamjl",
"vmIWy1WMEg",
"iclr_2022_fOsN52jn25l",
"oYEYRvqS4cY",
"s1AmwI_PpCH",
"2_a2i9t7fj",
"2NyP2go_Do... |
iclr_2022_dDjSKKA5TP1 | Incremental False Negative Detection for Contrastive Learning | Self-supervised learning has recently shown great potential in vision tasks through contrastive learning, which aims to discriminate each image, or instance, in the dataset. However, such instance-level learning ignores the semantic relationship among instances and sometimes undesirably repels the anchor from the seman... | Accept (Poster) | While the reviewers were somewhat split on this paper, they all found some strengths, and pointed out some weaknesses. Among these the main seems to be the somewhat incremental nature of the work, in particular with respect to PCL. As the authors point out, the differences w.r.t. PCL are meaningful and include the main... | train | [
"I2FmeR4KrUt",
"DdO39VlGyA6",
"hw0g1V0WZPg",
"AKZqG39UvY0",
"YCtce44sFz",
"Orb14usGgE",
"mSqttR6HO24",
"4Ki81F411An",
"19MC10qWZcj",
"kl2CIquvwqg",
"jOlJ5flf7C9",
"orJ95nqeOwH",
"urcnYCQAOkM"
] | [
"official_reviewer",
"author",
"official_reviewer",
"official_reviewer",
"author",
"author",
"official_reviewer",
"author",
"official_reviewer",
"author",
"author",
"author",
"official_reviewer"
] | [
"This paper deals with the very important topic of “false negatives” in SSL and shows that “false negatives” have a huge impact on performance of SSL methods. One of the most interesting experiments they show is that as they increase the number of classes in the dataset the performance drops more and more. They pro... | [
6,
-1,
8,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
5
] | [
4,
-1,
4,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
-1,
4
] | [
"iclr_2022_dDjSKKA5TP1",
"iclr_2022_dDjSKKA5TP1",
"iclr_2022_dDjSKKA5TP1",
"orJ95nqeOwH",
"I2FmeR4KrUt",
"mSqttR6HO24",
"4Ki81F411An",
"19MC10qWZcj",
"urcnYCQAOkM",
"urcnYCQAOkM",
"urcnYCQAOkM",
"hw0g1V0WZPg",
"iclr_2022_dDjSKKA5TP1"
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.