paper_id stringlengths 9 12 | venue stringclasses 139
values | year stringclasses 7
values | paper_title stringlengths 0 181 | paper_authors stringlengths 4 925 | paper_abstract stringlengths 1 5k | paper_keywords stringlengths 2 436 | paper_content stringlengths 0 100k | review_id stringlengths 9 12 | review_title stringlengths 0 500 | review_rating stringclasses 61
values | review_text stringlengths 2 28.3k | review_confidence stringclasses 13
values | text stringlengths 402 130k |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
SPxaJuM4Hbz | aclweb.org/ACL/2020/Workshop/NLP-COVID | 2020 | Document Classification for COVID-19 Literature | ["Bernal Jimenez Gutierrez", "Juncheng Zeng", "Dongdong Zhang", "Ping Zhang", "Yu Su"] | The global pandemic has made it more important than ever to quickly and accurately retrieve relevant scientific literature for effective consumption by researchers in a wide range of fields. We provide an analysis of several multi-label document classification models on the LitCovid dataset, a growing collection of 8,0... | ["COVID-19", "BERT", "LitCovid", "Document Classification", "Biomedical NLP", "BioBERT", "Longformer", "Coronavirus"] | Document Classification for COVID-19 LiteratureBernal Jim ́enez Guti ́errez, Juncheng Zeng, Dongdong Zhang, Ping Zhang, Yu SuThe Ohio State Universityfjimenezgutierrez.1,zeng.671,zhang.11069,zhang.10631,su.809 g@osu.eduAbstractThe global pandemic has made it more im-portant than ever to quickly and accurately re-trieve... | -SfWfzSmKc | Study of automatic classification of LitCovid categories. | 6: Marginally above acceptance threshold | This article presents a comparison of methods for the classification of articles from the LitCovid collection. From all the evaluated methods, one based on BioBert performs better compared to other evaluated methods. The contribution of the paper is a comprehensive study of a broad range of methods to the multi-class c... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Document Classification for COVID-19 Literature
### Paper Abstract
The global pandemic has made it more important than ever to quickly and accurately retrieve relevant s... |
HkuGJ3kCb | ICLR.cc/2018/Conference | 2018 | All-but-the-Top: Simple and Effective Postprocessing for Word Representations | ["Jiaqi Mu", "Pramod Viswanath"] | Real-valued word representations have transformed NLP applications; popular examples are word2vec and GloVe, recognized for their ability to capture linguistic regularities. In this paper, we demonstrate a {\em very simple}, and yet counter-intuitive, postprocessing technique -- eliminate the common mean vector and a f... | ["simple", "word representations", "effective postprocessing", "variety", "nlp applications", "popular examples", "ability", "linguistic regularities", "technique"] | ABSTRACTReal-valued word representations have transformed NLP applications; popularexamples are word2vec and GloVe, recognized for their ability to capture linguisticregularities. In this paper, we demonstrate a very simple , and yet counter-intuitive,postprocessing technique – eliminate the common mean vector and a fe... | HyvB9RKez | In-depth explanation of known phenomenon. Experiments need some strengthening. | 6: Marginally above acceptance threshold | This paper provides theoretical and empirical motivations for removing the top few principle components of commonly-used word embeddings.
The paper is well-written and I enjoyed reading it. However, it does not explain how significant this result is beyond that of (Bullinaria and Levy, 2012), who also removed the top ... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
All-but-the-Top: Simple and Effective Postprocessing for Word Representations
### Paper Abstract
Real-valued word representations have transformed NLP applications; popu... |
r1eX1yrKwB | ICLR.cc/2020/Conference | 2020 | Distribution Matching Prototypical Network for Unsupervised Domain Adaptation | ["Lei Zhu", "Wei Wang", "Mei Hui Zhang", "Beng Chin Ooi", "Chang Yao"] | State-of-the-art Unsupervised Domain Adaptation (UDA) methods learn transferable features by minimizing the feature distribution discrepancy between the source and target domains. Different from these methods which do not model the feature distributions explicitly, in this paper, we explore explicit feature distributio... | ["Deep Learning", "Unsupervised Domain Adaptation", "Distribution Modeling"] | ABSTRACTState-of-the-art Unsupervised Domain Adaptation (UDA) methods learn trans-ferable features by minimizing the feature distribution discrepancy between thesource and target domains. Different from these methods which do not model thefeature distributions explicitly, in this paper, we explore explicit feature dist... | SygAfV3gqH | Official Blind Review #1 | 3: Weak Reject | The paper develops a new method for adapting models trained on labeled data from some source domain to unlabeled data in a target domain. The authors accomplish this by adapting a technique from [1] and [2] enforcing that the deep features learned during training approximately follow a Gaussian mixture distribution. Wi... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Distribution Matching Prototypical Network for Unsupervised Domain Adaptation
### Paper Abstract
State-of-the-art Unsupervised Domain Adaptation (UDA) methods learn tran... | |
R6tNszN_QfA | ICLR.cc/2021/Conference | 2021 | Adversarial Problems for Generative Networks | ["Kalliopi Basioti", "George V. Moustakides"] | We are interested in the design of generative networks. The training of these mathematical structures is mostly performed with the help of adversarial (min-max) optimization problems. We propose a simple methodology for constructing such problems assuring, at the same time, consistency of the corresponding solution. We... | ["generative networks", "adversarial generative networks"] | ABSTRACTWe are interested in the design of generative networks. The training of these math-ematical structures is mostly performed with the help of adversarial (min-max)optimization problems. We propose a simple methodology for constructing suchproblems assuring, at the same time, consistency of the corresponding solut... | QBIqYE8PKkw | Interesting ideas, relationship to previous work unclear | 4: Ok but not good enough - rejection | Summary
========
In this paper, the authors set out to find what scalar functions will make for a “max” part of the “min-max” GAN objective. They then find such a class of functions, and show that only a ratio between two equal probabilities will be admitted as a solution.
Pros:
====
The paper nicely introduces a diff... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Adversarial Problems for Generative Networks
### Paper Abstract
We are interested in the design of generative networks. The training of these mathematical structures is ... |
H1gax6VtDB | ICLR.cc/2020/Conference | 2020 | Contrastive Learning of Structured World Models | ["Thomas Kipf", "Elise van der Pol", "Max Welling"] | A structured understanding of our world in terms of objects, relations, and hierarchies is an important component of human cognition. Learning such a structured world model from raw sensory data remains a challenge. As a step towards this goal, we introduce Contrastively-trained Structured World Models (C-SWMs). C-SWMs... | ["state representation learning", "graph neural networks", "model-based reinforcement learning", "relational learning", "object discovery"] | ABSTRACTA structured understanding of our world in terms of objects, relations, and hierar-chies is an important component of human cognition. Learning such a structuredworld model from raw sensory data remains a challenge. As a step towards thisgoal, we introduce Contrastively-trained Structured World Models (C-SWMs).... | H1l81UwCYS | Official Blind Review #2 | 8: Accept | This paper tackles the problem of learning an encoder and transition model of an environment, such that the representation learnt uses an object-centric representation which could favor compositionality and generalisation. This is trained using a contrastive max-margin loss, instead of a generative loss as previously e... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Contrastive Learning of Structured World Models
### Paper Abstract
A structured understanding of our world in terms of objects, relations, and hierarchies is an importan... | |
BJl_a2VYPH | ICLR.cc/2020/Conference | 2020 | Natural Language Adversarial Attack and Defense in Word Level | ["Xiaosen Wang", "Hao Jin", "Kun He"] | Up until very recently, inspired by a mass of researches on adversarial examples for computer vision, there has been a growing interest in designing adversarial attacks for Natural Language Processing (NLP) tasks, followed by very few works of adversarial defenses for NLP. To our knowledge, there exists no defense meth... | ["adversarial examples", "text adversarial defense", "text adversarial attack", "synonym encoding", "natural language processing"] | ABSTRACTUp until very recently, inspired by a mass of researches on adversarial examplesfor computer vision, there has been a growing interest in designing adversarialattacks for Natural Language Processing (NLP) tasks, followed by very few worksof adversarial defenses for NLP. To our knowledge, there exists no defense... | Skg9UyFAuH | Official Blind Review #2 | 6: Weak Accept | Summary:
The paper propose a novel adversarial attack and defense method for NLP models. The defense method proposes to smooth out the feature space such that features within an epsilon distance to the current features has the same output.
Comments
1. The paper is clear written and well motivated.
2. It virtual adv... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Natural Language Adversarial Attack and Defense in Word Level
### Paper Abstract
Up until very recently, inspired by a mass of researches on adversarial examples for com... | |
FGvJvxn2wWO | ICLR.cc/2021/Conference | 2021 | Domain Adaptation with Morphologic Segmentation | ["Jonathan Klein", "Soren Pirk", "Dominik Michels"] | We present a novel domain adaptation framework that uses morphologic segmentation to translate images from arbitrary input domains (real and synthetic) into a uniform output domain. Our framework is based on an established image-to-image translation pipeline that allows us to first transform the input image into a gene... | ["Domain Adaptation", "Morphologic Segmentation", "Image-to-image Translation"] | ABSTRACTWe present a novel domain adaptation framework that uses morphologic segmen-tation to translate images from arbitrary input domains (real and synthetic) intoa uniform output domain. Our framework is based on an established image-to-image translation pipeline that allows us to first transform the input image int... | U61FHICU2R3 | Unclear motivation and insufficient evaluation | 4: Ok but not good enough - rejection | **Summary**
The paper proposes a framework to adapt images from various visual domains to a single domain using image-to-image translation methods. Instead of training the image-to-image translation model from each input domain to the output domain, the paper proposes to first convert the input domain into an Edge-plu... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Domain Adaptation with Morphologic Segmentation
### Paper Abstract
We present a novel domain adaptation framework that uses morphologic segmentation to translate images ... |
gp5Uzbl-9C- | ICLR.cc/2021/Conference | 2021 | Systematic Evaluation of Causal Discovery in Visual Model Based Reinforcement Learning | ["Nan Rosemary Ke", "Aniket Rajiv Didolkar", "Sarthak Mittal", "Anirudh Goyal", "Guillaume Lajoie", "Stefan Bauer", "Danilo Jimenez Rezende", "Michael Curtis Mozer", "Yoshua Bengio", "Christopher Pal"] | Inducing causal relationships from observations is a classic problem in machine learning. Most work in causality starts from the premise that the causal variables themselves have known semantics or are observed. However, for AI agents such as robots trying to make sense of their environment, the only observables are lo... | ["causal induction", "model based RL"] | ABSTRACTInducing causal relationships from observations is a classic problem in machinelearning. Most work in causality starts from the premise that the causal variablesthemselves are observed. However, for AI agents such as robots trying to makesense of their environment, the only observables are low-level variables l... | STIcPRr_Q0s | Another benchmark to facilitate causal induction for RL with pixel-level observations. | 6: Marginally above acceptance threshold | Summary:
This paper proposes a suite of RL benchmarks to facilitate inducing causal relationships from visual observations. They show that structural inductive biases are beneficial for causal relationship learning and model-based RL by testing a variety of representation learning algorithms on this benchmark.
######... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Systematic Evaluation of Causal Discovery in Visual Model Based Reinforcement Learning
### Paper Abstract
Inducing causal relationships from observations is a classic pr... |
HkljfjFee | ICLR.cc/2017/conference | 2017 | Support Regularized Sparse Coding and Its Fast Encoder | ["Yingzhen Yang", "Jiahui Yu", "Pushmeet Kohli", "Jianchao Yang", "Thomas S. Huang"] | Sparse coding represents a signal by a linear combination of only a few atoms of a learned over-complete dictionary. While sparse coding exhibits compelling performance for various machine learning tasks, the process of obtaining sparse code with fixed dictionary is independent for each data point without considering t... | ["srsc", "sparse codes", "sparse coding", "manifold structure", "data", "support", "fast encoder support", "fast encoder sparse", "signal", "linear combination"] | ABSTRACTSparse coding represents a signal by a linear combination of only a few atomsof a learned over-complete dictionary. While sparse coding exhibits compellingperformance for various machine learning tasks, the process of obtaining sparsecode with fixed dictionary is independent for each data point without consider... | S1DiuduNx | Support Regularized Sparse Coding and Its Fast Encoder | 7: Good paper, accept | The work proposes to use the geometry of data (that is considered to be known a priori) in order to have more consistent sparse coding. Namely, two data samples that are similar or neighbours, should have a sparse code that is similar (in terms of support). The general idea is not unique, but it is an interesting one (... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Support Regularized Sparse Coding and Its Fast Encoder
### Paper Abstract
Sparse coding represents a signal by a linear combination of only a few atoms of a learned over... |
r1ltgp4FwS | ICLR.cc/2020/Conference | 2020 | Learning Temporal Coherence via Self-Supervision for GAN-based Video Generation | ["Mengyu Chu", "You Xie", "Jonas Mayer", "Laura Leal-Taix\u00e9", "Nils Th\u00fcrey"] | We focus on temporal self-supervision for GAN-based video generation tasks. While adversarial training successfully yields generative models for a variety of areas, temporal relationship in the generated data is much less explored. This is crucial for sequential generation tasks, e.g. video super-resolution and unpaire... | ["adversarial training", "generative models", "unpaired video translation", "video super-resolution", "temporal coherence", "self-supervision", "cycle-consistency"] | ABSTRACTWe focus on temporal self-supervision for GAN-based video generation tasks.While adversarial training successfully yields generative models for a variety ofareas, temporal relationship in the generated data is much less explored. This iscrucial for sequential generation tasks , e.g. video super-resolution and u... | HkxvTKqTKS | Official Blind Review #1 | 3: Weak Reject | Contribution
Augments the loss of video generation systems with a discriminator that considers multiple frames (as opposed to single frames independently) and a new objective termed ping-pong loss which is introduced in order to deal with “artifacts” that appear in video generation. The paper also proposes a few automa... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Learning Temporal Coherence via Self-Supervision for GAN-based Video Generation
### Paper Abstract
We focus on temporal self-supervision for GAN-based video generation t... | |
9mk7Quvo7L | AAAI.org/2023/Bridge/CCBridge | 2023 | From Continual Learning to Causal Discovery in Robotics | ["Luca Castri", "Sariah Mghames", "Nicola Bellotto"] | Reconstructing accurate causal models of dynamic systems from time-series of sensor data is a key problem in many real-world scenarios.
In this paper, we present an overview based on our experience about practical challenges that the causal analysis encounters when applied to autonomous robots and how Continual Learnin... | ["continual learning", "discovery", "robotics", "accurate causal models", "dynamic systems", "sensor data", "key problem", "many", "scenarios", "overview"] | From Continual Learning to Causal Discovery in Robotics*Luca Castri,1Sariah Mghames,1Nicola Bellotto1,21University of Lincoln, UK2University of Padua, Italy{lcastri, smghames }@lincoln.ac.uk, nbellotto@dei.unipd.itAbstractReconstructing accurate causal models of dynamic systemsfrom time-series of sensor data is a key p... | LlLRL5fSl7K | Review | 7: Good paper, accept | This paper describes Causal Robot Discovery (CRD), a continual learning approach for causal discovery. The approach can address current challenges in limited compute resources as well as efficient update of causal models continuously.
The paper gave a clear overview of their proposed method, which is good for applying... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
From Continual Learning to Causal Discovery in Robotics
### Paper Abstract
Reconstructing accurate causal models of dynamic systems from time-series of sensor data is a ... |
aa0705s2Qc | ICLR.cc/2021/Conference | 2021 | Measuring Visual Generalization in Continuous Control from Pixels | ["Jake Grigsby", "Yanjun Jane Qi"] | Self-supervised learning and data augmentation have significantly reduced the performance gap between state and image-based reinforcement learning agents in continuous control tasks. However, it is still unclear whether current techniques can face the variety of visual conditions required by real-world environments. We... | ["Continuous Control", "Reinforcement Learning"] | Under review as a conference paper at ICLR 2021Measuring Visual Generalization inContinuous Control from PixelsAnonymous authorsPaper under double-blind reviewAbstractSelf-supervised learning and data augmentation have signicantly reducedthe performance gap between state and image-based reinforcement learn-ing agents i... | IHCeoJcDP1M | This paper presents a "remastered" version of the Deep Mind Control suite that allows for testing of visual generalization. The paper presents an interesting augmentation to the DMC environment. | 5: Marginally below acceptance threshold | This paper presents a "remastered" version of the Deep Mind Control suite that allows for testing of visual generalization to a number of environmental changes (e.g., pattern/texture of the floor, background, and lighting).
I will begin my review by stating that I have somewhat limited knowledge of soft actor critic... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Measuring Visual Generalization in Continuous Control from Pixels
### Paper Abstract
Self-supervised learning and data augmentation have significantly reduced the perfor... |
S1AtgaPug | ICLR.cc/2017/workshop | 2017 | Adjusting for Dropout Variance in Batch Normalization and Weight Initialization | ["Dan Hendrycks", "Kevin Gimpel"] | We show how to adjust for the variance introduced by dropout with corrections to weight initialization and Batch Normalization, yielding higher accuracy. Though dropout can preserve the expected input to a neuron between train and test, the variance of the input differs. We thus propose a new weight initialization by c... | ["Deep learning"] | ABSTRACTWe show how to adjust for the variance introduced by dropout with corrections toweight initialization and Batch Normalization, yielding higher accuracy. Thoughdropout can preserve the expected input to a neuron between train and test, thevariance of the input differs. We thus propose a new weight initialization... | HkuEuBQig | A useful idea to emphasize, but not a significant contribution | 6: Marginally above acceptance threshold | Paper summary -
Several weight initialization schemes for deep neural networks aim to maintain
the variance of each layer's output to be one, in order to prevent an
exponential blow-up or decay with increasing depth. This paper describes a way
to adjust the scale of the weight initialization to take into account the f... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Adjusting for Dropout Variance in Batch Normalization and Weight Initialization
### Paper Abstract
We show how to adjust for the variance introduced by dropout with corr... |
Bpw_O132lWT | ICLR.cc/2021/Conference | 2021 | Dynamic of Stochastic Gradient Descent with State-dependent Noise | ["Qi Meng", "Shiqi Gong", "Wei Chen", "Zhi-Ming Ma", "Tie-Yan Liu"] | Stochastic gradient descent (SGD) and its variants are mainstream methods to train deep neural networks. Since neural networks are non-convex, more and more works study the dynamic behavior of SGD and its impact to generalization, especially the escaping efficiency from local minima. However, these works make the over-... | ["state-dependent noise", "power-law dynamic", "stochastic gradient descent", "generalization", "deep neural network", "heavy-tailed", "escape time"] | ABSTRACTStochastic gradient descent (SGD) and its variants are mainstream methods totrain deep neural networks. Since neural networks are non-convex, more andmore works study the dynamic behavior of SGD and its impact to generalization,especially the escaping efficiency from local minima. However, these worksmake the o... | 0KhY94zUnkY | Official Review of Reviewer #3 | 6: Marginally above acceptance threshold | ### Summary
This paper proposes an analysis for an approximate dynamic of SGD which captures the heavy-tailed noise distributions seen practically at local minima. The authors derive this new dynamic (which they call Power-law dynamic) using basic principles and the assumption that the noise variance depends on the sta... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Dynamic of Stochastic Gradient Descent with State-dependent Noise
### Paper Abstract
Stochastic gradient descent (SGD) and its variants are mainstream methods to train d... |
lGFed4WfQWg | NoDaLiDa/2023/Conference | 2023 | Quasi: a synthetic Question-Answering dataset in Swedish using GPT-3 and zero-shot learning | ["Dmytro Kalpakchi", "Johan Boye"] | This paper describes the creation and evaluation of a synthetic dataset of Swedish multiple-choice questions (MCQs) for reading comprehension using GPT-3. Although GPT-3 is trained mostly on English data, with only 0.11% of Swedish texts in its training material, the model still managed to generate MCQs in Swedish. Abo... | ["question generation", "question answering", "synthetic data", "data augmentation", "Swedish", "multiple-choice questions", "GPT-3", "zero-shot learning"] | Quasi: a synthetic Question-Answering dataset in Swedish using GPT-3and zero-shot learningDmytro KalpakchiDivision of Speech, Music and HearingKTH Royal Institute of TechnologyStockholm, Swedendmytroka@kth.seJohan BoyeDivision of Speech, Music and HearingKTH Royal Institute of TechnologyStockholm, Swedenjboye@kth.seAbs... | jPCJcJv-hsE | A GPT-3 created dataset for Swedish QA | 9: Top 15% of accepted papers, strong accept | In "Quasi: a synthetic Question-Answering dataset in Swedish using GPT-3 and zero-shot learning" the authors introduce a new dataset for Swedish QA that has been created with the help of GPT-3.
Even though 0.11% of GPT-3's training data was in Swedish, the model is capable of handling Swedish texts.
The authors feed G... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Quasi: a synthetic Question-Answering dataset in Swedish using GPT-3 and zero-shot learning
### Paper Abstract
This paper describes the creation and evaluation of a synt... |
3hGNqpI4WS | ICLR.cc/2021/Conference | 2021 | Deployment-Efficient Reinforcement Learning via Model-Based Offline Optimization | ["Tatsuya Matsushima", "Hiroki Furuta", "Yutaka Matsuo", "Ofir Nachum", "Shixiang Gu"] | Most reinforcement learning (RL) algorithms assume online access to the environment, in which one may readily interleave updates to the policy with experience collection using that policy. However, in many real-world applications such as health, education, dialogue agents, and robotics, the cost or potential risk of de... | ["Reinforcement Learning", "deployment-efficiency", "offline RL", "Model-based RL"] | ABSTRACTMost reinforcement learning (RL) algorithms assume online access to the environ-ment, in which one may readily interleave updates to the policy with experiencecollection using that policy. However, in many real-world applications such ashealth, education, dialogue agents, and robotics, the cost or potential ris... | hynMigrsjO4 | Justification of Deployment Efficiency metric is unclear | 5: Marginally below acceptance threshold | The paper has been well written and is easy to understand. The premise of the paper is clear, and methods have been evaluated properly.
Suggestions to improve the paper:
- While the deployment efficiency metric has been clearly explained, the justification of this metric has not been elaborated on sufficiently. For t... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Deployment-Efficient Reinforcement Learning via Model-Based Offline Optimization
### Paper Abstract
Most reinforcement learning (RL) algorithms assume online access to t... |
SyeHbtgSiN | ICML.cc/2019/Workshop/RL4RealLife | 2019 | Q-Learning for Continuous Actions with Cross-Entropy Guided Policies | ["Riley Simmons-Edler", "Ben Eisner", "Eric Mitchell", "Sebastian Seung", "Daniel Lee"] | Off-Policy reinforcement learning (RL) is an important class of methods for many problem domains, such as robotics, where the cost of collecting data is high and on-policy methods are consequently intractable. Standard methods for applying Q-learning to continuous-valued action domains involve iteratively sampling the ... | ["Reinforcement Learning", "Q-learning", "Sample-based optimization", "Deep learning", "Hyperparameter robustness", "Training stability"] | Q-Learning for Continuous Actions with Cross-Entropy Guided PoliciesRiley Simmons-Edler1 2 *Ben Eisner2 *Eric Mitchell2 *Sebastian Seung2Daniel Lee2AbstractOff-Policy reinforcement learning (RL) is an im-portant class of methods for many problem do-mains, such as robotics, where the cost of collect-ing data is high and... | Hklz_1kH6E | Not a generic or novel approach | 3: Borderline | In this paper, the authors study RL in MDPs where the action space is continuous. Continuous action space problems are prominent in the control task. The authors nicely motivate this task and describe the prior works on on-policy approaches. They argue that on-policy methods might not be sample efficient, therefore of... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Q-Learning for Continuous Actions with Cross-Entropy Guided Policies
### Paper Abstract
Off-Policy reinforcement learning (RL) is an important class of methods for many ... |
ByeLBj0qFQ | ICLR.cc/2019/Conference | 2019 | Unsupervised Image to Sequence Translation with Canvas-Drawer Networks | ["Kevin Frans", "Chin-Yi Cheng"] | Encoding images as a series of high-level constructs, such as brush strokes or discrete shapes, can often be key to both human and machine understanding. In many cases, however, data is only available in pixel form. We present a method for generating images directly in a high-level domain (e.g. brush strokes), without... | ["image", "translation", "unsupervised", "model-based"] | ABSTRACTEncoding images as a series of high-level constructs, such as brush strokes or dis-crete shapes, can often be key to both human and machine understanding. In manycases, however, data is only available in pixel form. We present a method for gen-erating images directly in a high-level domain (e.g. brush strokes),... | ByxIZdYF3Q | Simple and working idea, insufficient evaluations | 6: Marginally above acceptance threshold | This is an interesting paper with simple idea and good results. I like the fact that the authors adopt simple autoencoder-like models instead of GANs or RL.
Following are a couple questions that I am concerned about:
1. Is the ordering of strokes important at all? I suspect that a drawer model that outputs 10 strokes i... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Unsupervised Image to Sequence Translation with Canvas-Drawer Networks
### Paper Abstract
Encoding images as a series of high-level constructs, such as brush strokes or ... |
rJe5_CNtPB | ICLR.cc/2020/Conference | 2020 | Attention Forcing for Sequence-to-sequence Model Training | ["Qingyun Dou", "Yiting Lu", "Joshua Efiong", "Mark J.F. Gales"] | Auto-regressive sequence-to-sequence models with attention mechanism have achieved state-of-the-art performance in many tasks such as machine translation and speech synthesis. These models can be difficult to train. The standard approach, teacher forcing, guides a model with reference output history during training. Th... | ["deep learning", "sequence-to-sequence model", "attention mechanism", "speech synthesis", "machine translation"] | ABSTRACTAuto-regressive sequence-to-sequence models with attention mechanism haveachieved state-of-the-art performance in many tasks such as machine translationand speech synthesis. These models can be difficult to train. The standard ap-proach, teacher forcing, guides a model with reference output history during train... | BJejk40KFr | Official Blind Review #3 | 1: Reject | This paper proposes an alternative mechanism of training the attention values of a sequence to sequence learning model as applied to tasks like speech synthesis and translation. During training they compute two forms of attention: (1) the standard soft-attention from a decoder fed with teacher forced output, and (2) t... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Attention Forcing for Sequence-to-sequence Model Training
### Paper Abstract
Auto-regressive sequence-to-sequence models with attention mechanism have achieved state-of-... | |
gl3D-xY7wLq | ICLR.cc/2021/Conference | 2021 | Noise or Signal: The Role of Image Backgrounds in Object Recognition | ["Kai Yuanqing Xiao", "Logan Engstrom", "Andrew Ilyas", "Aleksander Madry"] | We assess the tendency of state-of-the-art object recognition models to depend on signals from image backgrounds. We create a toolkit for disentangling foreground and background signal on ImageNet images, and find that (a) models can achieve non-trivial accuracy by relying on the background alone, (b) models often misc... | ["Backgrounds", "Model Biases", "Robustness", "Computer Vision"] | ABSTRACTWe assess the tendency of state-of-the-art object recognition models to depend onsignals from image backgrounds. We create a toolkit for disentangling foregroundand background signal on ImageNet images, and find that (a) models can achievenon-trivial accuracy by relying on the background alone, (b) models often... | jNzCrbDiNln | A comprehensive study on the role of background in image classification | 7: Good paper, accept | The authors presented a comprehensive study on the role of background in image classification. They designed a new set of data and a lot of experiments to find answers to the following questions: (1) How much decrease in classification accuracy if the background signal is removed? (2) Can a model successfully classify ... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Noise or Signal: The Role of Image Backgrounds in Object Recognition
### Paper Abstract
We assess the tendency of state-of-the-art object recognition models to depend on... |
sjGBjudWib | ICLR.cc/2021/Conference | 2021 | FAST GRAPH ATTENTION NETWORKS USING EFFECTIVE RESISTANCE BASED GRAPH SPARSIFICATION | ["Rakshith Sharma Srinivasa", "Cao Xiao", "Lucas Glass", "Justin Romberg", "Jimeng Sun"] | The attention mechanism has demonstrated superior performance for inference over nodes in graph neural networks (GNNs), however, they result in a high computational burden during both training and inference. We propose FastGAT, a method to make attention based GNNs lightweight by using spectral sparsification to genera... | ["Graph neural networks", "Graph attention networks", "graph sparsification", "spectral sparsification"] | ABSTRACTThe attention mechanism has demonstrated superior performance for inferenceover nodes in graph neural networks (GNNs), however, they result in a highcomputational burden during both training and inference. We propose FastGAT ,amethod to make attention based GNNs lightweight by using spectral sparsificationto ge... | gkTL9mWfbn1 | Limited novelty | 4: Ok but not good enough - rejection | This work proposed to use graph-sparcification method to decrease the number of edges so as to support GAT training over large networks. The graph sparcification approach is based on resistance. Empirical results by comparing with standard GAT or randomly edge-dropping strategy demonstrate the effectiveness of their ap... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
FAST GRAPH ATTENTION NETWORKS USING EFFECTIVE RESISTANCE BASED GRAPH SPARSIFICATION
### Paper Abstract
The attention mechanism has demonstrated superior performance for ... |
TMUR2ovJfjE | ICLR.cc/2021/Conference | 2021 | Co-complexity: An Extended Perspective on Generalization Error | ["Rohan Ghosh", "Mehul Motani"] | It is well known that the complexity of a classifier's function space controls its generalization gap, with two important examples being VC-dimension and Rademacher complexity (R-Complexity). We note that these traditional generalization error bounds consider the ground truth label generating function (LGF) to be fixed... | ["Generalization", "Theoretical Machine Learning", "Convolutional Neural Networks"] | ABSTRACTIt is well known that the complexity of a classifier’s function space controls its gen-eralization gap, with two important examples being VC-dimension and Rademachercomplexity (R-Complexity). We note that these traditional generalization errorbounds consider the ground truth label generating function (LGF) to b... | n4VWg7Vi6kY | Interesting work but, in this form, the paper not ready for publication. | 4: Ok but not good enough - rejection | This paper aims to propose a new complexity measure, called co-complexity, to control classifiers' generalization gap. This new measure acts like a joint-entropy and leads to tighter bounds on the generalization error in this setting. The main idea is to extend the classical complexity measure of Barlett & Mendelson (2... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Co-complexity: An Extended Perspective on Generalization Error
### Paper Abstract
It is well known that the complexity of a classifier's function space controls its gene... |
h3Miqa_jqN | MIDL.io/2020/Conference | 2020 | Multiple resolution residual network for automatic thoracic organs-at-risk segmentation from CT | ["Hyemin Um", "Jue Jiang", "Maria Thor", "Andreas Rimner", "Leo Luo", "Joseph O. Deasy", "Harini Veeraraghavan"] | We implemented and evaluated a multiple resolution residual network (MRRN) for multiple normal organs-at-risk (OAR) segmentation from computed tomography (CT) images for thoracic radiotherapy treatment (RT) planning. Our approach simultaneously combines feature streams computed at multiple image resolutions and feature... | ["Multiple residual feature streams", "thoracic normal organs", "AAPM thoracic grand challenge dataset"] | Medical Imaging with Deep Learning 2020 1{5, 2020 MIDL 2020 { Short PaperMultiple resolution residual network for automatic thoracicorgans-at-risk segmentation from CTHyemin Um1Jue Jiang1Maria Thor1Andreas Rimner1Leo Luo1Joseph O.Deasy1Harini Veeraraghavan1veerarah@mskcc.org1Department of Medical Physics, Memorial Sloa... | zlNf5tqiRq | multi-scale method for organ segmentation | 2: Weak reject | #Summary
This work proposed a new deep-learning architecture for the segmentation of normal organs at risk in thoracic CT data. The authors introduce residual connections from downed scale to upper scales for skip connection of U-Net architecture. They explained these residual connections between down-scaled feature ma... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Multiple resolution residual network for automatic thoracic organs-at-risk segmentation from CT
### Paper Abstract
We implemented and evaluated a multiple resolution res... |
Bke4KsA5FX | ICLR.cc/2019/Conference | 2019 | Generative Code Modeling with Graphs | ["Marc Brockschmidt", "Miltiadis Allamanis", "Alexander L. Gaunt", "Oleksandr Polozov"] | Generative models forsource code are an interesting structured prediction problem, requiring to reason about both hard syntactic and semantic constraints as well as about natural, likely programs. We present a novel model for this problem that uses a graph to represent the intermediate state of the generated output. Ou... | ["Generative Model", "Source Code", "Graph Learning"] | ABSTRACTGenerative models for source code are an interesting structured prediction problem,requiring to reason about both hard syntactic and semantic constraints as well asabout natural, likely programs. We present a novel model for this problem thatuses a graph to represent the intermediate state of the generated outp... | SyxeYOdu3Q | Overall good ideas, but not quite ready | 7: Good paper, accept | The paper proposes a code completion task that given the rest of a program, predicts the content of an expression. This task has similarity to code completion tasks in the code editor of an IDE. The paper proposes an interesting problem, but the paper would benefit if writing and evaluation are significantly improved.
... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Generative Code Modeling with Graphs
### Paper Abstract
Generative models forsource code are an interesting structured prediction problem, requiring to reason about both... |
yaeJLwvTr | graphicsinterface.org/Graphics_Interface/2020/Conference | 2020 | Interactive Shape Based Brushing Technique for Trail Sets | ["Almoctar Hassoumi", "Mar\u00eda Jes\u00fas Lobo", "Gabriel Jarry", "Vsevolod Peysakhovich", "Christophe Hurter"] | Brushing techniques have a long history with the first interactive selection tools appearing in the 1990s. Since then, many additional techniques have been developed to address selection accuracy, scalability and flexibility issues. Selection is especially difficult in large datasets where many visual items tangle and ... | ["Input Techniques", "Visualization", "Content Analysis", "Eye Tracking", "Interaction Design", "Trajectory data"] | ABSTRACTBrushing techniques have a long history with the first interactiveselection tools appearing in the 1990s. Since then, many additionaltechniques have been developed to address selection accuracy, scala-bility and flexibility issues. Selection is especially difficult in largedatasets where many visual items tangl... | 5kEgUyLuBT | Interesting idea and implementation, but lacks maturity | 5: Marginally below acceptance threshold | This paper presents a technique for brushing trajectories based on two different metrics, and provides case studies illustrating the efficiency of the technique.
I am not an expert in brushing per se, but I find the related work accessible enough for a non-expert to grasp the research landscape in the area. The autho... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Interactive Shape Based Brushing Technique for Trail Sets
### Paper Abstract
Brushing techniques have a long history with the first interactive selection tools appearing... |
SkxueFsiFV | MIDL.io/2019/Conference/Abstract | 2019 | Chest CT Super-resolution and Domain-adaptation using Memory-efficient 3D Reversible GANs | ["Tycho F.A. van der Ouderaa", "Daniel E. Worrall", "Bram van Ginneken"] | Medical imaging data are typically large in size. As a result, it can be difficult to train deep neural models on them. The activations of invertible neural networks do not have to be stored to perform backpropagation, therefore such networks can be used to save memory when handling large data volumes. We use a techniq... | ["Reversible", "GAN", "Chest CT", "3D", "Memory-efficient", "Super-resolution", "Domain-adaptation", "Image-to-image Translation"] | Medical Imaging with Deep Learning 2019 MIDL 2019 { Extended Abstract TrackChest CT Super-resolution and Domain-adaptation usingMemory-ecient 3D Reversible GANsTycho F. A. van der Ouderaa1;2tychovdo@gmail.comDaniel E. Worrall1d.e.worrall@uva.nlBram van Ginneken2bram.vanginneken@radboudumc.nl1University of Amsterdam, Am... | S1x7kcxUoE | Clear and consise study with promising prelimirary results | 3: Accept | Image-to-image translation in the context of super-resolution and domain-adaptation is performed by a 3D-RevGAN model. The authors add reversible blocks to the 3D-RevGAN model, without increasing memory cost. This property is demonstrated in the experiments, along with slightly improved results in the context of chest ... | 2: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Chest CT Super-resolution and Domain-adaptation using Memory-efficient 3D Reversible GANs
### Paper Abstract
Medical imaging data are typically large in size. As a resul... |
vtWagx1luOX | auai.org/UAI/2022/Workshop/CRL | 2022 | Causal Class Activation Maps for Weakly-Supervised Semantic Segmentation | ["Yiping Wang"] | Weakly-supervised semantic segmentation (WSSS) has emerged in recent years due to its appealing requirements for training data, i.e., with only image-level labels available as supervision. Most existing WSSS methods exploit the class activation maps (CAMs) as the seeds and generate the pseudo-pixel-level ground truth t... | ["cams", "confounders", "semantic segmentation", "wsss", "adjustment", "recent years due", "appealing requirements", "data"] | Causal Class Activation Maps for Weakly-Supervised Semantic SegmentationYiping Wang11Computer Science Dept., University of Waterloo, Waterloo, Ontario, CanadaAbstractWeakly-supervised semantic segmentation (WSSS)has emerged in recent years due to its appealing re-quirements for training data, i.e., with only image-leve... | jwlIHwakyB8 | Interesting paper on CAMs inspired by causality for WSSS, could benefit from more aligned and rigorous evaluation | 7: Good paper, accept | _**Contributions**_
The authors propose a new Causal Inference framework for producing Class Activation Maps (CAMs) in Weakly Supervised Semantic Segmentation (WSSS). The proposed approach aims to de-confound the classification network which produces CAMs (Class Activation Maps) via front door adjustment, hoping to pr... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Causal Class Activation Maps for Weakly-Supervised Semantic Segmentation
### Paper Abstract
Weakly-supervised semantic segmentation (WSSS) has emerged in recent years du... |
ELiYxj9JlyW | ICLR.cc/2021/Conference | 2021 | ME-MOMENTUM: EXTRACTING HARD CONFIDENT EXAMPLES FROM NOISILY LABELED DATA | ["Yingbin Bai", "Tongliang Liu"] | Examples that are close to the decision boundary—that we term hard examples, are essential to shaping accurate classifiers. Extracting confident examples has been widely studied in the community of learning with noisy labels. However, it remains elusive how to extract hard confident examples from the noisy training ... | ["label noise", "hard confident examples"] | ABSTRACTExamples that are close to the decision boundary—that we term hard examples ,are essential to shaping accurate classifiers. Extracting confident examples hasbeen widely studied in the community of learning with noisy labels. However, itremains elusive how to extract hard confident examples from the noisy traini... | svTd45rxxj1 | Potentially interesting work, but technical contribution is limited, and the method is not theoretically solid. | 4: Ok but not good enough - rejection | *quality*
The organization of this paper is barely satisfactory. The technique proposed to extract hard confident examples in this paper is not convincing, even though the experimental results seem promising.
*clarity*
It is not difficult to understand the proposed method, however, the Figures in this paper are somewh... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
ME-MOMENTUM: EXTRACTING HARD CONFIDENT EXAMPLES FROM NOISILY LABELED DATA
### Paper Abstract
Examples that are close to the decision boundary—that we term hard examples... |
r1xBk6NKwr | ICLR.cc/2020/Conference | 2020 | Learning audio representations with self-supervision | ["M. Tagliasacchi", "B. Gfeller", "F. de Chaumont Quitry", "D. Roblek"] | We explore self-supervision as a way to learn general purpose audio
representations. Specifically, we propose self-supervised tasks that exploit the
temporal context in the spectrogram domain. The temporal gap task estimates
the distance between two short audio segments extracted at random from the same
audio clip. The... | ["Audio representations", "self-supervised learning"] | ABSTRACTWe explore self-supervision as a way to learn general purpose audio representations.Specifically, we propose self-supervised tasks that exploit the temporal context inthe spectrogram domain. The TemporalGap task estimates the distance betweentwo short audio segments extracted at random from the same audio clip.... | BylIMFGh9B | Official Blind Review #1 | 3: Weak Reject | *What is this paper about?*
This paper proposes a new unsupervised (or self-supervised) learning method to get general purpose audio representations. It proposes three main self supervised tasks, two of them analogous to word2vec’s SkipGram and CBOW. The third task is related to predicting the distance between two aud... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Learning audio representations with self-supervision
### Paper Abstract
We explore self-supervision as a way to learn general purpose audio representations. Specifically... | |
HkeQ6ANYDB | ICLR.cc/2020/Conference | 2020 | Blending Diverse Physical Priors with Neural Networks | ["Yunhao Ba", "Guangyuan Zhao", "Achuta Kadambi"] | Rethinking physics in the era of deep learning is an increasingly important topic. This topic is special because, in addition to data, one can leverage a vast library of physical prior models (e.g. kinematics, fluid flow, etc) to perform more robust inference. The nascent sub-field of physics-based learning (PBL) studi... | ["Physics-based learning", "Physics-aware learning"] | ABSTRACTRethinking physics in the era of deep learning is an increasingly important topic.This topic is special because, in addition to data, one can leverage a vast libraryof physical prior models (e.g. kinematics, fluid flow, etc) to perform more ro-bust inference. The nascent sub-field of physics-based learning (PBL... | H1lojMTjKH | Official Blind Review #3 | 6: Weak Accept | === Overall comments ===
This paper proposes to generalize approaches to physics-based learning (PBL) by performing network architecture search (NAS) over elements from PBL models found in the literature. This entails including physical inputs to the network and the incorporation of new operations to the NAS. I think t... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Blending Diverse Physical Priors with Neural Networks
### Paper Abstract
Rethinking physics in the era of deep learning is an increasingly important topic. This topic is... | |
Hy1d-ebAb | ICLR.cc/2018/Conference | 2018 | Learning Deep Generative Models of Graphs | ["Yujia Li", "Oriol Vinyals", "Chris Dyer", "Razvan Pascanu", "Peter Battaglia"] | Graphs are fundamental data structures required to model many important real-world data, from knowledge graphs, physical and social interactions to molecules and proteins. In this paper, we study the problem of learning generative models of graphs from a dataset of graphs of interest. After learning, these models can b... | ["Generative Model of Graphs"] | ABSTRACTGraphs are fundamental data structures required to model many important real-world data, from knowledge graphs, physical and social interactions to moleculesand proteins. In this paper, we study the problem of learning generative models ofgraphs from a dataset of graphs of interest. After learning, these models... | S1crSKYgM | Learning deep generative models for graphs | 5: Marginally below acceptance threshold | The paper introduces a generative model for graphs. The three main decision functions in the sequential process are computed with neural nets. The neural nets also compute node embeddings and graph embeddings and the embeddings of the current graph are used to compute the decisions at time step T. The paper is well wri... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Learning Deep Generative Models of Graphs
### Paper Abstract
Graphs are fundamental data structures required to model many important real-world data, from knowledge grap... |
QnzSSoqmAvB | ICLR.cc/2021/Conference | 2021 | Playing Nondeterministic Games through Planning with a Learned Model | ["Thomas Willkens", "Jordan Pollack"] | The MuZero algorithm is known for achieving high-level performance on traditional zero-sum two-player games of perfect information such as chess, Go, and shogi, as well as visual, non-zero sum, single-player environments such as the Atari suite. Despite lacking a perfect simulator and employing a learned model of envir... | ["reinforcement learning", "alphazero", "muzero", "mcts", "planning", "search"] | ABSTRACTThe MuZero algorithm is known for achieving high-level performance on tradi-tional zero-sum two-player games of perfect information such as chess, Go, andshogi, as well as visual, non-zero sum, single-player environments such as theAtari suite. Despite lacking a perfect simulator and employing a learned modelof... | wKE8jZRintM | Clear and enjoyable paper, but difficult to motivate | 3: Clear rejection | This paper introduces NDMZ, short for nondeterministic MuZero, a deep reinforcement learning algorithm for model-based RL that doesn't use the rules of the game to perform search. The paper's contribution is mostly focused on describing how to construct the algorithm, and experimental results are provided at the end. A... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Playing Nondeterministic Games through Planning with a Learned Model
### Paper Abstract
The MuZero algorithm is known for achieving high-level performance on traditional... |
tnq_O52RVbR | ICLR.cc/2021/Conference | 2021 | SHADOWCAST: Controllable Graph Generation with Explainability | ["Wesley Joon-Wie Tann", "Ee-Chien Chang", "Bryan Hooi"] | We introduce the problem of explaining graph generation, formulated as controlling the generative process to produce desired graphs with explainable structures. By directing this generative process, we can explain the observed outcomes. We propose SHADOWCAST, a controllable generative model capable of mimicking network... | ["Controllable Graph Generation", "Explainability", "Conditional Generative Adversarial Network"] | ABSTRACTWe introduce the problem of explaining graph generation, formulated as control-ling the generative process to produce desired graphs with explainable structures.By directing this generative process, we can explain the observed outcomes. Wepropose S HADOW CAST, a controllable generative model capable of mimickin... | zmJjXVJNrVF | An insufficiently convincing graph generation method | 5: Marginally below acceptance threshold | This paper introduces a graph generation method that aims at mimicking the structure of a given graph but meanwhile maintaining the possibility of altering it ("controllable" in the authors' wording).
There appears a strong footprint of NetGAN (Bojchevski et al., 2018) in the proposed method. NetGAN learns a distribut... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
SHADOWCAST: Controllable Graph Generation with Explainability
### Paper Abstract
We introduce the problem of explaining graph generation, formulated as controlling the g... |
kSXh83gWWy | robot-learning.org/CoRL/2023/Conference | 2023 | Context-Aware Deep Reinforcement Learning for Autonomous Robotic Navigation in Unknown Area | ["Jingsong Liang", "Zhichen Wang", "Yuhong Cao", "Jimmy Chiun", "Mengqi Zhang", "Guillaume Adrien Sartoretti"] | Mapless navigation refers to a challenging task where a mobile robot must rapidly navigate to a predefined destination using its partial knowledge of the environment, which is updated online along the way, instead of a prior map of the environment. Inspired by the recent developments in deep reinforcement learning (DRL... | ["deep reinforcement learning", "mapless navigation", "context-aware decision-making"] | Context-Aware Deep Reinforcement Learning forAutonomous Robotic Navigation in Unknown AreaJingsong Liang∗Zhichen Wang∗Yuhong Cao†∗Jimmy ChiunMengqi Zhang Guillaume SartorettiNational University of Singapore{jingsongliang, zhichenwang, caoyuhong, jimmy.chiun }@u.nus.edu{mpezmq, guillaume.sartoretti }@nus.edu.sgAbstract:... | KfwX0UqnYR | Strengths
Strong Apparent Performance Characteristics
The comparative results between the proposed DRL method and classical methods in the literature displays the utility of learning-based methods in providing reasonable solutions to what would otherwise be complex, time consuming queries. The model is shown to be
su... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Context-Aware Deep Reinforcement Learning for Autonomous Robotic Navigation in Unknown Area
### Paper Abstract
Mapless navigation refers to a challenging task where a mo... | ||
rylVYjqHdN | ICLR.cc/2019/Workshop/LLD | 2019 | A Pseudo-Label Method for Coarse-to-Fine Multi-Label Learning with Limited Supervision | ["Cheng-Yu Hsieh", "Miao Xu", "Gang Niu", "Hsuan-Tien Lin", "Masashi Sugiyama"] | The goal of multi-label learning (MLL) is to associate a given instance with its relevant labels from a set of concepts. Previous works of MLL mainly focused on the setting where the concept set is assumed to be fixed, while many real-world applications require introducing new concepts into the set to meet new demands.... | ["Multi-Label Learning", "Weakly-Supervised Learning", "Pseudo-Labels", "Meta-Learning"] | ABSTRACTThe goal of multi-label learning (MLL) is to associate a given instance with itsrelevant labels from a set of concepts. Previous works of MLL mainly focused onthe setting where the concept set is assumed to be fixed, while many real-worldapplications require introducing new concepts into the set to meet new dem... | SkeUhEJVtV | Review: A Pseudo-Label Method for Coarse-to-Fine Multi-Label Learning with Limited Supervision | 2: Marginally below acceptance threshold | The authors propose and implement a new meta-learning approach for multi-label classification where the labels are structured in a two-level hierarchy (i.e. if an example has a label y, it has the label y* where y* is any ancestor of y in the hierarchy).
The text in Figure 2 is very difficult to see since the font siz... | 2: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
A Pseudo-Label Method for Coarse-to-Fine Multi-Label Learning with Limited Supervision
### Paper Abstract
The goal of multi-label learning (MLL) is to associate a given ... |
akc8f5ampp | graphicsinterface.org/Graphics_Interface/2023/Conference_SD | 2023 | Challenges and Opportunities for Software Testing in Virtual Reality Application Development | ["Qing Liu", "Gustavo Alves", "Jian Zhao"] | Testing is a core process for the development of Virtual Reality (VR) software, which could ensure the delivery of high-quality VR products and experiences. As VR applications have become more popular in different fields, more challenges and difficulties have been raised during the testing phase. However, few studies h... | ["Virtual Reality", "end-user development", "AR/VR Development", "software testing", "challenges"] | ABSTRACTTesting is a core process for the development of Virtual Reality(VR) software, which could ensure the delivery of high-quality VRproducts and experiences. As VR applications have become morepopular in different fields, more challenges and difficulties have beenraised during the testing phase. However, few studi... | SPCB9v-vN8Y | A good contribution to the VR/AR Community | 7: Good paper, accept | Software testing for the development of VR applications has challenges and has not been explored much. To address this, the authors conducted a qualitative interview study with 14 professional VR developers and validated the findings of the qualitative study with a second survey study with 33 new participants.
The pap... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Challenges and Opportunities for Software Testing in Virtual Reality Application Development
### Paper Abstract
Testing is a core process for the development of Virtual ... |
S1xKZwFpKN | MIDL.io/2019/Conference/Abstract | 2019 | Applying simultaneous super-resolution and contrast synthesis to routine clinical magnetic resonance images for automated segmentation of knee joint cartilage | ["Ales Neubert", "Pierrick Bourgeat", "Jason Wood", "Craig Engstrom", "Shekhar S. Chandra", "Stuart Crozier", "Jurgen Fripp"] | High resolution 3D MR images are well suited for automated cartilage segmentation in the human knee joint. However, volumetric scans such as 3D Double-Echo Steady-State (DESS) images, are not routinely acquired. Instead, typical clinical knee MR imaging exams involve acquisition of a series of 2D turbo spin echo (TSE) ... | ["super-resolution", "contrast synthesis", "knee cartilage segmentation"] | Proceedings of Machine Learning Research { XXXX:1{4, 2019 Extended Abstract { MIDL 2019 submissionApplying simultaneous super-resolution and contrastsynthesis to routine clinical magnetic resonance images forautomated segmentation of knee joint cartilage: Data fromthe Osteoarthritis InitiativeAles Neubert1ales.neubert@... | r1l-13lVsN | Good performance but motivation for the proposed approach is unclear | 2: Reject | The paper proposes contrast-synthesis as an intermediate step for their automated segmentation tool, in which they achieve great performance improvement. However, I have a few concerns: firstly, since the network had to “synthesize” cartilage, it is very important to characterize the bias of the network (i.e. does the ... | 2: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Applying simultaneous super-resolution and contrast synthesis to routine clinical magnetic resonance images for automated segmentation of knee joint cartilage
### Paper ... |
S1lDkaEFwS | ICLR.cc/2020/Conference | 2020 | Thwarting finite difference adversarial attacks with output randomization | ["Haidar Khan", "Dan Park", "Azer Khan", "B\u00fclent Yener"] | Adversarial input poses a critical problem to deep neural networks (DNN). This problem is more severe in the "black box" setting where an adversary only needs to repeatedly query a DNN to estimate the gradients required to create adversarial examples. Current defense techniques against attacks in this setting are not... | ["black box adversarial attacks", "adversarial examples", "defense", "deep learning"] | ABSTRACTAdversarial input poses a critical problem to deep neural networks (DNN). Thisproblem is more severe in the "black box" setting where an adversary only needsto repeatedly query a DNN to estimate the gradients required to create adversarialexamples. Current defense techniques against attacks in this setting are ... | SkglhWCXcB | Official Blind Review #2 | 3: Weak Reject | This paper proposes to introduce randomness in a classifier’s predictions to mitigate black-box attacks that rely on gradient estimation through finite differences. The intuition behind the defense is correct: finite differences rely on the outputs of the neural network being non-deterministic and accurate to estimate ... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Thwarting finite difference adversarial attacks with output randomization
### Paper Abstract
Adversarial input poses a critical problem to deep neural networks (DNN). ... | |
rkl_Ch4YwS | ICLR.cc/2020/Conference | 2020 | A TWO-STAGE FRAMEWORK FOR MATHEMATICAL EXPRESSION RECOGNITION | ["Jin Zhang", "Weipeng Ming", "Pengfei Liu"] |
Although mathematical expressions (MEs) recognition have achieved great progress, the development of MEs recognition in real scenes is still unsatisfactory. Inspired by the recent work of neutral network, this paper proposes a novel two-stage approach which takes a printed mathematical expression image as input and ge... | ["mathematical expressions recognition", "seq2seq model"] | ABSTRACTAlthough mathematical expressions (MEs) recognition have achieved greatprogress, the development of MEs recognition in real scenes is still unsatisfac-tory. Inspired by the recent work of neutral network, this paper proposes a noveltwo-stage approach which takes a printed mathematical expression image as inputa... | r1lT36Q5cS | Official Blind Review #2 | 6: Weak Accept | This paper proposes to recognise a mathmatical expression using a two-stage framework, including object detection by YOLOv3, and encoder-decoder based translation. The paper is written well, and easy to read. In the experiments, the recognition and translation methods both work well on Homologous and non-Homologous tes... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
A TWO-STAGE FRAMEWORK FOR MATHEMATICAL EXPRESSION RECOGNITION
### Paper Abstract
Although mathematical expressions (MEs) recognition have achieved great progress, the d... | |
cI8WHuiW3x | MICCAI.org/2022/Challenge/FLARE | 2022 | A Simple Self-labeling Method for Semi-supervised Medical Image Segmentation | ["Ye Zhu"] | Leveraging a few labeled images and a large number of unlabeled images is crucial for medical image segmentation since labeling the medical data can be very expensive and time-consumed. Therefore we introduce a naïve but simple method to utilize the massive unlabeled
medical images for better training. We first use all... | ["Semi-supervise learning Pseudo labels"] | A Simple Self-labeling Method forSemi-supervised Medical Image SegmentationYe Zhu1and Hanlin Tian1The Chinese University of Hong Kong (Shenzhen), Chinazhuye1@cuhk.edu.cnAbstract. Leveraging a few labeled images and a large number of un-labeled images is crucial for medical image segmentation since labelingthe medical d... | m1GDl-hbT59 | The authors propose a self-labeling method that trains a Swin-Unet for segmenting the test set | 5: Marginally below acceptance threshold | Strengths: Simple and easy to implement method.
Weaknesses: The proposed method should be described in better detail. There is lack of discussion of the model and efficiency results.
Details:
- The metrics in the validation set are not included in the abstract
- Grammatical and spelling mistakes.
- There is no ex... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
A Simple Self-labeling Method for Semi-supervised Medical Image Segmentation
### Paper Abstract
Leveraging a few labeled images and a large number of unlabeled images is... |
Ig53hpHxS4 | ICLR.cc/2021/Conference | 2021 | Flowtron: an Autoregressive Flow-based Generative Network for Text-to-Speech Synthesis | ["Rafael Valle", "Kevin J. Shih", "Ryan Prenger", "Bryan Catanzaro"] | In this paper we propose Flowtron: an autoregressive flow-based generative network for text-to-speech synthesis with style transfer and speech variation. Flowtron borrows insights from Autoregressive Flows and revamps Tacotron 2 in order to provide high-quality and expressive mel-spectrogram synthesis. Flowtron is opti... | ["Text to speech synthesis", "normalizing flows", "deep learning"] | ABSTRACTIn this paper we propose Flowtron: an autoregressive flow-based generative networkfor text-to-speech synthesis with style transfer and speech variation. Flowtronborrows insights from Autoregressive Flows and revamps Tacotron 2 in orderto provide high-quality and expressive mel-spectrogram synthesis. Flowtron is... | RHN50cYZFEb | Accept. The idea is good and experiments are good. There are some concerns about the clarity of the paper but those can be worked on. | 9: Top 15% of accepted papers, strong accept | ### Summary
This paper introduces a novel application of normalizing flows to speech synthesis, allowing direct optimization of spectrogram log-likelihoods which results in more natural variation at inference compared to L1/L2 losses that model the mean. This setup also allows more control over non-textual information... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Flowtron: an Autoregressive Flow-based Generative Network for Text-to-Speech Synthesis
### Paper Abstract
In this paper we propose Flowtron: an autoregressive flow-based... |
D42ulHk5CU | NoDaLiDa/2023/Conference | 2023 | Identifying Token-Level Dialectal Features in Social Media | ["Jeremy Barnes", "Samia Touileb", "Petter M\u00e6hlum", "Pierre Lison"] | Dialectal variation is present in many human languages and is attracting a growing interest in NLP. Most previous work concentrated on either (1) classifying dialectal varieties at the document or sentence level or (2) performing standard NLP tasks on dialectal data. In this paper, we propose the novel task of token-le... | ["dialectal features", "token-level dialectal features", "dialect identification"] | Identifying Token-Level Dialectal Features in Social MediaJeremy Barnes1, Samia Touileb2, Petter Mæhlum3, Pierre Lison41University of the Basque Country,2University of Bergen,3University of Oslo,4Norwegian Computing Centerjeremy.barnes@ehu.eus, samia.touileb@uib.no,pettemae@ifi.uio.no, plison@nr.noAbstractDialectal var... | 7wDz4FFfoYi | A corpus of Norwegian tweets annotated on token level with dialect features and a series of experiments on the associated task of dialect feature identification | 8: Top 50% of accepted papers, clear accept | This paper presents a corpus of Norwegian tweets which are annotated with dialect features on token level. Several experiments are conducted on the associated task of dialect feature identification (i.e. not dialect identification), and the task is found to be challenging because of its multi-label nature and sparsity ... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Identifying Token-Level Dialectal Features in Social Media
### Paper Abstract
Dialectal variation is present in many human languages and is attracting a growing interest... |
iYbJn-wIvf | MIDL.io/2023/Short_Paper_Track | 2023 | Deep model-based optoacoustic image reconstruction (DeepMB) | ["Christoph Dehner", "Guillaume Zahnd", "Vasilis Ntziachristos", "Dominik J\u00fcstel"] | Multispectral optoacoustic tomography requires image feedback in real-time to locate and identify relevant tissue structures during clinical interventions. Backprojection methods are commonly used for optoacoustic image reconstruction in real-time but only afford imprecise images due to oversimplified modelling assumpt... | ["Multispectral Optoacoustic Tomography (MSOT)", "Model-based reconstruction", "Inverse problems", "Real-time imaging", "Synthesized training data"] | Medical Imaging with Deep Learning 2023Deep model-based optoacoustic image reconstruction(DeepMB)Christoph Dehner∗;1;2christoph.dehner@ithera-medical.comGuillaume Zahnd∗;1;2guillaume.zahnd@ithera-medical.comVasilis Ntziachristos†;1;3;4bioimaging.translatum@tum.deDominik J ustel†;1;3;5dominik.justel@helmholtz-munich.de1... | aQlhxdb-aci | Interesting solution to a real-world medical imaging problem | 8: Top 50% of accepted papers, clear accept | This short paper describes a deep-learning approach for optoacoustic image reconstruction in real time. Authors train the method using abundant natural images and validate using in-vivo images. As a reference, a state-of-the-art but slow model-based reconstruction method is used. Because there is no other ground truth,... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Deep model-based optoacoustic image reconstruction (DeepMB)
### Paper Abstract
Multispectral optoacoustic tomography requires image feedback in real-time to locate and i... |
SJeUm1HtDH | ICLR.cc/2020/Conference | 2020 | Swoosh! Rattle! Thump! - Actions that Sound | ["Dhiraj Gandhi", "Abhinav Gupta", "Lerrel Pinto"] | Truly intelligent agents need to capture the interplay of all their senses to build a rich physical understanding of their world. In robotics, we have seen tremendous progress in using visual and tactile perception; however we have often ignored a key sense: sound. This is primarily due to lack of data that captures th... | ["Sound", "Action", "Audio Representations"] | ABSTRACTTruly intelligent agents need to capture the interplay of all their senses to build arich physical understanding of their world. In robotics, we have seen tremendousprogress in using visual and tactile perception; however we have often ignored akey sense: sound. This is primarily due to lack of data that captur... | Hyxux2VH9r | Official Blind Review #3 | 3: Weak Reject | This paper presents audio-visual object classification and motion prediction work on a novel dataset of 60 different objects rolling around in a bin tilted to and fro by a robot, with video and 4-channel audio recordings of the object impacts. The data is rather novel, is large enough to do ML (around 17 hours of eve... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Swoosh! Rattle! Thump! - Actions that Sound
### Paper Abstract
Truly intelligent agents need to capture the interplay of all their senses to build a rich physical unders... | |
csHPh2kuVP | NeurIPS.cc/2021/Workshop/DLDE | 2021 | Actor-Critic Algorithm for High-dimensional PDEs | ["Xiaohan Zhang"] | We develop a machine learning model to effectively solve high-dimensional nonlinear parabolic partial differential equations (PDE). We use Feynman-Kac formula to reformulate PDE into the equivalent stochastic control problem governed by a Backward Stochastic Differential Equation (BSDE) system. Our model is designed to... | ["Partial Differential Equations", "Reinforcement learning", "Actor-critic"] | Actor-Critic Algorithm for High-dimensional PDEsAnonymous Author(s)AffiliationAddressemailAbstractWe develop a deep neural network model to solve high-dimensional nonlinear 1parabolic partial differential equations (PDE). Our model extends and advances 2the DBSDE model in the following 3 aspects: 1) the trainable param... | 9zHmAwCRcde | Review | This paper proposes to solve high-dimensional nonlinear parabolic partial differential equations by using an Actor-Critic neural network architecture. The performance benefits are well articulated although a bit repetitive.
The paper lacks background and literature review, with a corresponding lack of references.
The... | 2: The reviewer is willing to defend the evaluation, but it is quite likely that the reviewer did not understand central parts of the paper | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Actor-Critic Algorithm for High-dimensional PDEs
### Paper Abstract
We develop a machine learning model to effectively solve high-dimensional nonlinear parabolic partial... | |
esnzokeU0L3 | MICCAI.org/2022/Challenge/FLARE | 2022 | Self-training with Selective Re-training Improves Abdominal Organ Segmentation in CT Image | ["Fan Zhang", "Meihuan Wang", "Hua Yang"] | Abstract. Inspired by self-training learning via pseudo labeling, we construct self-training framework with selective re-training pseudo labels to improve semi-supervised abdominal organ segmentation. In this work, we carefully design the strong data augmentations (SDA) and test-time augmentations (TTA) to alleviate o... | ["Self-training \u00b7 Efficient segmentation learning \u00b7 Abdominal organ segmentation."] | Self-training with Selective Re-training ImprovesAbdominal Organ Segmentation in CT ImageFan Zhang1, Meihuan Wang2, and Hua Yang1()1Department of Radiological Algorithm, Fosun Aitrox Information Technology Co.,LTD., Shanghai, China.2College of Medicine and Biological Information Engineering, NortheasternUniversity, She... | aKrQBqrF36p | Self-training with Selective Re-training | 10: Top 5% of accepted papers, seminal paper | The quality, clarity, and description of the paper are good except for some minor issues.
* There is no qualitative analysis in section 4.2 (Qualitative results on validation set).
* The format of the table is different.
| 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Self-training with Selective Re-training Improves Abdominal Organ Segmentation in CT Image
### Paper Abstract
Abstract. Inspired by self-training learning via pseudo lab... |
3ht_dupP1b4 | vivoconference.org/VIVO/2021/Conference | 2021 | 30% of our Visits are Mobile, But We're Not Ready! Two Simple Ways We Worked Around the Problem | ["Anirvan Chatterjee", "Brian Turner", "Eric Meeks"] | The University of California, San Francisco has operated the UCSF Profiles RIS platform (https://profiles.ucsf.edu/) for over a decade.
In that time, the share of mobile visits has grown from 3% to 30%.
Unfortunately, our platform, Profiles RNS, was designed solely for desktop users, so mobile users would typically g... | ["UI/UX Design", "Web Analytics", "User Engagement"] | YdRK_Ei8cE8 | Differences to earlier presentation? | 6: Marginally above acceptance threshold | AMP was already suggested as a suggestion for user experience. The authors should make clear what's new to this submission. Additionally there are big concerns about AMP being a user tracking device, invading the user's privacy. In the light of https://stoptrackingscience.eu/ the authors should explain if and why perfo... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
30% of our Visits are Mobile, But We're Not Ready! Two Simple Ways We Worked Around the Problem
### Paper Abstract
The University of California, San Francisco has operat... | |
B1gXWCVtvr | ICLR.cc/2020/Conference | 2020 | Adapting Behaviour for Learning Progress | ["Tom Schaul", "Diana Borsa", "David Ding", "David Szepesvari", "Georg Ostrovski", "Will Dabney", "Simon Osindero"] | Determining what experience to generate to best facilitate learning (i.e. exploration) is one of the distinguishing features and open challenges in reinforcement learning. The advent of distributed agents that interact with parallel instances of the environment has enabled larger scale and greater flexibility, but has ... | ["adaptation", "behaviour", "reinforcement learning", "modulated behaviour", "exploration", "deep reinforcement learning"] | ABSTRACTDetermining what experience to generate to best facilitate learning (i.e. explo-ration) is one of the distinguishing features and open challenges in reinforcementlearning. The advent of distributed agents that interact with parallel instances ofthe environment has enabled larger scale and greater flexibility, b... | BJl2JOl9cr | Official Blind Review #4 | 3: Weak Reject | This in an interesting paper as it tries to alleviate the burden of hyper-parameters tuning for exploration strategies Deep Reinforcement learning.
The paper proposes an adaptive behaviour in order to shape the data generation process for effective learning. The paper considers a behaviour policy that is parametrized b... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Adapting Behaviour for Learning Progress
### Paper Abstract
Determining what experience to generate to best facilitate learning (i.e. exploration) is one of the distingu... | |
upQogJXuhQ | MIDL.io/2023/Short_Paper_Track | 2023 | Active learning for medical image segmentation with stochastic batches | ["M\u00e9lanie Gaillochet", "Christian Desrosiers", "Herve Lombaert"] | Active learning (AL) selects informative samples for annotation. This is becoming increasingly crucial to medical image segmentation since image annotation is hardly scalable to full pixel-level labeling of large datasets. However, most research focuses on classification or natural image segmentation. Uncertainty-based... | ["Active learning", "Segmentation", "Medical image analysis", "Uncertainty"] | Medical Imaging with Deep Learning 2023Active learning for medical image segmentation withstochastic batchesM ́ elanie Gaillochet1melanie.gaillochet.1@ens.etsmtl.caChristian Desrosiers1christian.desrosiers@etsmtl.caHerv ́ e Lombaert1herve.lombaert@etsmtl.ca1ETS Montr ́ ealAbstractActive learning (AL) selects informativ... | ImuYBg84WUa | simple (and/but) good results | 7: Good paper, accept | The authors propose an active learning strategy for medical image segmentation (a field that is still relatively unexplored in the deep learning era) based on computing uncertaintty at the batch level - in order to mitigate the problems associated with potentially selecting outliers, when operating at the sample level.... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Active learning for medical image segmentation with stochastic batches
### Paper Abstract
Active learning (AL) selects informative samples for annotation. This is becomi... |
S1vuO-bCW | ICLR.cc/2018/Conference | 2018 | Leave no Trace: Learning to Reset for Safe and Autonomous Reinforcement Learning | ["Benjamin Eysenbach", "Shixiang Gu", "Julian Ibarz", "Sergey Levine"] | Deep reinforcement learning algorithms can learn complex behavioral skills, but real-world application of these methods requires a considerable amount of experience to be collected by the agent. In practical settings, such as robotics, this involves repeatedly attempting a task, resetting the environment between each a... | ["manual reset", "continual learning", "reinforcement learning", "safety"] | ABSTRACTDeep reinforcement learning algorithms can learn complex behavioral skills, butreal-world application of these methods requires a large amount of experience to becollected by the agent. In practical settings, such as robotics, this involves repeat-edly attempting a task, resetting the environment between each a... | SyqiSbomf | Great idea but the write up needs to be made clearer | 7: Good paper, accept | This paper proposes the idea of having an agent learning a policy that resets the agent's state to one of the states drawn from the distribution of starting states. The agent learns such policy while also learning how to solve the actual task. This approach generates more autonomous agents that require fewer human inte... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Leave no Trace: Learning to Reset for Safe and Autonomous Reinforcement Learning
### Paper Abstract
Deep reinforcement learning algorithms can learn complex behavioral s... |
Hkgpnn4YvH | ICLR.cc/2020/Conference | 2020 | Graph Neural Networks For Multi-Image Matching | ["Stephen Phillips", "Kostas Daniilidis"] | In geometric computer vision applications, multi-image feature matching gives more accurate and robust solutions compared to simple two-image matching. In this work, we formulate multi-image matching as a graph embedding problem, then use a Graph Neural Network to learn an appropriate embedding function for aligning im... | ["Graph Neural Networks", "Multi-image Matching"] | ABSTRACTIn geometric computer vision applications, multi-image feature matching givesmore accurate and robust solutions compared to simple two-image matching. Inthis work, we formulate multi-image matching as a graph embedding problem,then use a Graph Neural Network to learn an appropriate embedding function foralignin... | S1egIL32FS | Official Blind Review #3 | 3: Weak Reject | The paper presents graph neural network approach for learning multi-view feature similarity.
The input data is local feature descriptors (SIFT), keypoint location, orientation and scale. The objective is to learn embedding such that features corresponding to the same 3d location will have similar embeddings, while dif... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Graph Neural Networks For Multi-Image Matching
### Paper Abstract
In geometric computer vision applications, multi-image feature matching gives more accurate and robust ... | |
Bklh0SiQiN | ICML.cc/2019/Workshop/RL4RealLife | 2019 | Contextual Markov Decision Processes using Generalized Linear Models | ["Aditya Modi", "Ambuj Tewari"] | We consider the recently proposed reinforcement learning (RL) framework of Contextual Markov Decision Processes (CMDP), where the agent has a sequence of episodic interactions with tabular environments chosen from a possibly infinite set. The parameters of these environments depend on a context vector that is available... | ["Reinforcement Learning", "Markov Decision Processes", "Online Learning", "Regret Analysis", "Contextual Learning"] | Contextual Markov Decision Processes using Generalized Linear ModelsAditya Modi1Ambuj Tewari2AbstractWe consider the recently proposed reinforcementlearning (RL) framework of Contextual MarkovDecision Processes (CMDP), where the agent hasa sequence of episodic interactions with tabularenvironments chosen from a possibl... | HkgMf9Tr6E | The paper generalizes previous contextual MDP models, develops an order-optimal algorithm leveraging known algorithms, and proves its performance bound. NOt sure how suitable the venue is given the purely theoretical nature of the work. | 4: Weak Accept | The authors study finite horizon contextual Markov Decision Process (MDP) problems using generalized linear models. In particular, at each episode, a context is given a priori. The transition matrix of the MDP is assumed to follow a generalized linear model for the given context. The model is shown to be general that s... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Contextual Markov Decision Processes using Generalized Linear Models
### Paper Abstract
We consider the recently proposed reinforcement learning (RL) framework of Contex... |
stOiAztdcH | ICML.cc/2020/Workshop/SAS | 2020 | Bootstrapping Unsupervised Deep Music Separation from Primitive Auditory Grouping Principles | ["Prem Seetharaman", "Gordon Wichern", "Jonathan Le Roux", "Bryan Pardo"] | Separating an audio scene, such as a cocktail party with multiple overlapping voices, into meaningful components (e.g., individual voices) is a core task in computer audition, analogous to image segmentation in computer vision. Deep networks are the state-of-the-art approach. They are typically trained on synthetic aud... | ["auditory scene analysis", "deep learning", "cocktail party problem", "self-supervised learning", "bootstrapping"] | Bootstrapping Unsupervised Deep Music Separation fromPrimitive Auditory Grouping PrinciplesPrem Seetharaman1Gordon Wichern2Jonathan Le Roux2Bryan Pardo1AbstractSeparating an audio scene, such as a cocktail partywith multiple overlapping voices, into meaningfulcomponents (e.g., individual voices) is a core taskin comput... | 7vHhzLoThAR | Novel idea with minor details missing | 7: Good paper, accept | The paper introduces a novel unsupervised method for music separation, which consists of a labeler, a teacher, and a student. By leveraging bootstrapping, the model performance on the MUSDB test set is promising.
Pros:
- The idea is novel, which encourages the usage of bootstrapping and unsupervised training for soun... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Bootstrapping Unsupervised Deep Music Separation from Primitive Auditory Grouping Principles
### Paper Abstract
Separating an audio scene, such as a cocktail party with ... |
ByloIiCqYQ | ICLR.cc/2019/Conference | 2019 | Maximal Divergence Sequential Autoencoder for Binary Software Vulnerability Detection | ["Tue Le", "Tuan Nguyen", "Trung Le", "Dinh Phung", "Paul Montague", "Olivier De Vel", "Lizhen Qu"] | Due to the sharp increase in the severity of the threat imposed by software vulnerabilities, the detection of vulnerabilities in binary code has become an important concern in the software industry, such as the embedded systems industry, and in the field of computer security. However, most of the work in binary code vu... | ["Vulnerabilities Detection", "Sequential Auto-Encoder", "Separable Representation"] | ABSTRACTDue to the sharp increase in the severity of the threat imposed by softwarevulnerabilities, the detection of vulnerabilities in binary code has become animportant concern in the software industry, such as the embedded systemsindustry, and in the field of computer security. However, most of the works inbinary co... | S1l--zJuTm | Using deep neural networks to learn embeddings for binary software vulnerability detection; unclear about its contribution for ML community | 6: Marginally above acceptance threshold | This paper proposes a variational autoencoder-based architecture for binary code embedding. For evaluation, they construct a dataset by compiling source code in the NDSS18 dataset. They evaluate their approaches against several neural network baselines, and demonstrate that their learned embeddings are more effective a... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Maximal Divergence Sequential Autoencoder for Binary Software Vulnerability Detection
### Paper Abstract
Due to the sharp increase in the severity of the threat imposed ... |
IrofNLZuWF | ICLR.cc/2021/Conference | 2021 | Stochastic Optimization with Non-stationary Noise: The Power of Moment Estimation | ["Jingzhao Zhang", "Hongzhou Lin", "Subhro Das", "Suvrit Sra", "Ali Jadbabaie"] | We investigate stochastic optimization under weaker assumptions on the distribution of noise than those used in usual analysis. Our assumptions are motivated by empirical observations in training neural networks. In particular, standard results on optimal convergence rates for stochastic optimization assume either the... | ["Stochastic optimization"] | ABSTRACTWe investigate stochastic optimization under weaker assumptions on the distribu-tion of noise than those used in usual analysis. Our assumptions are motivated byempirical observations in training neural networks. In particular, standard resultson optimal convergence rates for stochastic optimization assume eith... | bEBhBghH5Iy | New attempt to handle non-stationary gradient noise | 3: Clear rejection | ## Summary
The authors provide a new analysis of SGD and versions of RMSprop, taking into account possible non-stationarity of the gradient noise. In particular, the authors propose.
(i) the convergence analysis of SGD with stepsizes dependent on the second moment of stochastic gradients and a "norm" version of RMSpr... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Stochastic Optimization with Non-stationary Noise: The Power of Moment Estimation
### Paper Abstract
We investigate stochastic optimization under weaker assumptions on t... |
rkmDI85ge | ICLR.cc/2017/conference | 2017 | Efficient Softmax Approximation for GPUs | ["\u00c9douard Grave", "Armand Joulin", "Moustapha Ciss\u00e9", "David Grangier", "Herv\u00e9 J\u00e9gou"] | We propose an approximate strategy to efficiently train neural network based language models over very large vocabularies. Our approach, called adaptive softmax, circumvents the linear dependency on the vocabulary size by exploiting the unbalanced word distribution to form clusters that explicitly minimize the expectat... | ["Natural language processing"] | ABSTRACTWe propose an approximate strategy to efficiently train neural network basedlanguage models over very large vocabularies. Our approach, called adap-tive softmax, circumvents the linear dependency on the vocabulary size by ex-ploiting the unbalanced word distribution to form clusters that explicitly min-imize th... | B1mHNfPEx | Final Review: nice practical speed optimization of softmax for GPUs | 7: Good paper, accept | The authors introduce an adaptive softmax approximation tailored for faster performance on GPUs. The key idea, which is very sensible, is to use a class-based hierarchical softmax, but where the clusters/hierarchy are distributed such that the resulting matrix multiplications are optimally-sized for GPU computation, ba... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Efficient Softmax Approximation for GPUs
### Paper Abstract
We propose an approximate strategy to efficiently train neural network based language models over very large ... |
H1gsz30cKX | ICLR.cc/2019/Conference | 2019 | Fixup Initialization: Residual Learning Without Normalization | ["Hongyi Zhang", "Yann N. Dauphin", "Tengyu Ma"] | Normalization layers are a staple in state-of-the-art deep neural network architectures. They are widely believed to stabilize training, enable higher learning rate, accelerate convergence and improve generalization, though the reason for their effectiveness is still an active research topic. In this work, we challenge... | ["deep learning", "residual networks", "initialization", "batch normalization", "layer normalization"] | ABSTRACTNormalization layers are a staple in state-of-the-art deep neural network archi-tectures. They are widely believed to stabilize training, enable higher learningrate, accelerate convergence and improve generalization, though the reason fortheir effectiveness is still an active research topic. In this work, we ch... | SyeHla6K37 | Interesting, but unsure about the impact | 7: Good paper, accept | This paper proposes an exploration of the effect of normalization and initialization in residual networks. In particular, the Authors propose a novel way to initialize residual networks, which is motivated by the need to avoid exploding/vanishing gradients. The paper proposes some theoretical analysis of the benefits o... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Fixup Initialization: Residual Learning Without Normalization
### Paper Abstract
Normalization layers are a staple in state-of-the-art deep neural network architectures.... |
SJ8BZTjeg | ICLR.cc/2017/conference | 2017 | Unsupervised Learning Using Generative Adversarial Training And Clustering | ["Vittal Premachandran", "Alan L. Yuille"] | In this paper, we propose an unsupervised learning approach that makes use of two components; a deep hierarchical feature extractor, and a more traditional clustering algorithm. We train the feature extractor in a purely unsupervised manner using generative adversarial training and, in the process, study the strengths ... | ["generative adversarial training", "unsupervised", "clustering", "adversarial training", "unsupervised learning", "use", "components", "traditional clustering algorithm", "feature extractor"] | ABSTRACTIn this paper, we propose an unsupervised learning approach that makes use of twocomponents; a deep hierarchical feature extractor, and a more traditional cluster-ing algorithm. We train the feature extractor in a purely unsupervised mannerusing generative adversarial training and, in the process, study the str... | BkUsyJGEl | review | 3: Clear rejection | This paper proposed an unsupervised learning method based on running kmeans on the features learned by a discriminator network in a generative adversarial network setup. Unsupervised learning methods with GANs is certainly a relevant topic but this paper does not propose anything particularly novel as far as I can tell... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Unsupervised Learning Using Generative Adversarial Training And Clustering
### Paper Abstract
In this paper, we propose an unsupervised learning approach that makes use ... |
o82EXEK5hu6 | robot-learning.org/CoRL/2023/Conference | 2023 | Parting with Misconceptions about Learning-based Vehicle Motion Planning | ["Daniel Dauner", "Marcel Hallgarten", "Andreas Geiger", "Kashyap Chitta"] | The release of nuPlan marks a new era in vehicle motion planning research, offering the first large-scale real-world dataset and evaluation schemes requiring both precise short-term planning and long-horizon ego-forecasting. Existing systems struggle to simultaneously meet both requirements. Indeed, we find that these ... | ["Motion Planning", "Autonomous Driving", "Data-driven Simulation"] | Parting with Misconceptions aboutLearning-based Vehicle Motion PlanningDaniel Dauner1,2Marcel Hallgarten1,3Andreas Geiger1,2Kashyap Chitta1,21University of Tübingen2Tübingen AI Center3Robert Bosch GmbHhttps://github.com/autonomousvision/tuplan_garageAbstract: The release of nuPlan marks a new era in vehicle motion pl... | 1-2ygssPtUA | Strengths
-The paper is well-written and easy to follow with nice explanations of the technical components and a helpful discussion of results.
-The empirical study is quite extensive, containing ablation studies and in-depth discussions of the performance of the approaches evaluated.
-The findings are quite intrigu... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Parting with Misconceptions about Learning-based Vehicle Motion Planning
### Paper Abstract
The release of nuPlan marks a new era in vehicle motion planning research, of... | ||
-757TnNDwIn | ICLR.cc/2021/Conference | 2021 | Generative Adversarial Neural Architecture Search with Importance Sampling | ["SEYED SAEED CHANGIZ REZAEI", "Fred X. Han", "Di Niu", "Mohammad Salameh", "Keith G Mills", "Shangling Jui"] | Despite the empirical success of neural architecture search (NAS) in deep learning applications, the optimality, reproducibility and cost of NAS schemes remain hard to assess. The variation in search spaces adopted has further affected a fair comparison between search strategies. In this paper, we focus on search strat... | ["Nueral Architecture Search", "Deep Learning", "Generative Adversarial Network", "Graph Neural Network", "Computer Vision"] | ABSTRACTDespite the empirical success of neural architecture search (NAS) in deep learningapplications, the optimality, reproducibility and cost of NAS schemes remain hardto assess. The variation in search spaces adopted has further affected a fair compar-ison between search strategies. In this paper, we focus on searc... | 9Z-a9SdW3XV | A flexible but complex and expensive NAS method. | 5: Marginally below acceptance threshold | A flexible but complex and expensive NAS method.
Summary:
The authors introduce a method for NAS that repeatedly trains a generator to sample candidate architectures. The method is evaluated on three NAS oracle benchmarks as well as constrained NAS settings. While there are some promising experimental results, I lean ... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Generative Adversarial Neural Architecture Search with Importance Sampling
### Paper Abstract
Despite the empirical success of neural architecture search (NAS) in deep l... |
r1gfweBFPB | ICLR.cc/2020/Conference | 2020 | Learning by shaking: Computing policy gradients by physical forward-propagation | ["Arash Mehrjou", "Ashkan Soleymani", "Stefan Bauer", "Bernhard Sch\u00f6lkopf"] | Model-free and model-based reinforcement learning are two ends of a spectrum. Learning a good policy without a dynamic model can be prohibitively expensive. Learning the dynamic model of a system can reduce the cost of learning the policy, but it can also introduce bias if it is not accurate. We propose a middle ground... | ["Reinforcement Learning", "Control Theory"] | ABSTRACTModel-free and model-based reinforcement learning are two ends of a spectrum.Learning a good policy without a dynamic model can be prohibitively expensive.Learning the dynamic model of a system can reduce the cost of learning the pol-icy, but it can also introduce bias if it is not accurate. We propose a middle... | S1eCOIEOqS | Official Blind Review #4 | 1: Reject | The paper proposes learning physical derivatives, the derivative of the trajectory distribution with respect to policy parameters. The proposed method estimates changes in trajectories at a particular theta by using finite differences,
then fitting Gaussian Processes per timestep to generalize to new dtheta's. The pap... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Learning by shaking: Computing policy gradients by physical forward-propagation
### Paper Abstract
Model-free and model-based reinforcement learning are two ends of a sp... | |
om1guSP_ray | ICLR.cc/2021/Conference | 2021 | Graph Pooling by Edge Cut | ["Alexis Galland", "marc lelarge"] | Graph neural networks (GNNs) are very efficient at solving several tasks in graphs such as node classification or graph classification. They come from an adaptation of convolutional neural networks on images to graph structured data. These models are very effective at finding patterns in images that can discriminate im... | ["graph", "deep", "learning", "pooling"] | ABSTRACTGraph neural networks (GNNs) are very efficient at solving several tasks in graphssuch as node classification or graph classification. They come from an adaptationof convolutional neural networks on images to graph structured data. These modelsare very effective at finding patterns in images that can discrimina... | K58H8p9iIec | Straightforward idea and weak empirical results | 3: Clear rejection | Summary:
This paper presents a graph pooling operator by first predicting scores on edges, then performing min-cut to separate subgraphs, and finally construct coarsened graphs. Authors perform experiments on several small datasets to verify their claims.
Pros:
1, The problem of defining pooling operators on graphs ... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Graph Pooling by Edge Cut
### Paper Abstract
Graph neural networks (GNNs) are very efficient at solving several tasks in graphs such as node classification or graph clas... |
3HQMxxEmiJn | thecvf.com/ECCV/2020/Workshop/BIC | 2020 | DoubleU-Net: Colorectal Cancer Diagnosis and Gland Instance Segmentation with Text-Guided Feature Control | ["Pei WANG", "Albert Chung"] | With the rapid therapeutic advancement in personalized medicine, the role of pathologists for colorectal cancer has greatly expanded from morphologists to clinical consultants. In addition to cancer diagnosis, pathologists are responsible for multiple assessments based on glandular morphology statistics, like selecting... | ["Cancer diagnosis", "gland segmentation", "morphological feature guidance"] | 000001002003004005006007008009010011012013014015016017018019020021022023024025026027028029030031032033034035036037038039040041042043044000001002003004005006007008009010011012013014015016017018019020021022023024025026027028029030031032033034035036037038039040041042043044ECCV#6ECCV#6DoubleU-Net: Colorectal Cancer Diagnos... | 7KFxnpBYI7K | Solid work, great approach, good results | 9: Top 15% of accepted papers, strong accept | I think the presented approach of improving segmentation results of glands in histopathological images by additionally taking clues from textual descriptions is a great idea and the proposed method is indeed improving compared to a baseline not using the additional text.
The proposed method does, in fact, reach state-o... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
DoubleU-Net: Colorectal Cancer Diagnosis and Gland Instance Segmentation with Text-Guided Feature Control
### Paper Abstract
With the rapid therapeutic advancement in pe... |
Hk__kGbCW | ICLR.cc/2018/Conference | 2018 | DENSELY CONNECTED RECURRENT NEURAL NETWORK FOR SEQUENCE-TO-SEQUENCE LEARNING | ["Fei Tian"] | Deep neural networks based sequence-to-sequence learning has achieved remarkable progress in applications like machine translation and text summarization. However, sequence-to-sequence models suffer from severe inefficiency in training process, requiring huge amount of training time as well as memory usage. In this wor... | ["learning", "training time", "mechanism", "recurrent neural network", "memory usage", "remarkable progress", "applications", "machine translation", "text summarization"] | ABSTRACTDeep neural networks based sequence-to-sequence learning has achieved remark-able progress in applications like machine translation and text summarization.However, sequence-to-sequence models suffer from severe inefficiency in train-ing process, requiring huge amount of training time as well as memory usage. In... | Bkr5fKdlf | Dense skip-connections between layers improve recurrent networks | 5: Marginally below acceptance threshold | The article proposes to use dense skip-connections on the "vertical" (between-layers) connections of recurrent networks. Moreover, the article proposes to use separate attention-heads that run on the outputs of each encoder's layer, with each attention selecting other regions in the input to attend to.
The experiments... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
DENSELY CONNECTED RECURRENT NEURAL NETWORK FOR SEQUENCE-TO-SEQUENCE LEARNING
### Paper Abstract
Deep neural networks based sequence-to-sequence learning has achieved rem... |
HklbTjRcKX | ICLR.cc/2019/Conference | 2019 | What Information Does a ResNet Compress? | ["Luke Nicholas Darlow", "Amos Storkey"] | The information bottleneck principle (Shwartz-Ziv & Tishby, 2017) suggests that SGD-based training of deep neural networks results in optimally compressed hidden layers, from an information theoretic perspective. However, this claim was established on toy data. The goal of the work we present here is to test these clai... | ["Deep Learning", "Information Bottleneck", "Residual Neural Networks", "Information Theory"] | ABSTRACTThe information bottleneck principle (Shwartz-Ziv & Tishby, 2017) suggests thatSGD-based training of deep neural networks results in optimally compressed hid-den layers, from an information theoretic perspective. However, this claim wasestablished on toy data. The goal of the work we present here is to test whe... | HJlXGqI53Q | Interesting empirical work on compression in Resnets with partially inconclusive results | 6: Marginally above acceptance threshold | -> Summary
The authors propose to extend the analysis of Shwartz-Ziv & Tishby on the information bottleneck principle in artificial neural network training to realistic large-scale settings. They do so by replacing otherwise intractable quantities with tractable bounds in forms of classifiers for I(y;h) and Pixel CNNs... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
What Information Does a ResNet Compress?
### Paper Abstract
The information bottleneck principle (Shwartz-Ziv & Tishby, 2017) suggests that SGD-based training of deep ne... |
BJpv46DGNl_ | NeurIPS.cc/2021/Workshop/SVRHM | 2021 | Category-orthogonal object features guide information processing in recurrent neural networks trained for object categorization | ["Sushrut Thorat", "Giacomo Aldegheri", "Tim C Kietzmann"] | Recurrent neural networks (RNNs) have been shown to perform better than feedforward architectures in visual object categorization tasks, especially in challenging conditions such as cluttered images. However, little is known about the exact computational role of recurrent information flow in these conditions. Here we t... | ["inductive bias", "recurrent processing", "explainability", "emergent processes"] | Category-orthogonal object features guideinformation processing in recurrent neural networkstrained for object categorizationSushrut Thoratsushrut.thorat94@gmail.comGiacomo Aldegherigiacomo.aldegheri@gmail.comTim C. Kietzmanntim.kietzmann@donders.ru.nlDonders Institute for Brain, Cognition and Behaviour6525 AJ Nijmegen... | A-b5vu6j4Zr | An interesting study on context-dependent processing in an RNN network; however, the relevance for biological and machine learning systems is not made clear. | 6: Marginally above acceptance threshold | The submission studies how category-orthogonal information guides object recognition in a small recurrent network (RNN). The authors hypothesize that category-orthogonal information (e.g., location, size) is not reduced during processing, but is instead conditioned on in order to process category-relevant information (... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Category-orthogonal object features guide information processing in recurrent neural networks trained for object categorization
### Paper Abstract
Recurrent neural netwo... |
XtPeiGx6BwC | ICLR.cc/2021/Conference | 2021 | Robustness against Relational Adversary | ["Yizhen Wang", "Xiaozhu Meng", "Mihai Christodorescu", "Somesh Jha", "Ke Wang"] | Test-time adversarial attacks have posed serious challenges to the robustness of machine-learning models, and in many settings the adversarial perturbation need not be bounded by small lp-norms. Motivated by the semantics-preserving attacks in vision and security domain, we investigate relational adversaries, a broad ... | ["Adversarial Machine Learning", "Semantics-Preserving Attacks", "Logic Relations", "Normalization"] | ABSTRACTTest-time adversarial attacks have posed serious challenges to the robustnessof machine-learning models, and in many settings the adversarial perturbationneed not be bounded by small �p-norms. Motivated by the semantics-preservingattacks in vision and security domain, we investigaterelational adversaries, a bro... | dloPldWXBnQ | Interesting formulation to defend against adversarial attacks | 6: Marginally above acceptance threshold | The paper proposes to define robustness against adversarial attacks using a binary relation over inputs. The binary relation R states which input examples should be classified the same way and each connected component defined by R needs to be assigned the same label leading to robust classification. The idea of using a... | 2: The reviewer is willing to defend the evaluation, but it is quite likely that the reviewer did not understand central parts of the paper | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Robustness against Relational Adversary
### Paper Abstract
Test-time adversarial attacks have posed serious challenges to the robustness of machine-learning models, and ... |
B1lG42C9Km | ICLR.cc/2019/Conference | 2019 | Intrinsic Social Motivation via Causal Influence in Multi-Agent RL | ["Natasha Jaques", "Angeliki Lazaridou", "Edward Hughes", "Caglar Gulcehre", "Pedro A. Ortega", "DJ Strouse", "Joel Z. Leibo", "Nando de Freitas"] | We derive a new intrinsic social motivation for multi-agent reinforcement learning (MARL), in which agents are rewarded for having causal influence over another agent's actions, where causal influence is assessed using counterfactual reasoning. The reward does not depend on observing another agent's reward function, an... | ["multi-agent reinforcement learning", "causal inference", "game theory", "social dilemma", "intrinsic motivation", "counterfactual reasoning", "empowerment", "communication"] | ABSTRACTWe derive a new intrinsic social motivation for multi-agent reinforcement learning(MARL), in which agents are rewarded for having causal influence over another agent’sactions. Causal influence is assessed using counterfactual reasoning. The reward doesnot depend on observing another agent’s reward function, and... | Skg0DcDc2X | Review of Paper "Intrinsic Social Motivation via Causal Influence in Multi-Agent RL" | 6: Marginally above acceptance threshold | This paper proposes an approach to model social influence in a scenario-independent manner by instantiating the concept of intrinsic motivation and combine various human abilities as part of a reinforcement learning function in order to improve the agent's operation in social dilemma scenarios.
Agents are operational... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Intrinsic Social Motivation via Causal Influence in Multi-Agent RL
### Paper Abstract
We derive a new intrinsic social motivation for multi-agent reinforcement learning ... |
BJJ9bz-0- | ICLR.cc/2018/Conference | 2018 | Reinforcement Learning from Imperfect Demonstrations | ["Yang Gao", "Huazhe(Harry) Xu", "Ji Lin", "Fisher Yu", "Sergey Levine", "Trevor Darrell"] | Robust real-world learning should benefit from both demonstrations and interaction with the environment. Current approaches to learning from demonstration and reward perform supervised learning on expert demonstration data and use reinforcement learning to further improve performance based on reward from the environm... | ["learning from demonstration", "reinforcement learning", "maximum entropy learning"] | ABSTRACTRobust real-world learning should benefit from both demonstrations and inter-action with the environment. Current approaches to learning from demonstra-tion and reward perform supervised learning on expert demonstration data anduse reinforcement learning to further improve performance based on reward fromthe en... | ryE3gOulz | Similar to previous work - Experimental result not so convincing | 6: Marginally above acceptance threshold | This paper proposes a method to learn a control policy from both interactions with an environment and demonstrations. The method is inspired by the recent work on max entropy reinforcement learning and links between Q-learning and policy gradient methods. Especially the work builds upon the recent work by Haarnoja et a... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Reinforcement Learning from Imperfect Demonstrations
### Paper Abstract
Robust real-world learning should benefit from both demonstrations and interaction with the envi... |
HJgLLyrYwB | ICLR.cc/2020/Conference | 2020 | State-only Imitation with Transition Dynamics Mismatch | ["Tanmay Gangwani", "Jian Peng"] | Imitation Learning (IL) is a popular paradigm for training agents to achieve complicated goals by leveraging expert behavior, rather than dealing with the hardships of designing a correct reward function. With the environment modeled as a Markov Decision Process (MDP), most of the existing IL algorithms are contingent ... | ["Imitation learning", "Reinforcement Learning", "Inverse Reinforcement Learning"] | ABSTRACTImitation Learning (IL) is a popular paradigm for training agents to achieve com-plicated goals by leveraging expert behavior, rather than dealing with the hard-ships of designing a correct reward function. With the environment modeled asa Markov Decision Process (MDP), most of the existing IL algorithms are co... | SJeATxLnYr | Official Blind Review #1 | 6: Weak Accept | Summary:
The manuscript considers the problem of imitation learning when the system dynamics of the agent are different from the dynamics of the expert. The paper proposes Indirect Imitation Learning (I2L), which aims to perform imitation learning with respect to a trajectory buffer that contains some of the previous t... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
State-only Imitation with Transition Dynamics Mismatch
### Paper Abstract
Imitation Learning (IL) is a popular paradigm for training agents to achieve complicated goals ... | |
HJedho0qFX | ICLR.cc/2019/Conference | 2019 | Using Word Embeddings to Explore the Learned Representations of Convolutional Neural Networks | ["Dhanush Dharmaretnam", "Chris Foster", "Alona Fyshe"] | As deep neural net architectures minimize loss, they build up information in a hierarchy of learned representations that ultimately serve their final goal. Different architectures tackle this problem in slightly different ways, but all models aim to create representational spaces that accumulate information through t... | ["Distributional Semantics", "word embeddings", "cnns", "interpretability"] | ABSTRACTAs deep neural net architectures minimize loss, they build up information in ahierarchy of learned representations that ultimately serve their final goal. Differentarchitectures tackle this problem in slightly different ways, but all models aimto create representational spaces that accumulate information throug... | H1g9GJoN37 | Limited contribution. In addition, it's hard to identify the contribution of this paper. | 4: Ok but not good enough - rejection | This paper extends the previous work (Dharmaretnam & Fyshe, 2018), which provided a analytic tool for understanding CNNs through word embeddings of class labels. By analyzing correlations between each CNN layers and class labels, enables to investigates how each layer of CNNs work, how much it performed well, or how to... | 2: The reviewer is willing to defend the evaluation, but it is quite likely that the reviewer did not understand central parts of the paper | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Using Word Embeddings to Explore the Learned Representations of Convolutional Neural Networks
### Paper Abstract
As deep neural net architectures minimize loss, they bui... |
HJ-PorsiG | MIDL.amsterdam/2018/Conference | 2018 | Fully Automated Segmentation of the Left Ventricle in Small Animal Cardiac MRI | ["Hao Xu", "Jurgen E. Schneider", "Alistair A. Young", "Vicente Grau"] | The study of cardiovascular diseases requires viable animal models, and small animals have become models of choice due to their significant advantages. Segmentation from cine-MR image has become the gold standard for cardiac function assessment. While many image analysis methods have been developed for clinical studies... | ["segmentation", "cardiac MR", "cine MR", "deep learning", "U-Net", "small data"] | Fully Automated Segmentation of the Left Ventriclein Small Animal Cardiac MRIHao XuDepartment of Engineering ScienceUniversity of Oxfordhao.xu@eng.ox.ac.ukJurgen E. SchneiderInstitute of Cardiovascular and Metabolic MedicineUniversity of Leedsj.e.schneider@leeds.ac.ukAlistair A. YoungDepartment of Anatomy with Radiolog... | BkzN16gCf | An efficient 3D segmentation method is proposed for small animal cardiac MR images. A combination of long-axis and short-axis images is used, by combining convolutional neural networks and a guide-point modelling method. Due to small dataset used thus far and the presented results, we view the paper as preliminary. | 2: Marginally below acceptance threshold | In this paper, an efficient 3D segmentation method is proposed for small animal cardiac MR images. A combination of long-axis and short-axis images is used, by combining convolutional neural networks
and a guide-point modelling method. A small training set is shown to give accurate results. The segmentation method is s... | 2: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Fully Automated Segmentation of the Left Ventricle in Small Animal Cardiac MRI
### Paper Abstract
The study of cardiovascular diseases requires viable animal models, and... |
9WlOIHve8dU | ICLR.cc/2021/Conference | 2021 | Learning Binary Trees via Sparse Relaxation | ["Valentina Zantedeschi", "Matt Kusner", "Vlad Niculae"] | One of the most classical problems in machine learning is how to learn binary trees that split data into meaningful partitions. From classification/regression via decision trees to hierarchical clustering, binary trees are useful because they (a) are often easy to visualize; (b) make computationally-efficient predictio... | ["optimization", "binary trees"] | ABSTRACTOne of the most classical problems in machine learning is how to learn binary treesthat split data into useful partitions. From classification/regression via decision treesto hierarchical clustering, binary trees are useful because they (a) are often easy tovisualize; (b) make computationally-efficient predicti... | YiG1x_10a8q | Recommending Weak Accept. | 6: Marginally above acceptance threshold | ## Summary
This paper presents a new approach to learn decision trees via sparse relaxation. The approach starts from a mixed-integer program that can simultaneously induce and prune optimal decision trees from data. The proposed technique aims to solve a continuous relaxation of this problem by combining (1) a tree i... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Learning Binary Trees via Sparse Relaxation
### Paper Abstract
One of the most classical problems in machine learning is how to learn binary trees that split data into m... |
S1X7nhsxl | ICLR.cc/2017/conference | 2017 | Improving Generative Adversarial Networks with Denoising Feature Matching | ["David Warde-Farley", "Yoshua Bengio"] | We propose an augmented training procedure for generative adversarial networks designed to address shortcomings of the original by directing the generator towards probable configurations of abstract discriminator features. We estimate and track the distribution of these features, as computed from data, with a denoising... | ["Deep learning", "Unsupervised Learning"] | ABSTRACTWe propose an augmented training procedure for generative adversarial networksdesigned to address shortcomings of the original by directing the generator to-wards probable configurations of abstract discriminator features. We estimate andtrack the distribution of these features, as computed from data, with a de... | BJYbdoXVl | 7: Good paper, accept | This paper is well written, and well presented. This method is using denoise autoencoder to learn an implicit probability distribution helps reduce training difficulty, which is neat. In my view, joint training with an auto-encoder is providing extra auxiliary gradient information to improve generator. Providing auxili... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Improving Generative Adversarial Networks with Denoising Feature Matching
### Paper Abstract
We propose an augmented training procedure for generative adversarial networ... | |
Bys_NzbC- | ICLR.cc/2018/Conference | 2018 | Achieving Strong Regularization for Deep Neural Networks | ["Dae Hoon Park", "Chiu Man Ho", "Yi Chang"] | L1 and L2 regularizers are critical tools in machine learning due to their ability to simplify solutions. However, imposing strong L1 or L2 regularization with gradient descent method easily fails, and this limits the generalization ability of the underlying neural networks. To understand this phenomenon, we investigat... | ["deep learning", "regularization"] | ABSTRACTL1 and L2 regularizers are critical tools in machine learnin g due to their abilityto simplify solutions. However, imposing strong L1 or L2 reg ularization withgradient descent method easily fails, and this limits the ge neralization ability of theunderlying neural networks. To understand this phenomenon , we i... | H1U0Hpcef | An interesting diagnosis but a rudimentary cure | 5: Marginally below acceptance threshold | The work was prompted by an interesting observation: a phase transition can be observed in deep learning with stochastic gradient descent and Tikhonov regularization. When the regularization parameter exceeds a (data-dependent) threshold, the parameters of the model are driven to zero, thereby preventing any learning.... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Achieving Strong Regularization for Deep Neural Networks
### Paper Abstract
L1 and L2 regularizers are critical tools in machine learning due to their ability to simplif... |
hnJSgY7p33a | ICLR.cc/2021/Conference | 2021 | Graph-Based Neural Network Models with Multiple Self-Supervised Auxiliary Tasks | ["Franco Manessi", "Alessandro Rozza"] | Self-supervised learning is currently gaining a lot of attention, as it allows neural networks to learn robust representations from large quantities of unlabeled data. Additionally, multi-task learning can further improve representation learning by training networks simultaneously on related tasks, leading to significa... | ["Graph Neural Networks", "Self-supervised learning", "Multi-task learning", "Graph Convolutional Networks", "Semi-supervised learning"] | ABSTRACTSelf-supervised learning is currently gaining a lot of attention, as it allows neuralnetworks to learn robust representations from large quantities of unlabeled data.Additionally, multi-task learning can further improve representation learning bytraining networks simultaneously on related tasks, leading to sign... | ly_Xd7jzci8 | This paper proposes a general framework to improve graph-based neural network models by introducing self-supervised auxiliary learning tasks. Experiments can be further improved. | 5: Marginally below acceptance threshold | This paper proposes a general framework to improve graph-based neural network models by introducing self-supervised auxiliary learning tasks including auto encoding, corrupted feature reconstruction and corrupted embedding construction. The overall framework is composed of a shared encoder and four output heads, one fo... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Graph-Based Neural Network Models with Multiple Self-Supervised Auxiliary Tasks
### Paper Abstract
Self-supervised learning is currently gaining a lot of attention, as i... |
2NU7a9AHo-6 | ICLR.cc/2021/Conference | 2021 | AUL is a better optimization metric in PU learning | ["Shangchuan Huang", "Songtao Wang", "Dan Li", "Liwei Jiang"] | Traditional binary classification models are trained and evaluated with fully labeled data which is not common in real life. In non-ideal dataset, only a small fraction of positive data are labeled. Training a model from such partially labeled data is named as positive-unlabeled (PU) learning. A naive solution of PU le... | ["data", "aul", "better optimization metric", "model performance", "pu learning aul", "common", "real life", "dataset"] | ABSTRACTTraditional binary classification models are trained and evaluated with fully la-beled data which is not common in real life. In non-ideal dataset, only a smallfraction of positive data are labeled. Training a model from such partially labeleddata is named as positive-unlabeled (PU) learning. A naive solution o... | 9a3sNQhJchC | An interesting proposition but work is too premature | 3: Clear rejection | This paper examines the positive-unlabeled (PU) learning setting, and recommends the usage of the area under the lift curve, or AUL, as an unbiased estimate of the AUL under the fully labeled setting. This justifies proposing to use an AUL-optimization algorithm to train binary classifiers.
Although I am not entirely ... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
AUL is a better optimization metric in PU learning
### Paper Abstract
Traditional binary classification models are trained and evaluated with fully labeled data which i... |
rkeJzpNtPS | ICLR.cc/2020/Conference | 2020 | Locally adaptive activation functions with slope recovery term for deep and physics-informed neural networks | ["Ameya D. Jagtap", "Kenji Kawaguchi", "George Em Karniadakis"] | We propose two approaches of locally adaptive activation functions namely, layer-wise and neuron-wise locally adaptive activation functions, which improve the performance of deep and physics-informed neural networks. The local adaptation of activation function is achieved by introducing scalable hyper-parameters in eac... | ["PINN", "machine learning", "stochastic gradients", "accelerated training"] | ABSTRACTWe propose two approaches of locally adaptive activation functions namely, layer-wise and neuron-wise locally adaptive activation functions, which improve theperformance of deep and physics-informed neural networks. The local adapta-tion of activation function is achieved by introducing scalable hyper-parameter... | rJgmHeyOqr | Official Blind Review #4 | 3: Weak Reject | Interesting paper! The paper is well organized, well written and easy to read.
The main idea proposed in the paper is to modify the activation function $\sigma(x)$ in a neural network to $\sigma(ax)$ where $a$ is a learnable parameter. Every neuron could have it's own $a$, or $a$ could be shared across all neurons in a... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Locally adaptive activation functions with slope recovery term for deep and physics-informed neural networks
### Paper Abstract
We propose two approaches of locally adap... | |
diOr96f65N | robot-learning.org/CoRL/2023/Conference | 2023 | Quantifying Assistive Robustness Via the Natural-Adversarial Frontier | ["Jerry Zhi-Yang He", "Daniel S. Brown", "Zackory Erickson", "Anca Dragan"] | Our ultimate goal is to build robust policies for robots that assist people. What makes this hard is that people can behave unexpectedly at test time, potentially interacting with the robot outside its training distribution and leading to failures. Even just measuring robustness is a challenge. Adversarial perturbation... | ["assistive robots", "safety", "human-robot interaction", "adversarial robustness"] | Quantifying Assistive Robustness Via theNatural-Adversarial FrontierJerry Zhi-Yang He1, Zackory Erickson2, Daniel S. Brown3, and Anca D. Dragan11University of California Berkeley,2Carnegie Mellon University,3University of Utah,{hzyjerry,anca}@berkeley.edu ,dsbrown@cs.utah.edu ,zerickso@cmu.eduAbstract: Our ultimate goa... | HlnR0AxJhIq | # Comments
In this paper, the authors employ an adversarial learning approach to learning human policy for evaluating the control policies of assistive robots. The objective of this approach is to obstruct the robot's policies with natural movements. In this point, the proposed method is related to robust adversarial ... | 2: The reviewer is willing to defend the evaluation, but it is quite likely that the reviewer did not understand central parts of the paper | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Quantifying Assistive Robustness Via the Natural-Adversarial Frontier
### Paper Abstract
Our ultimate goal is to build robust policies for robots that assist people. Wha... | ||
Syl38yrFwr | ICLR.cc/2020/Conference | 2020 | Near-Zero-Cost Differentially Private Deep Learning with Teacher Ensembles | ["Lichao Sun", "Yingbo Zhou", "Jia Li", "Richard Socher", "Philip S. Yu", "Caiming Xiong"] | Ensuring the privacy of sensitive data used to train modern machine learning models is of paramount importance in many areas of practice. One approach to study these concerns is through the lens of differential privacy. In this framework, privacy guarantees are generally obtained by perturbing models in such a way that... | ["sensitive data", "teacher", "student", "private deep learning", "teacher ensembles", "model", "information", "privacy", "paramount importance"] | ABSTRACTEnsuring the privacy of sensitive data used to train modern machine learning mod-els is of paramount importance in many areas of practice. One approach to studythese concerns is through the lens of differential privacy. In this framework, pri-vacy guarantees are generally obtained by perturbing models in such a... | SyeKgE2JqS | Official Blind Review #1 | 1: Reject | To improve the privacy-utility tradeoff, this manuscript proposes a voting mechanism used in a teacher-student model, where there is an ensemble of teachers, from which the student can get gradient information for utility improvement. The main idea of the proposed approach is to add a constant C to the maximum count co... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Near-Zero-Cost Differentially Private Deep Learning with Teacher Ensembles
### Paper Abstract
Ensuring the privacy of sensitive data used to train modern machine learnin... | |
H1gmHaEKwB | ICLR.cc/2020/Conference | 2020 | Data-Independent Neural Pruning via Coresets | ["Ben Mussay", "Margarita Osadchy", "Vladimir Braverman", "Samson Zhou", "Dan Feldman"] | Previous work showed empirically that large neural networks can be significantly reduced in size while preserving their accuracy. Model compression became a central research topic, as it is crucial for deployment of neural networks on devices with limited computational and memory resources. The majority of the compress... | ["coresets", "neural pruning", "network compression"] | ABSTRACTPrevious work showed empirically that large neural networks can be significantlyreduced in size while preserving their accuracy. Model compression became acentral research topic, as it is crucial for deployment of neural networks on de-vices with limited computational and memory resources. The majority of theco... | SkgB97XXqB | Official Blind Review #3 | 8: Accept | ------ after reading authors' response ------
Thanks for the complete response and revision. The new Fig 4 is very nice, and this helps address my concerns. I'm more favorable of the paper now, changing from "Weak Accept" to "Accept".
Note a small typo in the displayed equation just above Fig 4: the sum is over $P_{te... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Data-Independent Neural Pruning via Coresets
### Paper Abstract
Previous work showed empirically that large neural networks can be significantly reduced in size while pr... | |
CYHMIhbuLFl | ICLR.cc/2021/Conference | 2021 | Contextual HyperNetworks for Novel Feature Adaptation | ["Angus Lamb", "Evgeny Saveliev", "Yingzhen Li", "Sebastian Tschiatschek", "Camilla Longden", "Simon Woodhead", "Jos\u00e9 Miguel Hern\u00e1ndez-Lobato", "Richard E Turner", "Pashmina Cameron", "Cheng Zhang"] | While deep learning has obtained state-of-the-art results in many applications, the adaptation of neural network architectures to incorporate new features remains a research challenge. This issue is particularly severe in online learning settings, where new features are added continually with few or no associated obser... | ["Meta learning", "few-shot learning", "continual learning", "recommender systems", "deep learning"] | ABSTRACTWhile deep learning has obtained state-of-the-art results in many applications,the adaptation of neural network architectures to incorporate new output featuresremains a challenge, as a neural networks are commonly trained to produce a fixedoutput dimension. This issue is particularly severe in online learning ... | sCpwKREHi_- | An okay submission but with limited novelty. | 5: Marginally below acceptance threshold | This submission focuses on the cold start problem of new entities (new items in a recommender system, new treatments in a medical application, etc.). It combines the strengths of the *relations* between a new entity and the existing entities, and the *content* features of the new entity, by fusing the two kinds of info... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Contextual HyperNetworks for Novel Feature Adaptation
### Paper Abstract
While deep learning has obtained state-of-the-art results in many applications, the adaptation ... |
SJg7spEYDS | ICLR.cc/2020/Conference | 2020 | Generative Ratio Matching Networks | ["Akash Srivastava", "Kai Xu", "Michael U. Gutmann", "Charles Sutton"] | Deep generative models can learn to generate realistic-looking images, but many of the most effective methods are adversarial and involve a saddlepoint optimization, which requires a careful balancing of training between a generator network and a critic network. Maximum mean discrepancy networks (MMD-nets) avoid this i... | ["deep generative model", "deep learning", "maximum mean discrepancy", "density ratio estimation"] | ABSTRACTDeep generative models can learn to generate realistic-looking images, but many ofthe most effective methods are adversarial and involve a saddlepoint optimization,which requires a careful balancing of training between a generator network anda critic network. Maximum mean discrepancy networks (MMD-nets) avoid t... | BkgbEu5dKS | Official Blind Review #3 | 6: Weak Accept | In this paper, authors propose a new generative adversarial networks based on density-ratio estimation. More specifically, the squared-loss between the true ratio function and the model (a.k.a., Pearson divergence) is employed to train model. Through experiments, authors demonstrate that the proposed density-ratio bas... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Generative Ratio Matching Networks
### Paper Abstract
Deep generative models can learn to generate realistic-looking images, but many of the most effective methods are a... | |
SJMO2iCct7 | ICLR.cc/2019/Conference | 2019 | A NOVEL VARIATIONAL FAMILY FOR HIDDEN NON-LINEAR MARKOV MODELS | ["Daniel Hernandez Diaz", "Antonio Khalil Moretti", "Ziqiang Wei", "Shreya Saxena", "John Cunningham", "Liam Paninski"] | Latent variable models have been widely applied for the analysis and visualization of large datasets. In the case of sequential data, closed-form inference is possible when the transition and observation functions are linear. However, approximate inference techniques are usually necessary when dealing with nonlinear ev... | ["variational inference", "time series", "nonlinear dynamics", "neuroscience"] | ABSTRACTLatent variable models have been widely applied for the analysis and visualizationof large datasets. In the case of sequential data, closed-form inference is possiblewhen the transition and observation functions are linear. However, approximateinference techniques are usually necessary when dealing with nonline... | rJgBQzFjnX | Nice algorithm but need better motivation | 5: Marginally below acceptance threshold | This paper discusses a algorithm for variational inference of a non-linear dynamical models. In this paper model assumption is to use single stage Markov model in latent space with every latent variable Z_t to be defined Gaussian distributed with mean depends on Z_(t-1) and time invariant variance matrix lambda. The no... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
A NOVEL VARIATIONAL FAMILY FOR HIDDEN NON-LINEAR MARKOV MODELS
### Paper Abstract
Latent variable models have been widely applied for the analysis and visualization of l... |
rygjcsR9Y7 | ICLR.cc/2019/Conference | 2019 | SOM-VAE: Interpretable Discrete Representation Learning on Time Series | ["Vincent Fortuin", "Matthias H\u00fcser", "Francesco Locatello", "Heiko Strathmann", "Gunnar R\u00e4tsch"] | High-dimensional time series are common in many domains. Since human cognition is not optimized to work well in high-dimensional spaces, these areas could benefit from interpretable low-dimensional representations. However, most representation learning algorithms for time series data are difficult to interpret. This is... | ["deep learning", "self-organizing map", "variational autoencoder", "representation learning", "time series", "machine learning", "interpretability"] | ABSTRACTHigh-dimensional time series are common in many domains. Since human cognition isnot optimized to work well in high-dimensional spaces, these areas could benefit frominterpretable low-dimensional representations. However, most representation learningalgorithms for time series data are difficult to interpret. Th... | SJeYiFul3X | Interesting work, methodological aspects and implementation details to be clarified. Improving comparison with state of the art | 6: Marginally above acceptance threshold | This work addresses the problem of learning latent embeddings of high-dimensional time series data. The paper emphasises the need of interpretable representations accounting for the correlated nature of temporal data. To this scope, the study proposes to cluster the data in a latent space estimated through an auto-enco... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
SOM-VAE: Interpretable Discrete Representation Learning on Time Series
### Paper Abstract
High-dimensional time series are common in many domains. Since human cognition ... |
r1eCukHYDH | ICLR.cc/2020/Conference | 2020 | Manifold Learning and Alignment with Generative Adversarial Networks | ["Jiseob Kim", "Seungjae Jung", "Hyundo Lee", "Byoung-Tak Zhang"] | We present a generative adversarial network (GAN) that conducts manifold learning and alignment (MLA): A task to learn the multi-manifold structure underlying data and to align those manifolds without any correspondence information. Our main idea is to exploit the powerful abstraction ability of encoder architecture. S... | ["Generative Adversarial Networks", "Manifold Learning", "Manifold Alignment"] | ABSTRACTWe present a generative adversarial network (GAN) that conducts manifold learn-ing and alignment (MLA): A task to learn the multi-manifold structure underlyingdata and to align those manifolds without any correspondence information. Ourmain idea is to exploit the powerful abstraction ability of encoder architec... | BJlU_xM-5S | Official Blind Review #1 | 6: Weak Accept | EDIT: Updated score to weak Accept in lieu of author's response. See below for more details.
The authors propose a GAN architecture that aims to align the latent representations of the GAN with different interpretable degrees of freedom of the underlying data (e.g., size, pose). While the text, motivation, and exper... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Manifold Learning and Alignment with Generative Adversarial Networks
### Paper Abstract
We present a generative adversarial network (GAN) that conducts manifold learning... | |
H1lDbaVYvH | ICLR.cc/2020/Conference | 2020 | SMiRL: Surprise Minimizing RL in Entropic Environments | ["Glen Berseth", "Daniel Geng", "Coline Devin", "Dinesh Jayaraman", "Chelsea Finn", "Sergey Levine"] | All living organisms struggle against the forces of nature to carve out niches where
they can maintain relative stasis. We propose that such a search for order amidst
chaos might offer a unifying principle for the emergence of useful behaviors in
artificial agents. We formalize this idea into an unsupervised reinforcem... | ["intrinsic motivation", "reinforcement learning", "unsurpervised RL"] | ABSTRACTAll living organisms struggle against the forces of nature to carve out niches wherethey can maintain relative stasis. We propose that such a search for order amidstchaos might offer a unifying principle for the emergence of useful behaviors inartificial agents. We formalize this idea into an unsupervised reinf... | HJgOkLqitS | Official Blind Review #3 | 6: Weak Accept | This paper proposes Surprise Minimizing RL (SMiRL), a conceptual framework for training a reinforcement learning agent to seek out states with high likelihood under a density model trained on visited states. They qualitatively and quantitatively explore various aspects of the behaviour of these agents and argue that th... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
SMiRL: Surprise Minimizing RL in Entropic Environments
### Paper Abstract
All living organisms struggle against the forces of nature to carve out niches where they can m... | |
BJlvCtU45E | MIDL.io/2019/Conference/Abstract | 2019 | From Point Annotations to Epithelial Cell Detection in Breast Cancer Histopathology using RetinaNet | ["Caner Mercan", "Maschenka Balkenhol", "Jeroen van der Laak", "Francesco Ciompi"] | Detection of epithelial cells has powerful implications such as being an integral part of nuclear pleomorphism scoring for breast cancer grading. We exploit the point annotations inside nuclei boundaries to estimate their bounding boxes using empirical analysis on the cell bodies and the coarse instance segmentation ma... | ["Epithelial cell detection", "RetinaNet", "breast histopathology"] | Proceedings of Machine Learning Research { Under Review:1{4, 2019 Extended Abstract { MIDL 2019 submissionFrom Point Annotations to Epithelial Cell Detection inBreast Cancer Histopathology using RetinaNetCaner Mercan Caner.Mercan@radboudumc.nlMaschenka Balkenhol Maschenka.Balkenhol@radboudumc.nlJeroen van der Laak Jero... | Hkg4PP-wsE | Application-oriented abstract using RetinaNet to detect epithelial cells in breast cancer histopathology | 2: Reject | The RetinaNet with different optimizers (Adam vs. AdaBound) and bounding boxes (Grabcut vs. Fixed) was trained, validated, and tested in this work on a total of 39 whole-slide breast histopathology images of 39 patients. The best methods achieved mean average precision scores of ~71.4% and ~59.7% on tumor and benign ce... | 3: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
From Point Annotations to Epithelial Cell Detection in Breast Cancer Histopathology using RetinaNet
### Paper Abstract
Detection of epithelial cells has powerful implica... |
rMSlLb33Gb | KDD.org/2023/Workshop/epiDAMIK | 2023 | A Snapshot of COVID-19 Incidence, Hospitalizations, and Mortality from Indirect Survey Data in China in January 2023 (Extended Abstract) | ["Juan Marcos Ramirez", "Sergio Diaz-Aranda", "Jose Aguilar", "Oluwasegun Ojo", "Rosa Elvira Lillo", "Antonio Fernandez Anta"] | The estimation of incidence has been a crucial component for monitoring COVID-19 dissemination. This has become challenging when official data are unavailable or insufficiently reliable. Hence, the implementation of efficient, inexpensive, and secure techniques that capture information about epidemic indicators is requ... | ["COVID-19", "incidence estimation", "indirect surveys", "NSUM"] | ABSTRACTThe estimation of incidence has been a crucial component for moni-toring COVID-19 dissemination. This has become challenging whenofficial data are unavailable or insufficiently reliable. Hence, the im-plementation of efficient, inexpensive, and secure techniques thatcapture information about epidemic indicators... | i7lQlnCLyYO | Paper provides a succinct and accessible method to estimate disease incidence | 4: Good paper, accept | ### Summary
This paper seeks to improve disease incidence estimation methods using information from surveys about contacts, rather than the respondents direct experience. They can obtain much more information by asking about multiple individuals the respondent knows about rather than gather information about only one i... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
A Snapshot of COVID-19 Incidence, Hospitalizations, and Mortality from Indirect Survey Data in China in January 2023 (Extended Abstract)
### Paper Abstract
The estimatio... |
SNeLAhVuDZ9 | ICLR.cc/2022/Workshop/DGM4HSD | 2022 | Score-Based Generative Models for Wireless Channel Modeling and Estimation | ["Marius Arvinte", "Jonathan Tamir"] | In this work, we investigate score-based models for learning the distribution of multiple-input multiple-output (MIMO) wireless channels in structured stochastic environments, using either clean or corrupted (noisy) data for training. We find that score-based models are capable of generating high-quality synthetic chan... | ["score-based", "generative", "wireless", "estimation"] | ABSTRACTIn this work, we investigate score-based models for learning the distribution ofmultiple-input multiple-output (MIMO) wireless channels in structured stochasticenvironments, using either clean or corrupted (noisy) data for training. We findthat score-based models are capable of generating high-quality synthetic... | rflr0LCmM9 | Good work fitting the workshops's objectives but lacking some practical details | 7: Good paper, accept | The paper uses score-based models and GANs in multiple-input multiple-output (MIMO) channel estimation. Despite the work seems to be well formulated and evaluated, I have a limited knowledge on the specific application field of wireless channel modelling to be entirely sure about my review. Indeed, most of my comments ... | 2: The reviewer is willing to defend the evaluation, but it is quite likely that the reviewer did not understand central parts of the paper | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Score-Based Generative Models for Wireless Channel Modeling and Estimation
### Paper Abstract
In this work, we investigate score-based models for learning the distributi... |
SkgGjRVKDS | ICLR.cc/2020/Conference | 2020 | Towards Stabilizing Batch Statistics in Backward Propagation of Batch Normalization | ["Junjie Yan", "Ruosi Wan", "Xiangyu Zhang", "Wei Zhang", "Yichen Wei", "Jian Sun"] | Batch Normalization (BN) is one of the most widely used techniques in Deep Learning field. But its performance can awfully degrade with insufficient batch size. This weakness limits the usage of BN on many computer vision tasks like detection or segmentation, where batch size is usually small due to the constraint of m... | ["batch normalization", "small batch size", "backward propagation"] | ABSTRACTBatch Normalization (BN) is one of the most widely used techniques in DeepLearning field. But its performance can awfully degrade with insufficient batchsize. This weakness limits the usage of BN on many computer vision tasks like de-tection or segmentation, where batch size is usually small due to the constrai... | Bke_CWyaFH | Official Blind Review #1 | 8: Accept | The paper proposes a new approach for batch-normalization. Standard approaches are sensitive to the batch size, because small batches will lead to unstable statistics. So when the mini-batch is small, the performance can drop significantly. The paper addresses this issue by analyzing extra statistics in the batch norma... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Towards Stabilizing Batch Statistics in Backward Propagation of Batch Normalization
### Paper Abstract
Batch Normalization (BN) is one of the most widely used techniques... | |
_CrmWaJ2uvP | ICLR.cc/2021/Conference | 2021 | Recurrent Neural Network Architecture based on Dynamic Systems Theory for Data Driven Modelling of Complex Physical Systems | ["Deniz Neufeld"] | While dynamic systems can be modelled as sequence-to-sequence tasks by deep learning using different network architectures like DNN, CNN, RNNs or neural ODEs, the resulting models often provide poor understanding of the underlying system properties. We propose a new recurrent network architecture, the Dynamic Recurre... | ["dynamic system identification", "recurrent networks", "explainable AI", "time series modelling"] | ABSTRACTWhile dynamic systems can be modeled as sequence-to-sequence tasks by deeplearning using different network architectures like DNN, CNN, RNNs or neuralODEs, the resulting models often provide poor understanding of the underlyingsystem properties. We propose a new recurrent network architecture, the DynamicRecurr... | tGHf_ON5HBh | Where’s the interpretability exactly? | 4: Ok but not good enough - rejection | The theory of dynamical systems has such a rich and extensive history dating to more than a century ago, that Poincaré himself would marvel to see what we can do now with modern technology. Perhaps he’d not be pleased to see that extensive research in dynamical systems from the previous century is ignored in this pape... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Recurrent Neural Network Architecture based on Dynamic Systems Theory for Data Driven Modelling of Complex Physical Systems
### Paper Abstract
While dynamic systems can ... |
QcjTNc_afvH | ICLR.cc/2021/Conference | 2021 | A StyleMap-Based Generator for Real-Time Image Projection and Local Editing | ["Hyunsu Kim", "Yunjey Choi", "Junho Kim", "Sungjoo Yoo", "Youngjung Uh"] | Generative adversarial networks (GANs) have been successful in synthesizing and manipulating synthetic but realistic images from latent vectors. However, it is still challenging for GANs to manipulate real images, especially in real-time. State-of-the-art GAN-based methods for editing real images suffer from time-consu... | ["Generative Adversarial Network", "Real-time Image Projection", "Image Manipulation", "Local Editing", "Deep Learning"] | ABSTRACTGenerative adversarial networks (GANs) have been successful in synthesizing andmanipulating synthetic but realistic images from latent vectors. However, it is stillchallenging for GANs to manipulate real images, especially in real-time. State-of-the-art GAN-based methods for editing real images suffer from time... | _YlkTM5heoX | Misleading terminology and unclear experiments | 3: Clear rejection | Summary: This paper proposes a new method for image generation and editing. A smaller network generates multi-scale feature maps, which are used to modulate (affine transform) the features of a larger network in corresponding scales. Besides, an encoder is trained to project real images into the lowest-resolution featu... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
A StyleMap-Based Generator for Real-Time Image Projection and Local Editing
### Paper Abstract
Generative adversarial networks (GANs) have been successful in synthesizin... |
HJtN5K9gx | ICLR.cc/2017/conference | 2017 | Learning Disentangled Representations in Deep Generative Models | ["N. Siddharth", "Brooks Paige", "Alban Desmaison", "Jan-Willem van de Meent", "Frank Wood", "Noah D. Goodman", "Pushmeet Kohli", "Philip H.S. Torr"] | Deep generative models provide a powerful and flexible means to learn complex distributions over data by incorporating neural networks into latent-variable models. Variational approaches to training such models introduce a probabilistic encoder that casts data, typically unsupervised, into an entangled and unstructured... | ["Semi-Supervised Learning", "Deep learning", "Computer vision"] | ABSTRACTDeep generative models provide a powerful and flexible means to learn com-plex distributions over data by incorporating neural networks into latent-variablemodels. Variational approaches to training such models introduce a probabilisticencoder that casts data, typically unsupervised, into an entangled represent... | B1FyYJmVe | A variant of the semi-supervised VAE. | 6: Marginally above acceptance threshold | This paper proposed a variant of the semi-supervised VAE model which leads to a unified objective for supervised and unsupervised VAE. This variant gives software implementation of these VAE models more flexibility in specifying which variables are supervised and which are not.
This development introduces a few extra... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Learning Disentangled Representations in Deep Generative Models
### Paper Abstract
Deep generative models provide a powerful and flexible means to learn complex distribu... |
kB8DkEKSDH | ICLR.cc/2021/Conference | 2021 | Hellinger Distance Constrained Regression | ["Egor Rotinov"] | This paper introduces an off-policy reinforcement learning method that uses Hellinger distance between sampling policy (from what samples were collected) and current policy (policy being optimized) as a constraint.
Hellinger distance squared multiplied by two is greater than or equal to total variation distance square... | ["offline", "Reinforcement Learning", "off-policy", "control"] | ABSTRACTThis paper introduces an off-policy reinforcement learning method that usesHellinger distance between sampling policy (from what samples were collected)and current policy (policy being optimized) as a constraint. Hellinger distancesquared multiplied by two is greater than or equal to total variation distancesqu... | VBIr6lYuNZO | Insufficient contribution and experimental validation | 3: Clear rejection | The authors propose the use of the Hellinger distance instead of KL divergence to constrain new policies to remain close to the behavior policy. The technical aspects are straightforward, noting that Hellinger provides tighter bounds on total variation than KL, and can straightforwardly be plugged into the CPI/TRPO bou... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Hellinger Distance Constrained Regression
### Paper Abstract
This paper introduces an off-policy reinforcement learning method that uses Hellinger distance between sampl... |
EdXhmWvvQV | ICLR.cc/2021/Conference | 2021 | Center-wise Local Image Mixture For Contrastive Representation Learning | ["Hao Li", "XIAOPENG ZHANG", "Ruoyu Sun", "Hongkai Xiong", "Qi Tian"] | Recent advances in unsupervised representation learning have experienced remarkable progress, especially with the achievements of contrastive learning, which regards each image as well its augmentations as a separate class, while does not consider the semantic similarity among images. This paper proposes a new kind of ... | ["Self-supervised Learning", "Data Mixing", "Contrastive Learning"] | ABSTRACTRecent advances in unsupervised representation learning have experienced re-markable progress, especially with the achievements of contrastive learning,which regards each image as well its augmentations as a separate class, whiledoes not consider the semantic similarity among images. This paper proposes anew ki... | PmHb63JS32u | Simple method, very strong results, but not very novel and some justifications are missing. | 6: Marginally above acceptance threshold | Center-wise Local Image Mixture For Contrastive Representation Learning
The paper introduces a new contrastive learning method for unsupervised representation learning. The main idea is to consider the semantic similarity between different images and incorporate it in the learning procedure, in contrast to the many co... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Center-wise Local Image Mixture For Contrastive Representation Learning
### Paper Abstract
Recent advances in unsupervised representation learning have experienced remar... |
ryggIs0cYQ | ICLR.cc/2019/Conference | 2019 | Differentiable Learning-to-Normalize via Switchable Normalization | ["Ping Luo", "Jiamin Ren", "Zhanglin Peng", "Ruimao Zhang", "Jingyu Li"] | We address a learning-to-normalize problem by proposing Switchable Normalization (SN), which learns to select different normalizers for different normalization layers of a deep neural network. SN employs three distinct scopes to compute statistics (means and variances) including a channel, a layer, and a minibatch. SN ... | ["normalization", "deep learning", "CNN", "computer vision"] | ABSTRACTWe address a learning-to-normalize problem by proposing Switchable Normaliza-tion (SN), which learns to select different normalizers for different normalizationlayers of a deep neural network. SN employs three distinct scopes to computestatistics (means and variances) including a channel, a layer, and a minibat... | BygLp1wrpm | Simple approach that works | 7: Good paper, accept | This paper considers normalization by learning to average across well-known normalization techniques.
This meta-procedure almost by definition can yield better results, and this is demonstrated nicely by the authors.
The idea is simple and easy to implement, and easy works in this case.
I would like to hear more about... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Differentiable Learning-to-Normalize via Switchable Normalization
### Paper Abstract
We address a learning-to-normalize problem by proposing Switchable Normalization (SN... |
dKwmCtp6YI | ICLR.cc/2021/Conference | 2021 | Representation and Bias in Multilingual NLP: Insights from Controlled Experiments on Conditional Language Modeling | ["Ada Wan"] | Inspired by the phenomenon of performance disparity between languages in machine translation, we investigate whether and to what extent languages are equally hard to "conditional-language-model". Our goal is to improve our understanding and expectation of the relationship between language, data representation, size, an... | ["multilinguality", "science for NLP", "fundamental science in the era of AI/DL", "representation learning for language", "conditional language modeling", "Transformer", "Double Descent", "non-monotonicity", "fairness", "meta evaluation", "visualization or interpretation of learned representations"] | ABSTRACTInspired by the phenomenon of performance disparity between languages in ma-chine translation, we investigate whether and to what extent languages are equallyhard to “conditional-language-model”. Our goal is to improve our understandingand expectation of the relationship between language, data representation, s... | ZTDrA_qwKLm | An interesting and thorough exaimination of an important problem. Writing is sometimes too complicated and some more high level analysis could help. | 6: Marginally above acceptance threshold | The paper provides an empirical investigation of an important problem: the transferability of language modeling signals across languages in the transformer model. This is an important question because it can teach us both on the relations between languages and the properties of the transformer model (although it is not... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Representation and Bias in Multilingual NLP: Insights from Controlled Experiments on Conditional Language Modeling
### Paper Abstract
Inspired by the phenomenon of perfo... |
SylwBpNKDr | ICLR.cc/2020/Conference | 2020 | Boosting Network: Learn by Growing Filters and Layers via SplitLBI | ["Zuyuan Zhong", "Chen Liu", "Yanwei Fu", "Yuan Yao"] | Network structures are important to learning good representations of many tasks in computer vision and machine learning communities. These structures are either manually designed, or searched by Neural Architecture Search (NAS) in previous works, which however requires either expert-level efforts, or prohibitive comput... | ["network", "layers", "learn", "filters", "splitlbi", "structures", "splitlbi network structures", "important", "good representations", "many tasks"] | ABSTRACTNetwork structures are important to learning good representations of many tasksin computer vision and machine learning communities. These structures are eithermanually designed, or searched by Neural Architecture Search (NAS) in previousworks, which however requires either expert-level efforts, or prohibitive c... | HygAtPPZ5S | Official Blind Review #2 | 6: Weak Accept | This paper proposes an architecture search method for deep convolutional neural network models that progressively increases the number of filters per layer as well as the number of layers, and the authors refer to this general approach as boosting networks. The algorithm for increasing the number of filters is based on... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Boosting Network: Learn by Growing Filters and Layers via SplitLBI
### Paper Abstract
Network structures are important to learning good representations of many tasks in ... | |
cAvgPMAA3hb | ICLR.cc/2021/Conference | 2021 | GRF: Learning a General Radiance Field for 3D Scene Representation and Rendering | ["Alex Trevithick", "Bo Yang"] | We present a simple yet powerful implicit neural function that can represent and render arbitrarily complex 3D scenes in a single network only from 2D observations. The function models 3D scenes as a general radiance field, which takes a set of 2D images with camera poses and intrinsics as input, constructs an internal... | ["3D scene representation", "novel view synthesis", "neural rendering"] | ABSTRACTWe present a simple yet powerful implicit neural function that can represent andrender arbitrarily complex 3D scenes in a single network only from 2D observa-tions. The function models 3D scenes as a general radiance field, which takes aset of 2D images with cameraposes andintrinsics as input, constructs an int... | Qk0GBDPv5K | Recommendation to weakly Reject | 4: Ok but not good enough - rejection | Summary:
The paper proposed an extension on Neural radiance field for better generalization across novel scenes by introducing multi-view pixel aligned features as additional input to NeRF. To fuse multi-view information and implicit reason about occlusion, an attention aggregation module is applied.
Strengths:
+ Th... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
GRF: Learning a General Radiance Field for 3D Scene Representation and Rendering
### Paper Abstract
We present a simple yet powerful implicit neural function that can re... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.