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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
r1xapAEKwS | ICLR.cc/2020/Conference | 2020 | SDGM: Sparse Bayesian Classifier Based on a Discriminative Gaussian Mixture Model | ["Hideaki Hayashi", "Seiichi Uchida"] | In probabilistic classification, a discriminative model based on Gaussian mixture exhibits flexible fitting capability. Nevertheless, it is difficult to determine the number of components. We propose a sparse classifier based on a discriminative Gaussian mixture model (GMM), which is named sparse discriminative Gaussia... | ["classification", "sparse Bayesian learning", "Gaussian mixture model"] | ABSTRACTIn probabilistic classification, a discriminative model based on Gaussian mixtureexhibits flexible fitting capability. Nevertheless, it is difficult to determine thenumber of components. We propose a sparse classifier based on a discriminativeGaussian mixture model (GMM), which is named sparse discriminative Ga... | SygQvluBFS | Official Blind Review #3 | 3: Weak Reject | This paper proposes a classifier, called SDGM, based on discriminative Gaussian mixture and its sparse parameter estimation.
The method aims to have a flexible decision boundary due to the mixture model as well as to generalize well owing to the sparse learning strategy.
While the method incorporates distinct ideas i... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
SDGM: Sparse Bayesian Classifier Based on a Discriminative Gaussian Mixture Model
### Paper Abstract
In probabilistic classification, a discriminative model based on Gau... | |
SJeYe0NtvH | ICLR.cc/2020/Conference | 2020 | Neural Text Generation With Unlikelihood Training | ["Sean Welleck", "Ilia Kulikov", "Stephen Roller", "Emily Dinan", "Kyunghyun Cho", "Jason Weston"] | Neural text generation is a key tool in natural language applications, but it is well known there are major problems at its core. In particular, standard likelihood training and decoding leads to dull and repetitive outputs. While some post-hoc fixes have been proposed, in particular top-k and nucleus sampling, they do... | ["language modeling", "machine learning"] | ABSTRACTNeural text generation is a key tool in natural language applications, but it is wellknown there are major problems at its core. In particular, standard likelihoodtraining and decoding leads to dull and repetitive outputs (Holtzman et al., 2019).While some post-hoc fixes have been proposed, in particular top- k... | ryeEqCDG5B | Official Blind Review #1 | 3: Weak Reject | This paper targets on solving the dull and repetitive outputs in MLE training for neural text generation. The authors propose a new unlikelyhood training to avoid assigning much probability to frequent and repetitive words. The authors combine the proposed algorithms and beam search and state that the results improved ... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Neural Text Generation With Unlikelihood Training
### Paper Abstract
Neural text generation is a key tool in natural language applications, but it is well known there ar... | |
HyoST_9xl | ICLR.cc/2017/conference | 2017 | DSD: Dense-Sparse-Dense Training for Deep Neural Networks | ["Song Han", "Jeff Pool", "Sharan Narang", "Huizi Mao", "Enhao Gong", "Shijian Tang", "Erich Elsen", "Peter Vajda", "Manohar Paluri", "John Tran", "Bryan Catanzaro", "William J. Dally"] | Modern deep neural networks have a large number of parameters, making them very hard to train. We propose DSD, a dense-sparse-dense training flow, for regularizing deep neural networks and achieving better optimization performance. In the first D (Dense) step, we train a dense network to learn connection weights and im... | ["Deep learning"] | ABSTRACTModern deep neural networks have a large number of parameters, making themvery hard to train. We propose DSD, a dense-sparse-dense training flow, forregularizing deep neural networks and achieving better optimization performance.In the first D (Dense) step, we train a dense network to learn connection weightsan... | HydNSx9Xg | Interesting training strategy for deep networks | 5: Marginally below acceptance threshold | This paper presents a training strategy for deep networks. First, the network is trained in a standard fashion. Second, small magnitude weights are clamped to 0; the rest of the weights continue to be trained. Finally, all the weights are again jointly trained. Experiments on a variety of image, text, and speech da... | 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
DSD: Dense-Sparse-Dense Training for Deep Neural Networks
### Paper Abstract
Modern deep neural networks have a large number of parameters, making them very hard to trai... |
dcktlmtcM7 | ICLR.cc/2021/Conference | 2021 | Neural Time-Dependent Partial Differential Equation | ["Yihao Hu", "Tong Zhao", "Zhiliang Xu", "Lizhen Lin"] | Partial differential equations (PDEs) play a crucial role in studying a vast number of problems in science and engineering. Numerically solving nonlinear and/or high-dimensional PDEs is frequently a challenging task. Inspired by the traditional finite difference and finite elements methods and emerging advancements in ... | ["Numerical analysis", "Deep learning", "Partial differential equation", "Machine learning", "Predictive modeling"] | ABSTRACTPartial differential equations (PDEs) play a crucial role in studying a vast numberof problems in science and engineering. Numerically solving nonlinear and/or high-dimensional PDEs is frequently a challenging task. Inspired by the traditional finitedifference and finite elements methods and emerging advancemen... | WqoR7svzhhO | Need the comparison with Conv-LSTM | 3: Clear rejection | Summary:
The paper proposes to use LSTM to learn partial differential equations. Experiments include Wave equation, Heat equation, Burgers' equation, and Navier-Stokes equation. They compare with PINN on Allen-Cahn and Burgers equation.
The paper is nicely written and easy to understand. But I have several concerns:
... | 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
Neural Time-Dependent Partial Differential Equation
### Paper Abstract
Partial differential equations (PDEs) play a crucial role in studying a vast number of problems in... |
vu_oPlZSDi | ICML.cc/2020/Workshop/SAS | 2020 | Audio ALBERT: A Lite BERT for Self-supervised Learning of Audio Representation | ["Pohan Chi", "PeiHung Chung", "Tsung Han Wu", "Chun-Cheng Hsieh", "Shang-Wen Li", "Hung-yi Lee"] | For self-supervised speech processing, it is crucial to use pre-trained models as speech representation extractors. In recent works, increasing the size of the model has been utilized in acoustic model training in order to achieve better performance. In this paper, we propose Audio ALBERT, a lite version of the self-su... | ["audio albert", "lite bert", "learning", "models", "downstream tasks", "speaker", "phoneme", "latent representations", "audio representation"] | Audio ALBERT:A Lite BERT for Self-supervised Learning of Audio RepresentationPo-Han Chi1Pei-Hung Chung1Tsung-Han Wu1Chun-Cheng Hsieh1Shang-Wen Li2 3Hung-yi Lee1AbstractFor self-supervised speech processing, it is crucialto use pretrained models as speech representationextractors. In recent works, increasing the sizeof ... | 157ac1MuHdW | Review | 4: Ok but not good enough - rejection | Authors applied previously introduced BERT-like architectures as well as a similar pre-training strategy to BERT in order to obtain representations of audio for downstream tasks in the standard application setting of self-supervision.
While authors report important empirical findings in that parameter-efficient models... | 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
Audio ALBERT: A Lite BERT for Self-supervised Learning of Audio Representation
### Paper Abstract
For self-supervised speech processing, it is crucial to use pre-trained... |
cKnKJcTPRcV | ICLR.cc/2021/Conference | 2021 | HyperSAGE: Generalizing Inductive Representation Learning on Hypergraphs | ["Devanshu Arya", "Deepak Gupta", "Stevan Rudinac", "Marcel Worring"] | Graphs are the most ubiquitous form of structured data representation used in machine learning. They model, however, only pairwise relations between nodes and are not designed for encoding the higher-order relations found in many real-world datasets. To model such complex relations, hypergraphs have proven to be a natu... | ["Hypergraph", "Representation Learning", "Inductive Learning", "Geometric Deep Learning", "Aggregation Methods"] | ABSTRACTGraphs are the most ubiquitous form of structured data representation used in ma-chine learning. They model, however, only pairwise relations between nodes andare not designed for encoding the higher-order relations found in many real-worlddatasets. To model such complex relations, hypergraphs have proven to be... | VsrOjoyx3Iz | An extension of GraphSage towards Hypergraphs | 6: Marginally above acceptance threshold | Summary
In this paper, the authors study the problem of learning node embeddings for hypergraphs. While most of the existing studies consider reducing hyper-graphs into graphs, this paper studies learning embeddings directly on the hypergraphs using two stages of aggregations / sampling. The efficacy of the proposed m... | 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
HyperSAGE: Generalizing Inductive Representation Learning on Hypergraphs
### Paper Abstract
Graphs are the most ubiquitous form of structured data representation used in... |
SylCrnCcFX | ICLR.cc/2019/Conference | 2019 | Towards Robust, Locally Linear Deep Networks | ["Guang-He Lee", "David Alvarez-Melis", "Tommi S. Jaakkola"] | Deep networks realize complex mappings that are often understood by their locally linear behavior at or around points of interest. For example, we use the derivative of the mapping with respect to its inputs for sensitivity analysis, or to explain (obtain coordinate relevance for) a prediction. One key challenge is tha... | ["robust derivatives", "transparency", "interpretability"] | ABSTRACTDeep networks realize complex mappings that are often understood by their lo-cally linear behavior at or around points of interest. For example, we use thederivative of the mapping with respect to its inputs for sensitivity analysis, or toexplain (obtain coordinate relevance for) a prediction. One key challenge... | H1eDBfaunm | It seems to be an interesting problem but is the solution proposed practical? | 8: Top 50% of accepted papers, clear accept |
########## Updated Review ##########
The author(s) have presented a very good rebuttal, and I am impressed. My concerns have been addressed and my confusions have been clarified. To reflect this, I am raising my points to 8. It is a good paper, job well done. I enthusiastically recommend acceptance.
###############... | 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
Towards Robust, Locally Linear Deep Networks
### Paper Abstract
Deep networks realize complex mappings that are often understood by their locally linear behavior at or a... |
BJ_MGwqlg | ICLR.cc/2017/conference | 2017 | Rethinking Numerical Representations for Deep Neural Networks | ["Parker Hill", "Babak Zamirai", "Shengshuo Lu", "Yu-Wei Chao", "Michael Laurenzano", "Mehrzad Samadi", "Marios Papaefthymiou", "Scott Mahlke", "Thomas Wenisch", "Jia Deng", "Lingjia Tang", "Jason Mars"] | With ever-increasing computational demand for deep learning, it is critical to investigate the implications of the numeric representation and precision of DNN model weights and activations on computational efficiency. In this work, we explore unconventional narrow-precision floating-point representations as it relates ... | ["Deep learning"] | ABSTRACTWith ever-increasing computational demand for deep learning, it is critical to in-vestigate the implications of the numeric representation and precision of DNNmodel weights and activations on computational efficiency. In this work, we ex-plore unconventional narrow-precision floating-point representations as it... | SJIZu9_Sg | Review | 6: Marginally above acceptance threshold | The paper studies the impact of using customized number representations on accuracy, speed, and energy consumption of neural network inference. Several standard computer vision architectures including VGG and GoogleNet are considered for the experiments, and it is concluded that floating point representations are prefe... | 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
Rethinking Numerical Representations for Deep Neural Networks
### Paper Abstract
With ever-increasing computational demand for deep learning, it is critical to investiga... |
4RbdgBh9gE | ICLR.cc/2021/Conference | 2021 | Teaching with Commentaries | ["Aniruddh Raghu", "Maithra Raghu", "Simon Kornblith", "David Duvenaud", "Geoffrey Hinton"] | Effective training of deep neural networks can be challenging, and there remain many open questions on how to best learn these models. Recently developed methods to improve neural network training examine teaching: providing learned information during the training process to improve downstream model performance. In thi... | ["learning to teach", "metalearning", "hypergradients"] | ABSTRACTEffective training of deep neural networks can be challenging, and there remainmany open questions on how to best learn these models. Recently developed meth-ods to improve neural network training examine teaching : providing learned in-formation during the training process to improve downstream model performan... | 4qqqQxtiBja | Interesting concept that ties together disparate ways of using meta-information for network training, concerns around scalability | 6: Marginally above acceptance threshold | Summary: This paper proposes a framework for incorporating and optimising meta-information as additional parameters when optimising the validation loss for a specific network on a given task. More precisely, they instantiate a teacher network that supplies the meta-information (could be anything from attention maps, to... | 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
Teaching with Commentaries
### Paper Abstract
Effective training of deep neural networks can be challenging, and there remain many open questions on how to best learn th... |
YU0Upy5k1Te | MICCAI.org/2022/Challenge/FLARE | 2022 | DLUNet: Semi-supervised Learning based Dual-Light UNet for Multi-organ Segmentation | ["Haoran Lai", "Tao Wang", "Shuoling Zhou"] | The manual ground truth of abdominal multi-organ is labor-intensive. In order to make full use of CT data, we developed a semi-supervised learning based dual-light UNet. In the training phase, it consists of two light UNets, which make full use of label and unlabeled data simultaneously by using consistent-based learn... | ["Semi-supervised learning", "UNet", "Robust segmentation loss"] | DLUNet: Semi-supervised Learning basedDual-Light UNet for Multi-organ SegmentationHaoran Lai1, Tao Wang2, and Shuoling Zhou31Guangdong Provincial Key Laboratory of Medical Image Processing, SouthernMedical University, Guangzhou, 510515, Chinahanranlai@163.com2Guangdong Provincial Key Laboratory of Medical Image Process... | 9XBIZM1tXtn | Points are clear generally speaking, just need some minor changes | 7: Good paper, accept | The paper's structure and grammar are good. It's very easy to read about. And I have some suggestion for you about the format.
1. The format of Tabel 3.,Tabel 4. and the Tabel 5. makes me look inconvenient. and you can combine them together.
2. As for the Fig3. , you can add the case without ssl. | 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
DLUNet: Semi-supervised Learning based Dual-Light UNet for Multi-organ Segmentation
### Paper Abstract
The manual ground truth of abdominal multi-organ is labor-intensi... |
SkFvV0yC- | ICLR.cc/2018/Conference | 2018 | Network Iterative Learning for Dynamic Deep Neural Networks via Morphism | ["Tao Wei", "Changhu Wang", "Chang Wen Chen"] | In this research, we present a novel learning scheme called network iterative learning for deep neural networks. Different from traditional optimization algorithms that usually optimize directly on a static objective function, we propose in this work to optimize a dynamic objective function in an iterative fashion capa... | ["Network Iterative Learning", "Morphism"] | ABSTRACTIn this research, we present a novel learning scheme called network iterative learn-ingfor deep neural networks. Different from traditional optimization algorithmsthat usually optimize directly on a static objective function, we propose in this workto optimize a dynamic objective function in an iterative fashio... | By4qQIQ-f | Interesting paper, but requires more clarifications on novelty and experiment results | 5: Marginally below acceptance threshold | This paper proposed an iterative learning scheme to train a very deep convolutional neural network. Instead of learning a deep network from scratch, the authors proposed to gradually increase the depth of the network while transferring the knowledge obtained from the shallower network by applying network morphism.
Ov... | 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
Network Iterative Learning for Dynamic Deep Neural Networks via Morphism
### Paper Abstract
In this research, we present a novel learning scheme called network iterative... |
wiSgdeJ29ee | ICLR.cc/2021/Conference | 2021 | Fine-Tuning Offline Reinforcement Learning with Model-Based Policy Optimization | ["Adam Villaflor", "John Dolan", "Jeff Schneider"] | In offline reinforcement learning (RL), we attempt to learn a control policy from a fixed dataset of environment interactions. This setting has the potential benefit of allowing us to learn effective policies without needing to collect additional interactive data, which can be expensive or dangerous in real-world syste... | ["Offline Reinforcement Learning", "Model-Based Reinforcement Learning", "Off-policy Reinforcement Learning", "uncertainty estimation"] | ABSTRACTIn offline reinforcement learning (RL), we attempt to learn a control policy from afixed dataset of environment interactions. This setting has the potential benefit ofallowing us to learn effective policies without needing to collect additional inter-active data, which can be expensive or dangerous in real-worl... | kY_l2uOooiH | Some part of the projects feel rushed and fail to clearly show the benefits over baseline. | 4: Ok but not good enough - rejection | ### Summary
The paper describes a novel method to combine two different school of thoughts for improving offline reinforcement learning. The model free part relies on keeping policy changes local and model based part uses model uncertainties to ‘fine tune’ to obtain final offline RL policy.
### Strong Points
1. Pap... | 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
Fine-Tuning Offline Reinforcement Learning with Model-Based Policy Optimization
### Paper Abstract
In offline reinforcement learning (RL), we attempt to learn a control ... |
J40FkbdldTX | ICLR.cc/2021/Conference | 2021 | Exploring single-path Architecture Search ranking correlations | ["Kevin Alexander Laube", "Andreas Zell"] | Recently presented benchmarks for Neural Architecture Search (NAS) provide the results of training thousands of different architectures in a specific search space, thus enabling the fair and rapid comparison of different methods.
Based on these results, we quantify the ranking correlations of single-path architecture s... | ["Neural Architecture Search", "AutoML", "Neural Networks"] | ABSTRACTRecently presented benchmarks for Neural Architecture Search (NAS) provide theresults of training thousands of different architectures in a specific search space,thus enabling the fair and rapid comparison of different methods. Based on theseresults, we quantify the ranking correlations of single-path architect... | vJzEWZvDffO | An empirical study on the ranking correlation of several NAS mechanisms in the singe-path setup | 5: Marginally below acceptance threshold | This paper introduces an empirical study on the ranking correlation in the singe-path setup. Following the paradigm of NAS-Bench-201, the authors test the Kendall Tau correlation of networks from NAS training and networks from standalone training.
In general, I appreciate the authors' effort in bringing more infrast... | 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
Exploring single-path Architecture Search ranking correlations
### Paper Abstract
Recently presented benchmarks for Neural Architecture Search (NAS) provide the results ... |
ZlIfK1wCubc | ICLR.cc/2021/Conference | 2021 | Contrasting distinct structured views to learn sentence embeddings | ["Antoine Simoulin", "Benoit Crabb\u00e9"] | We propose a self-supervised method that builds sentence embeddings from the combination of diverse explicit syntactic structures of a sentence. We assume structure is crucial to build consistent representations as we expect sentence meaning to be a function from both syntax and semantic aspects. In this perspective, ... | ["Sentence", "Embeddings", "Structure", "Contrastive", "Multi-views"] | ABSTRACTWe propose a self-supervised method that builds sentence embeddings from thecombination of diverse explicit syntactic structures of a sentence. We assumestructure is crucial to build consistent representations as we expect sentence mean-ing to be a function from both syntax and semantic aspects. In this perspec... | cbeO1Xh7MJE | Is explicit syntax really necessary? | 3: Clear rejection | The authors introduce a pretrianing paradigm based on contrastive learning between multiple syntactic views of the same sentence. The method maximizes representations between different setence encoders when given the same sentence, and minimize the similarity to all other sentence repre sentations. The results on the 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
Contrasting distinct structured views to learn sentence embeddings
### Paper Abstract
We propose a self-supervised method that builds sentence embeddings from the combi... |
ryl1r1BYDS | ICLR.cc/2020/Conference | 2020 | Multiagent Reinforcement Learning in Games with an Iterated Dominance Solution | ["Yoram Bachrach", "Tor Lattimore", "Marta Garnelo", "Julien Perolat", "David Balduzzi", "Thomas Anthony", "Satinder Singh", "Thore Graepel"] | Multiagent reinforcement learning (MARL) attempts to optimize policies of intelligent agents interacting in the same environment. However, it may fail to converge to a Nash equilibrium in some games. We study independent MARL under the more demanding solution concept of iterated elimination of strictly dominated strat... | ["multiagent", "reinforcement learning", "iterated dominance", "mechanism design", "Nash equilibrium"] | ABSTRACTMultiagent reinforcement learning (MARL) attempts to optimize policies of intelli-gent agents interacting in the same environment. However, it may fail to convergeto a Nash equilibrium in some games. We study independent MARL under themore demanding solution concept of iterated elimination of strictly dominated... | HJeqKrkBcS | Official Blind Review #2 | 3: Weak Reject | This paper studies independent multi-agent reinforcement learning (MARL) in dominance solvable games. The main contribution of this paper is that the authors have proved the convergence to the iterated dominance solution for two RL algorithms: REINFORCE (Section 3.1, binary action case only) and Importance Weighted Mon... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Multiagent Reinforcement Learning in Games with an Iterated Dominance Solution
### Paper Abstract
Multiagent reinforcement learning (MARL) attempts to optimize policies ... | |
HkeccWB854 | vivoconference.org/VIVO/2019/Conference | 2019 | Results of VIVO Community Feedback Survey | ["Michael Conlon", "Kristi Holmes", "Daniel W Hook", "Dean B Krafft", "Mark P Newton", "Julia Trimmer"] | In early 2018, the VIVO Leadership group brought together parties from across the broader VIVO community to Duke University to discuss critical aspects of VIVO as both a product and a community. At the meeting, a number of working groups were created to do deeper work on a set of focus areas to help inform the VIVO le... | ["VIVO Community", "guiding principles", "community feedback"] | HJlu1ANlRV | Community Feedback Survey | 10: Top 5% of accepted papers, seminal paper | This proposal is highly relevant to the VIVO community and to the conference. The work that will be reported here represents an important inflection point for the VIVO project on its path toward achieving the next level of maturity in relevance, governance, process, and product. As such, it's necessary to reflect on ... | 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
Results of VIVO Community Feedback Survey
### Paper Abstract
In early 2018, the VIVO Leadership group brought together parties from across the broader VIVO community to ... | |
HkNN7nR5Ym | ICLR.cc/2019/Conference | 2019 | Associate Normalization | ["Song-Hao Jia", "Ding-Jie Chen", "Hwann-Tzong Chen"] | Normalization is a key technique for training deep neural networks. It improves the stability of the training process and thus makes the networks easier to train. However, in typical normalization methods, the rescaling parameters that control the mean and variance of the output do not associate with any input informat... | ["assocnorm", "parameters", "normalization module", "key technique", "deep neural networks", "stability", "training process", "networks easier", "typical normalization methods"] | ABSTRACTNormalization is a key technique to improve the training of deep neural net-works. However, the existing normalization methods often merely rely on theback-propagation processes to learn the rescaling parameters. That is, the exist-ing methods treat all different input features under the same distribution, whic... | Bkg0R-Yp3Q | Input-dependent normalization method | 3: Clear rejection | The authors proposed a normalization method that the rescaling parameters are predicted by the input data. It is shown to use an auto-encoder like network. As the proposed approach uses input-level normalization, it seems to be robust on changing the batch size during training and also showing better performance compar... | 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
Associate Normalization
### Paper Abstract
Normalization is a key technique for training deep neural networks. It improves the stability of the training process and thus... |
nkyvisAIEC | NoDaLiDa/2023/Conference | 2023 | Abstractive Text Summarization for Icelandic | ["\u00de\u00f3r Sverrisson", "Hafsteinn Einarsson"] | In this work, we studied methods for automatic abstractive summarization in a low-resource setting using Icelandic text, which is morphologically rich and has limited data compared to languages such as English. We collected and published the first publicly available abstractive summarization dataset for Icelandic and u... | ["abstractive summarization", "language model", "corpus"] | Abstractive Text Summarization for IcelandicÞór SverrissonDepartment of Computer ScienceUniversity of IcelandIcelandths220@hi.isHafsteinn EinarssonDepartment of Computer ScienceUniversity of IcelandIcelandhafsteinne@hi.isAbstractIn this work, we studied methods forautomatic abstractive summarization in alow-resource s... | 7NRiGmjKhc | Solid contribution with thorough model exploration + novel and useful resource creation | 8: Top 50% of accepted papers, clear accept | ## Summary
This paper explores for abstractive summarization models for the Icelandic language. With the lack of a suitable dataset for abstractive modeling (prior work being extractive) a new dataset (RRN) is introduced, collecting and manually curating 4k articles from the Icelandic national broadcasting service (RÚV... | 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
Abstractive Text Summarization for Icelandic
### Paper Abstract
In this work, we studied methods for automatic abstractive summarization in a low-resource setting using ... |
BkZu9wooz | MIDL.amsterdam/2018/Conference | 2018 | OBELISK - One Kernel to Solve Nearly Everything: Unified 3D Binary Convolutions for Image Analysis | ["Mattias P. Heinrich", "Ozan Oktay", "Nassim Bouteldja"] | Deep networks have set the state-of-the-art in most image analysis tasks by replacing handcrafted features with learned convolution filters within end-to-end trainable architectures. Still, the specifications of a convolutional network are subject to much manual design - the shape and size of the receptive field for co... | ["deformable convolution", "sparsity", "deep learning", "segmentation"] | OBELISK – One Kernel to Solve Nearly Everything:Unified 3D Binary Convolutions for Image AnalysisMattias P. Heinrich1, Ozan Oktay2, and Nassim Bouteldja11Institute of Medical Informatics, University of Lübeck, DE2Babylon Health, London, UKheinrich@imi.uni-luebeck.deAbstractDeep networks have set the state-of-the-art i... | HJeQFu8lyQ | Review | 5: Top 15% of accepted papers, strong accept | This paper is interesting because learning filter offsets directly is still relatively unexplored. The authors follow ref 3, is this the only paper that tried to do this?
The binary part, the B in OBELISK, is not well explained. The paper says: "The unary variant refers to using only one 3D offset per filter coeffici... | 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
OBELISK - One Kernel to Solve Nearly Everything: Unified 3D Binary Convolutions for Image Analysis
### Paper Abstract
Deep networks have set the state-of-the-art in most... |
SJeS16EKPr | ICLR.cc/2020/Conference | 2020 | Learning relevant features for statistical inference | ["C\u00e9dric B\u00e9ny"] | We introduce an new technique to learn correlations between two types of data.
The learned representation can be used to directly compute the expectations of functions over one type of data conditioned on the other, such as Bayesian estimators and their standard deviations.
Specifically, our loss function teaches two ... | ["unsupervised learning", "non-parametric probabilistic model", "singular value decomposition", "fisher information metric", "chi-squared distance"] | ABSTRACTGiven two views of data, we consider the problem of finding the features of oneview which can be most faithfully inferred from the other. We find that these arealso the most correlated variables in the sense of deep canonical correlation anal-ysis (DCCA). Moreover, we show that these variables can be used to co... | rklS9UB6YS | Official Blind Review #1 | 3: Weak Reject | The paper considers how to learn correlations between two spaces, e.g., input/output, in order to generate data in one space conditioned on values from the other. This is performed by modeling features with neural networks and optimizing an objective function that maximizes a measure of correlation between the features... | <|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 relevant features for statistical inference
### Paper Abstract
We introduce an new technique to learn correlations between two types of data. The learned repres... | |
1o5Bzo7BO3v | AAAI.org/2022/Workshop/AdvML | 2022 | Attention for Adversarial Attacks: Learning from your Mistakes | ["Florian Jaeckle", "Aleksandr Agadzhanov", "JINGYUE LU", "M. Pawan Kumar"] | In order to apply Neural Networks in safety-critical settings, such as healthcare or autonomous driving, we need to be able to analyse their robustness against adversarial attacks. As complete verification is often computationally prohibitive, we rely on cheap and effective adversarial attacks to estimate their robustn... | ["Adversarial Examples", "Attention Mechanism", "Graph Neural Networks", "PGD Attack", "Adversarial Attacks"] | Attention for Adversarial Attacks: Learning from your MistakesFlorian Jaeckle*, Aleksandr Agadzhanov*, Jingyue Lu, M. Pawan Kumar,Department of Engineering ScienceUniversity of Oxford{florian,pawan }@robots.ox.ac.uk, alexagadzhanov97@gmail.com, jingyue.lu@spc.ox.ac.ukAbstractIn order to apply Neural Networks in safety-... | _PrRUNkRt7U | A good idea to introduce GNN into adversarial attacks. The efficiency of regular adversarial attackers is improved. | 8: Top 50% of accepted papers, clear accept | Pros:
1. The writing is good.
2. The explanation of using GNN is clear.
3. The experimental results are convincing.
4. Using GNN as an attention mechanism to find the pitfalls of a classifier is a good idea. Maybe it can be used in more important scenarios. | 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
Attention for Adversarial Attacks: Learning from your Mistakes
### Paper Abstract
In order to apply Neural Networks in safety-critical settings, such as healthcare or au... |
Hkx6hANtwH | ICLR.cc/2020/Conference | 2020 | LambdaNet: Probabilistic Type Inference using Graph Neural Networks | ["Jiayi Wei", "Maruth Goyal", "Greg Durrett", "Isil Dillig"] | As gradual typing becomes increasingly popular in languages like Python and TypeScript, there is a growing need to infer type annotations automatically. While type annotations help with tasks like code completion and static error catching, these annotations cannot be fully inferred by compilers and are tedious to annot... | ["Type inference", "Graph neural network", "Programming languages", "Pointer network"] | ABSTRACTAsgradual typing becomes increasingly popular in languages like Python andTypeScript, there is a growing need to infer type annotations automatically. Whiletype annotations help with tasks like code completion and static error catching,these annotations cannot be fully determined by compilers and are tedious to... | BJgF7g7jFr | Official Blind Review #2 | 8: Accept | = Summary
A method to predict likely type of program variables in TypeScript is presented. It consists of a translation of a program's type constraints and defined objects into a (hyper)graph, and a specialised neural message passing architecture to learn from the generated graphs. Experiments show that the method subs... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
LambdaNet: Probabilistic Type Inference using Graph Neural Networks
### Paper Abstract
As gradual typing becomes increasingly popular in languages like Python and TypeSc... | |
Skf5qiC5KQ | ICLR.cc/2019/Conference | 2019 | A Unified View of Deep Metric Learning via Gradient Analysis | ["Xun Wang", "Xintong Han", "Weilin Huang", "Dengke Dong", "Matthew R. Scott"] | Loss functions play a pivotal role in deep metric learning (DML). A large variety of loss functions have been proposed in DML recently. However, it remains difficult to answer this question: what are the intrinsic differences among these loss functions?This paper answers this question by proposing a unified perspective... | ["metric learning", "gradient equivalence", "image retrieval"] | ABSTRACTLoss functions play a pivotal role in deep metric learning (DML). A large varietyof loss functions have been proposed in DML recently. However, it remains dif-ficult to answer this question: what are the intrinsic differences among these lossfunctions? This paper answers this question by proposing a unified per... | rkexzyc82Q | Strong experimental results but the theory is not that strong | 6: Marginally above acceptance threshold | The paper analyses the gradients some deep metric learning (DML) approaches. Most of them optimize over pair-wise or triplet-wise constraints.
The main claim of the paper is that, during training, most DML approaches ignore some informative pairs of examples or do not take into account the value of the distance gap bet... | 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 Unified View of Deep Metric Learning via Gradient Analysis
### Paper Abstract
Loss functions play a pivotal role in deep metric learning (DML). A large variety of loss... |
b1tl3aOt2R2 | robot-learning.org/CoRL/2023/Conference | 2023 | GNFactor: Multi-Task Real Robot Learning with Generalizable Neural Feature Fields | ["Yanjie Ze", "Ge Yan", "Yueh-Hua Wu", "Annabella Macaluso", "Yuying Ge", "Jianglong Ye", "Nicklas Hansen", "Li Erran Li", "Xiaolong Wang"] | It is a long-standing problem in robotics to develop agents capable of executing diverse manipulation tasks from visual observations in unstructured real-world environments. To achieve this goal, the robot will need to have a comprehensive understanding of the 3D structure and semantics of the scene. In this work, we ... | ["Robotic Manipulation", "Neural Radiance Field", "Behavior Cloning"] | GNFactor: Multi-Task Real Robot Learning withGeneralizable Neural Feature FieldsYanjie Ze1∗Ge Yan2∗Yueh-Hua Wu2∗Annabella Macaluso2Yuying Ge3Jianglong Ye2Nicklas Hansen2Li Erran Li4Xiaolong Wang21Shanghai Jiao Tong University2UC San Diego3University of Hong Kong4AWS AI, Amazon∗Equal Contributionyanjieze.com/GNFactorAbs... | -RQZNwAhrqs | Strengths:
- Using NeRF rendering and diffusion consistency to improve the generalization performance of behavior-cloning agents is an interesting and novel approach. Prior works have either looked at 3D reconstruction (eg. MIRA) or pre-trained features (R3M) to improve policies, but GNFactor simply puts them together.... | 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
GNFactor: Multi-Task Real Robot Learning with Generalizable Neural Feature Fields
### Paper Abstract
It is a long-standing problem in robotics to develop agents capable ... | ||
vLaHRtHvfFp | ICLR.cc/2021/Conference | 2021 | PDE-Driven Spatiotemporal Disentanglement | ["J\u00e9r\u00e9mie Don\u00e0", "Jean-Yves Franceschi", "sylvain lamprier", "patrick gallinari"] | A recent line of work in the machine learning community addresses the problem of predicting high-dimensional spatiotemporal phenomena by leveraging specific tools from the differential equations theory. Following this direction, we propose in this article a novel and general paradigm for this task based on a resolution... | ["disentanglement", "spatiotemporal prediction", "representation learning", "dynamical systems", "separation of variables"] | ABSTRACTA recent line of work in the machine learning community addresses the problemof predicting high-dimensional spatiotemporal phenomena by leveraging specifictools from the differential equations theory. Following this direction, we propose inthis article a novel and general paradigm for this task based on a resol... | unq7FMrkb8n | Well-written paper with interesting but insufficient experiments | 7: Good paper, accept | This paper proposes a new framework for spatiotemporal disentanglement. In particular, it contributes to the disentanglement of content and dynamics using neural ODEs.
Strength:
1. I think it is a well-written paper with interesting experimental design, especially the “swap” and “multi-view” experiments that measure... | 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
PDE-Driven Spatiotemporal Disentanglement
### Paper Abstract
A recent line of work in the machine learning community addresses the problem of predicting high-dimensional... |
tiqI7w64JG2 | ICLR.cc/2021/Conference | 2021 | On Graph Neural Networks versus Graph-Augmented MLPs | ["Lei Chen", "Zhengdao Chen", "Joan Bruna"] | From the perspectives of expressive power and learning, this work compares multi-layer Graph Neural Networks (GNNs) with a simplified alternative that we call Graph-Augmented Multi-Layer Perceptrons (GA-MLPs), which first augments node features with certain multi-hop operators on the graph and then applies learnable no... | ["Graph Neural Networks", "expressive power", "feature propagation", "rooted graphs", "attributed walks", "community detection", "depth separation"] | ABSTRACTFrom the perspectives of expressive power and learning, this work comparesmulti-layer Graph Neural Networks (GNNs) with a simplified alternative thatwe call Graph-Augmented Multi-Layer Perceptrons (GA-MLPs), which first aug-ments node features with certain multi-hop operators on the graph and then applieslearna... | ZQCmV_0Gzxc | Expressvity of graph MLPs with operators | 7: Good paper, accept | The paper compares graph neural networks (GNNs) with graph-augmented multi-layer perceptrons (GA-MLPs) where GA-MLPs are MLPs over nodes with additional node features computed over the graph. The paper contains theoretical results and experimental results for graph isomorphism testing and for node level functions. In 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
On Graph Neural Networks versus Graph-Augmented MLPs
### Paper Abstract
From the perspectives of expressive power and learning, this work compares multi-layer Graph Neur... |
BJg_fnRqF7 | ICLR.cc/2019/Conference | 2019 | Deep clustering based on a mixture of autoencoders | ["Shlomo E. Chazan", "Sharon Gannot", "Jacob Goldberger"] | In this paper we propose a Deep Autoencoder Mixture Clustering (DAMIC) algorithm. It is based on a mixture of deep autoencoders where each cluster is represented by an autoencoder. A clustering network transforms the data into another space and then selects one of the clusters. Next, the autoencoder associated with thi... | ["deep clustering", "mixture of experts", "mixture of autoencoders"] | ABSTRACTIn this paper we propose a Deep Autoencoder Mixture Clustering (DAMIC) al-gorithm. It is based on a mixture of deep autoencoders where each cluster isrepresented by an autoencoder. A clustering network transforms the data into an-other space and then selects one of the clusters. Next, the autoencoder associated... | B1esAOTZhm | interesting but the performance not good enough. | 5: Marginally below acceptance threshold | The paper presents a deep clustering method which represents each cluster with different auto-encoders. The idea is simple but seems interesting. Benefiting from the neural network, the proposed model works in an end-to-end manner. It also can be used to cluster new incoming data without redoing the whole clustering pr... | 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
Deep clustering based on a mixture of autoencoders
### Paper Abstract
In this paper we propose a Deep Autoencoder Mixture Clustering (DAMIC) algorithm. It is based on a ... |
H1xAXpnX5E | icaps-conference.org/ICAPS/2019/Workshop/XAIP | 2019 | Combining Cognitive and Affective Measures with Epistemic Planning for Explanation Generation | ["Anonymous"] | This paper presents an overview of the EPSRC-funded project Start Making Sense, which is investigating explainability and trust maintenance in interactive and autonomous systems. This project brings together experimental research in cognitive science involving cooperative joint action with the practical construction of... | ["Explainable planning", "epistemic planning", "cognitive science", "affective measure in human communication"] | Combining Cognitive and Affective Measures with Epistemic Planning forExplanation GenerationRonald P. A. PetrickDepartment of Computer ScienceHeriot-Watt UniversityEdinburgh, Scotland, United KingdomR.Petrick@hw.ac.ukSara Dalzel-Job andRobin L. HillSchool of InformaticsUniversity of EdinburghEdinburgh, Scotland, United... | Syeg9Lqbh4 | Interesting Position Paper discussing combining Epistemic Planning and HSR to pursue effective explanations | 4: Top 50% of accepted papers, clear accept | I'm not a fan of seeing the abstract almost verbatim later on in the introduction section. I get it, but why have it in here twice?
Related Work:
I'm surprised you don't have any mention of Epistemic Planning here yet, or how it ties in. What part of Epistemic Planning is this paper going to take? Can I just use it ... | 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
Combining Cognitive and Affective Measures with Epistemic Planning for Explanation Generation
### Paper Abstract
This paper presents an overview of the EPSRC-funded proj... |
ryj0790hb | ICLR.cc/2018/Conference | 2018 | Incremental Learning through Deep Adaptation | ["Amir Rosenfeld", "John K. Tsotsos"] | Given an existing trained neural network, it is often desirable to learn new capabilities without hindering performance of those already learned. Existing approaches either learn sub-optimal solutions, require joint training, or incur a substantial increment in the number of parameters for each added task, typically as... | ["Transfer Learning", "Learning without forgetting", "Multitask Learning"] | ABSTRACTGiven an existing trained neural network, it is often desirable to learn new capabil-ities without hindering performance of those already learned. Existing approacheseither learn sub-optimal solutions, require joint training, or incur a substantial in-crement in the number of parameters for each added task, typ... | BJJTve9gM | - | 6: Marginally above acceptance threshold | This paper proposes to adapt convnet representations to new tasks while avoiding catastrophic forgetting by learning a per-task “controller” specifying weightings of the convolution-al filters throughout the network while keeping the filters themselves fixed.
Pros
The proposed approach is novel and broadly applicabl... | 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
Incremental Learning through Deep Adaptation
### Paper Abstract
Given an existing trained neural network, it is often desirable to learn new capabilities without hinderi... |
VscdYkKgwdH | robot-learning.org/CoRL/2023/Conference | 2023 | Neural Graph Control Barrier Functions Guided Distributed Collision-avoidance Multi-agent Control | ["Songyuan Zhang", "Kunal Garg", "Chuchu Fan"] | We consider the problem of designing distributed collision-avoidance multi-agent control in large-scale environments with potentially moving obstacles, where a large number of agents are required to maintain safety using only local information and reach their goals. This paper addresses the problem of collision avoidan... | ["Distributed control", "Control barrier functions", "Graph neural networks"] | Neural Graph Control Barrier Functions GuidedDistributed Collision-avoidance Multi-agent ControlSongyuan ZhangDepartment of Aeronautics and AstronauticsMassachusetts Institute of Technologyszhang21@mit.eduKunal GargDepartment of Aeronautics and AstronauticsMassachusetts Institute of Technologykgarg@mit.eduChuchu FanDep... | koSrYEYoTlG | ### Strengths
S1) Interesting deep learning approach combining multiple different methods like CBFs, GNNs and attention mechanism.
S2) Highly scalable.
S3) Training can be done in a relatively short amount of time (only 1 hour).
### Weaknesses
W1) The paper states that its controller is safe, although it has no safe... | 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
Neural Graph Control Barrier Functions Guided Distributed Collision-avoidance Multi-agent Control
### Paper Abstract
We consider the problem of designing distributed col... | ||
Skft7cijM | MIDL.amsterdam/2018/Conference | 2018 | Attention U-Net: Learning Where to Look for the Pancreas | ["Ozan Oktay", "Jo Schlemper", "Loic Le Folgoc", "Matthew Lee", "Mattias Heinrich", "Kazunari Misawa", "Kensaku Mori", "Steven McDonagh", "Nils Y Hammerla", "Bernhard Kainz", "Ben Glocker", "Daniel Rueckert"] | We propose a novel attention gate (AG) model for medical imaging that automatically learns to focus on target structures of varying shapes and sizes. Models trained with AGs implicitly learn to suppress irrelevant regions in an input image while highlighting salient features useful for a specific task. This enables us ... | ["Convolutional Neural Networks", "Attention Gating", "Medical Image Segmentation"] | Attention U-Net:Learning Where to Look for the PancreasOzan Oktay1,5, Jo Schlemper1, Loic Le Folgoc1, Matthew Lee4, Mattias Heinrich3,Kazunari Misawa2, Kensaku Mori2, Steven McDonagh1, Nils Y Hammerla5,Bernhard Kainz1, Ben Glocker1, and Daniel Rueckert11Biomedical Image Analysis Group, Imperial College London, London, ... | H1y2ckb0z | Attention U-Net: Learning Where to Look for the Pancreas | 5: Top 15% of accepted papers, strong accept | The paper presents a simple yet effective gate block, which can be plugged in to implement an attention mechanisms in existing convolutional neural network architectures.
The U-Net architecture is considered in this paper, applied to the specific problem of organs segmentation in thorax CT images.
Results show the effe... | 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
Attention U-Net: Learning Where to Look for the Pancreas
### Paper Abstract
We propose a novel attention gate (AG) model for medical imaging that automatically learns to... |
4IU_xHbLiH | ICLR.cc/2021/Conference | 2021 | Fast and Differentiable Matrix Inverse and Its Extension to SVD | ["Xingyu Xie", "Hao Kong", "Jianlong Wu", "Guangcan Liu", "Zhouchen Lin"] | Matrix inverse (Minv) and singular value decomposition (SVD) are among the most widely used matrix operations in massive data analysis, machine learning, and statistics. Although well-studied, they still encounter difficulties in practical use due to inefficiency and non-differentiability. In this paper, we aim at so... | ["learning-based optimization", "learning-based iterative method", "differentiable matrix inverse", "differentiable singular value decomposition", "convergence", "generalization"] | ABSTRACTMatrix inverse (Minv) and singular value decomposition (SVD) are among themost widely used matrix operations in massive data analysis, machine learning,and statistics. Although well-studied, they still encounter difficulties in practicaluse due to inefficiency and non-differentiability. In this paper, we aim at... | lsDy1rq1Ihn | Need more experimental comparisons | 5: Marginally below acceptance threshold | This paper proposes a deep neural network-based approach that computes matrix inverse and SVD for massive data. Let A be a given matrix and X be a matrix corresponding to A^-1, the proposed approach is based on the observation that (AX)^-1 can be represented as the Neumann series. By introducing a learnable coefficient... | 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
Fast and Differentiable Matrix Inverse and Its Extension to SVD
### Paper Abstract
Matrix inverse (Minv) and singular value decomposition (SVD) are among the most widely... |
ioa3UOmt-VH | ICLR.cc/2023/BlogPosts | 2023 | Autoregressive Renaissance in Neural PDE Solvers | ["Yolanne Yi Ran Lee"] | Recent developments in the field of neural partial differential equation (PDE) solvers have placed a strong emphasis on neural operators. However, the paper "Message Passing Neural PDE Solver" by Brandstetter et al. published in ICLR 2022 revisits autoregressive models and designs a message passing graph neural network... | ["neural PDE solvers", "message passing", "autoregressive models", "graph neural network"] | fZc4prhbmO | An interesting read that could be improved | 6: Marginally above acceptance threshold | ## Quality and novelty:
### *clarity and pedagogy of the exposition, new theoretical or practical insights, reproduction/extension of experiments, etc.*
This blogpost is an interesting read that is mostly self-contained and provides a nice introduction to neural PDE solver, with a particular emphasis on the ICLR 2022... | 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
Autoregressive Renaissance in Neural PDE Solvers
### Paper Abstract
Recent developments in the field of neural partial differential equation (PDE) solvers have placed a ... | |
Bk8BvDqex | ICLR.cc/2017/conference | 2017 | Metacontrol for Adaptive Imagination-Based Optimization | ["Jessica B. Hamrick", "Andrew J. Ballard", "Razvan Pascanu", "Oriol Vinyals", "Nicolas Heess", "Peter W. Battaglia"] | Many machine learning systems are built to solve the hardest examples of a particular task, which often makes them large and expensive to run---especially with respect to the easier examples, which might require much less computation. For an agent with a limited computational budget, this "one-size-fits-all" approach m... | ["Deep learning", "Reinforcement Learning", "Optimization"] | ABSTRACTMany machine learning systems are built to solve the hardest examples of a par-ticular task, which often makes them large and expensive to run—especially withrespect to the easier examples, which might require much less computation. Foran agent with a limited computational budget, this “one-size-fits-all” appro... | H1sj61Qrx | Pondering on expert advice for control | 7: Good paper, accept | This paper introduces an approach to reinforcement learning and control wherein, rather than training a single controller to perform a task, a metacontroller with access to a base-level controller and a number of accessory « experts » is utilized. The job of the metacontroller is to decide how many times to call the co... | 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
Metacontrol for Adaptive Imagination-Based Optimization
### Paper Abstract
Many machine learning systems are built to solve the hardest examples of a particular task, wh... |
1EuxRTe0WN | ICML.cc/2021/Workshop/AutoML | 2021 | PonderNet: Learning to Ponder | ["Andrea Banino", "Jan Balaguer", "Charles Blundell"] | In standard neural networks the amount of computation used grows with the size of the inputs, but not with the complexity of the problem being learnt. To overcome this limitation we introduce PonderNet, a new algorithm that learns to adapt the amount of computation based on the complexity of the problem at hand. Ponder... | ["adaptive computation", "probabilistic", "recurrent network", "transformer", "ponder"] | 8th ICML Workshop on Automated Machine Learning (2021)PonderNet: Learning to PonderAndrea BaninoDeepMindLondon, UKabanino@deepmind.comJan Balaguer*DeepMindLondon, UKjan@deepmind.comCharles BlundellDeepMindLondon, UKcblundell@deepmind.comAbstractIn standard neural networks the amount of computation used grows with the s... | a0py0yTzsiT | Interesting contribution, although the impact is yet unclear | 6: Marginally above acceptance threshold | The paper proposes a novel method that halts the number of computational steps in order to trade-off between accuracy and computation cost. The method is an incremental contribution on top of the Adaptive Computation Time approach (Graves, 2016) with a novel focus on handling its stability. Concretely, the paper intrid... | 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
PonderNet: Learning to Ponder
### Paper Abstract
In standard neural networks the amount of computation used grows with the size of the inputs, but not with the complexit... |
pOVaPJkP0E | logconference.io/LOG/2022/Conference | 2022 | Graph Neural Network Expressivity and Meta-Learning for Molecular Property Regression | ["Haitz S\u00e1ez de Oc\u00e1riz Borde", "Federico Barbero"] | We demonstrate the applicability of model-agnostic algorithms for meta-learning, specifically Reptile, to GNN models in molecular regression tasks. Using meta-learning we are able to learn new chemical prediction tasks with only a few model updates, as compared to using randomly initialized GNNs which require learning ... | ["Meta-Learning", "Molecular Property Prediction", "Graph Neural Networks"] | Graph Neural Network Expressivity and Meta-Learning forMolecular Property RegressionHaitz Sáez de Ocáriz Borde∗University of CambridgeFederico Barbero∗University of CambridgeAbstractWe demonstrate the applicability of model-agnostic algorithms for meta-learning,specifically Reptile, to GNN models in molecular regress... | Vsy4-RMTj7 | Summary:
This work investigates meta learning for molecular property predictions with GNNs. In particular, the authors adopt Reptile on quantum chemistry (QM) property predictions with GCN, GAT, and MPNN. Experiments show that meta-learning significantly improves the performance of the GNNs under a few-shot learning s... | <|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 Network Expressivity and Meta-Learning for Molecular Property Regression
### Paper Abstract
We demonstrate the applicability of model-agnostic algorithms fo... | |||
I86Og5lqg | graphicsinterface.org/Graphics_Interface/2020/Conference | 2020 | Scope and Impact of Visualization in Training Professionals in Academic Medicine | ["Venkat Bandi", "Debajyoti Mondal", "Brent Thoma"] | Professional training often requires need-based scheduling and observation-based assessment. In this paper, we present a visualization platform for managing such training data in a medical education domain, where the learners are resident physicians and the educators are certified doctors. The system was developed thro... | ["Visualization in Education", "Qualitative Evaluation", "Task and Requirements Analysis", "Visual Design", "Visualization System and Toolkit Design"] | ABSTRACTProfessional training often requires need-based scheduling andobservation-based assessment. In this paper, we present a visualiza-tion platform for managing such training data in a medical educationdomain, where the learners are resident physicians and the educatorsare certified doctors. The system was develope... | 1GFncdPt6t | review | 7: Good paper, accept | This paper presents a design study of a visualization system for tracking and reviewing resident physicians’ performance in a medical training program. The design of the system was through four focus groups and the evaluation was through a four-month deployment study. User behavior changes were observed after introduci... | 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
Scope and Impact of Visualization in Training Professionals in Academic Medicine
### Paper Abstract
Professional training often requires need-based scheduling and obse... |
EfhpoMWSqUN | AAAI.org/2022/Workshop/AdvML | 2022 | Broad Adversarial Training with Data Augmentation in the Output Space | ["Nils Worzyk", "Stella Yu"] | In image classification, data augmentation and the usage of additional data has been shown to increase the efficiency of clean training and the accuracy of the resulting model. However, this does not prevent models from being fooled by adversarial manipulations. To increase the robustness, Adversarial Training (AT) is ... | ["Adversarial Training", "Adversarial", "Adversarial Defence", "Output Space"] | Broad Adversarial Training with Data Augmentation in the Output SpaceNils Worzyk Stella X. YuUC Berkeley / International Computer Science Institute, Berkeley, CA, USAnilswo@icsi.berkeley.edu , stellayu@berkeley.eduAbstractIn image classification, data augmentation and the usage ofadditional data has been shown to incre... | KPriC9-sIVw | Review for the paper "Broad Adversarial Training with Data Augmentation in the Output Space" | 7: Good paper, accept | This paper proposes an interesting approach for adversarial training, that the data augmentation is applied to the output space.
Advantages:
- Figure 1 is clear to show the motivation of this paper, while is also reasonable.
- The proposed technique is simple and general for various adversarial training paradigms, inc... | 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
Broad Adversarial Training with Data Augmentation in the Output Space
### Paper Abstract
In image classification, data augmentation and the usage of additional data has ... |
BJema04FvS | ICLR.cc/2020/Conference | 2020 | VISUALIZING POINT CLOUD CLASSIFIERS BY MORPHING POINT CLOUDS INTO POTATOES | ["Ziwen Chen", "Wenxuan Wu", "Zhongang Qi", "Fuxin Li"] | Recently, various networks that operate directly on point clouds have been proposed. It is of interest to us what features are utilized in those classifiers for their predictions. In this paper, we propose a novel approach to visualize important features used in classification decisions from point cloud networks. Follo... | ["point cloud", "3D computer vision", "visualization"] | ABSTRACTRecently, various networks that operate directly on point clouds have been pro- 1posed. It is of interest to us what features are utilized in those classifiers for their 2predictions. In this paper, we propose a novel approach to visualize important 3features used in classification decisions from point cloud ne... | B1lNI9_MiB | Official Blind Review #4 | 3: Weak Reject | This paper introduces an approach to compute the saliency map of a 3D point cloud, where the saliency is motivated from removing a region could result in a decrease in performance in shape classification. This idea was introduced in the image domain, and prior works leverage the so-called the natural neighborhood of im... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
VISUALIZING POINT CLOUD CLASSIFIERS BY MORPHING POINT CLOUDS INTO POTATOES
### Paper Abstract
Recently, various networks that operate directly on point clouds have been ... | |
BJeWVpNtwr | ICLR.cc/2020/Conference | 2020 | On the Pareto Efficiency of Quantized CNN | ["Ting-Wu Chin", "Pierce I-Jen Chuang", "Vikas Chandra", "Diana Marculescu"] | Weight Quantization for deep convolutional neural networks (CNNs) has shown promising results in compressing and accelerating CNN-powered applications such as semantic segmentation, gesture recognition, and scene understanding. Prior art has shown that different datasets, tasks, and network architectures admit differen... | ["convolutional neural networks quantization", "model compression", "efficient neural network"] | ABSTRACTWeight quantization for deep convolutional neural networks (CNNs) has shownpromising results in compressing and accelerating CNN-powered applicationssuch as semantic segmentation, gesture recognition, and scene understanding.Prior art has shown that different datasets, tasks, and network architectures ad-mit di... | S1xxtycQ5B | Official Blind Review #1 | 3: Weak Reject | This paper investigated the Pareto efficiency of quantized convolutional neural networks (CNNs).
The authors pointed out that when the number of channels of a network are changed in an iso-model size scenario, the lower precision value can Pareto dominate the higher ones for CNNs with standard convolutions. Furthermo... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
On the Pareto Efficiency of Quantized CNN
### Paper Abstract
Weight Quantization for deep convolutional neural networks (CNNs) has shown promising results in compressing... | |
RpprvYz0xTM | ICLR.cc/2021/Conference | 2021 | A Flexible Framework for Discovering Novel Categories with Contrastive Learning | ["Xuhui Jia", "Kai Han", "Yukun Zhu", "Bradley Green"] | This paper studies the problem of novel category discovery on single- and multi-modal data with labels from different but relevant categories. We present a generic, end-to-end framework to jointly learn a reliable representation and assign clusters to unlabelled data. To avoid over-fitting the learnt embedding to label... | ["deep learning", "novel classes", "clustering", "self-supervised learning", "unsupervised learning"] | ABSTRACTThis paper studies the problem of novel category discovery on single- and multi-modal data with labels from different but relevant categories. We present a generic,end-to-end framework to jointly learn a reliable representation and assign clustersto unlabelled data. To avoid over-fitting the learnt embedding to... | lgwIKzg-vqU | Review | 4: Ok but not good enough - rejection | [Summary]
This paper addresses the problem of clustering unseen classes with the help of labeled data from an exclusive set of classes. It proposes an end-to-end learning strategy that combines a set of optimization objectives for training a deep neural network that can output a cluster index for the unlabeled data. Th... | 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 Flexible Framework for Discovering Novel Categories with Contrastive Learning
### Paper Abstract
This paper studies the problem of novel category discovery on single- ... |
ryevtyHtPr | ICLR.cc/2020/Conference | 2020 | Do Deep Neural Networks for Segmentation Understand Insideness? | ["Kimberly M Villalobos", "Vilim Stih", "Amineh Ahmadinejad", "Jamell Dozier", "Andrew Francl", "Frederico Azevedo", "Tomotake Sasaki", "Xavier Boix"] | Image segmentation aims at grouping pixels that belong to the same object or region. At the heart of image segmentation lies the problem of determining whether a pixel is inside or outside a region, which we denote as the "insideness" problem. Many Deep Neural Networks (DNNs) variants excel in segmentation benchmarks, ... | ["Image Segmentation", "Deep Networks for Spatial Relationships", "Visual Routines", "Recurrent Neural Networks"] | ABSTRACTImage segmentation aims at grouping pixels that belong to the same object or re-gion. At the heart of image segmentation lies the problem of determining whethera pixel is inside or outside a region, which we denote as the “insideness” problem.Many Deep Neural Networks (DNNs) variants excel in segmentation bench... | HkgS6FcY9H | Official Blind Review #1 | 6: Weak Accept | The paper shows that deep-nets can actually learn to solve the problem of "what is inside a curve" by using a sort of progressive filling of the space outside the curve. The paper suceeds in explaining that and in pointing out the limitations if standard learning to address this problem.
However,
(1) the proposed dem... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Do Deep Neural Networks for Segmentation Understand Insideness?
### Paper Abstract
Image segmentation aims at grouping pixels that belong to the same object or region. A... | |
B1esygHFwS | ICLR.cc/2020/Conference | 2020 | Detecting Change in Seasonal Pattern via Autoencoder and Temporal Regularization | ["Raphael Fettaya", "Dor Bank", "Rachel Lemberg", "Linoy Barel"] | Change-point detection problem consists of discovering abrupt property changes in the generation process of time-series. Most state-of-the-art models are optimizing the power of a kernel two-sample test, with only a few assumptions on the distribution of the data. Unfortunately, because they presume the samples are dis... | ["Autoencoder", "Change Point Detection", "Timeseries"] | ABSTRACTChange-point detection problem consists of discovering abrupt property changesin the generation process of time-series. Most state-of-the-art models are opti-mizing the power of a kernel two-sample test, with only a few assumptions onthe distribution of the data. Unfortunately, because they presume the samples ... | B1lt_1QroB | Official Blind Review #3 | 1: Reject | I am quite disappointed with the presentation and technical quality of the paper.
There are numerous grammatical errors that make the reading unpleasant. The mathematical notations are also inconsistent throughout different places in the paper.
The extensive literature of modelling time series with seasonality trend... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Detecting Change in Seasonal Pattern via Autoencoder and Temporal Regularization
### Paper Abstract
Change-point detection problem consists of discovering abrupt propert... | |
S1g2V3Cct7 | ICLR.cc/2019/Conference | 2019 | Experience replay for continual learning | ["David Rolnick", "Arun Ahuja", "Jonathan Schwarz", "Timothy P. Lillicrap", "Greg Wayne"] | Continual learning is the problem of learning new tasks or knowledge while protecting old knowledge and ideally generalizing from old experience to learn new tasks faster. Neural networks trained by stochastic gradient descent often degrade on old tasks when trained successively on new tasks with different data distrib... | ["continual learning", "catastrophic forgetting", "lifelong learning", "behavioral cloning", "reinforcement learning", "interference", "stability-plasticity"] | ABSTRACTContinual learning is the problem of learning new tasks or knowledge while pro-tecting old knowledge and ideally generalizing from old experience to learn newtasks faster. Neural networks trained by stochastic gradient descent often degradeon old tasks when trained successively on new tasks with different data ... | Hklq4yMYn7 | Reduction in Catastrophic Forgetting by Augmented Experience Replay | 5: Marginally below acceptance threshold | The authors propose an approach to augment experience replay buffers with properties that can alleviate issues with catastrophic forgetting. The buffers are augmented by storing both new and historical experiences, along with the desired historical policy & value distribution. The AC learning now couples two additional... | 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
Experience replay for continual learning
### Paper Abstract
Continual learning is the problem of learning new tasks or knowledge while protecting old knowledge and ideal... |
Hke4_JrYDr | ICLR.cc/2020/Conference | 2020 | Global-Local Network for Learning Depth with Very Sparse Supervision | ["Antonio Loquercio", "Alexey Dosovitskiy", "Davide Scaramuzza"] | Natural intelligent agents learn to perceive the three dimensional structure of the world without training on large datasets and are unlikely to have the precise equations of projective geometry hard-wired in the brain. Such skill would also be valuable to artificial systems in order to avoid the expensive c... | ["Depth Perception", "Learning from Sparse Supervision", "Learning from Interaction."] | ABSTRACTNatural intelligent agents learn to perceive the three dimensional structure ofthe world without training on large datasets and are unlikely to have the preciseequations of projective geometry hard-wired in the brain. Such skill would alsobe valuable to artificial systems in order to avoid the expensive collect... | H1l1jzoOtB | Official Blind Review #1 | 1: Reject | This paper presents a method for learning to predict a depth map given a pair of images. Training is done is an unsupervised way except for a small set of image locations for which the absolute 3D locations are known. The optical flow between the 2 images is computed automatically. A network is trained to predict the c... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Global-Local Network for Learning Depth with Very Sparse Supervision
### Paper Abstract
Natural intelligent agents learn to perceive the three dimensional structure of t... | |
B1l6nnEtwr | ICLR.cc/2020/Conference | 2020 | AN EFFICIENT HOMOTOPY TRAINING ALGORITHM FOR NEURAL NETWORKS | ["Qipin Chen", "Wenrui Hao"] | We present a Homotopy Training Algorithm (HTA) to solve optimization problems arising from neural networks. The HTA starts with several decoupled systems with low dimensional structure and tracks the solution to the high dimensional coupled system. The decoupled systems are easy to solve due to the low dimensionality b... | ["Homotopy training algorithm", "Convergence analysis", "Neural networks"] | ABSTRACTWe present a Homotopy Training Algorithm (HTA) to solve optimization prob-lems arising from neural networks. The HTA starts with several decoupled sys-tems with low dimensional structure and tracks the solution to the high dimen-sional coupled system. The decoupled systems are easy to solve due to the lowdimens... | Skla9ft35S | Official Blind Review #2 | 1: Reject | Summary
This paper proposes an algorithm to address the issue of nonlinear optimization
in high dimensions and applies it to convolutional neural networks (VGG models) on CIFAR 10.
They show 11% relative reduction in error for this particular task with this
particular network. In addition, they prove additional theor... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
AN EFFICIENT HOMOTOPY TRAINING ALGORITHM FOR NEURAL NETWORKS
### Paper Abstract
We present a Homotopy Training Algorithm (HTA) to solve optimization problems arising fro... | |
BJl-5pNKDB | ICLR.cc/2020/Conference | 2020 | On Computation and Generalization of Generative Adversarial Imitation Learning | ["Minshuo Chen", "Yizhou Wang", "Tianyi Liu", "Zhuoran Yang", "Xingguo Li", "Zhaoran Wang", "Tuo Zhao"] | Generative Adversarial Imitation Learning (GAIL) is a powerful and practical approach for learning sequential decision-making policies. Different from Reinforcement Learning (RL), GAIL takes advantage of demonstration data by experts (e.g., human), and learns both the policy and reward function of the unknown environme... | ["gail", "generalization", "computation", "generative adversarial imitation", "demonstration data", "theory", "powerful", "practical", "sequential"] | ABSTRACTGenerative Adversarial Imitation Learning (GAIL) is a powerful and practical ap-proach for learning sequential decision-making policies. Different from Rein-forcement Learning (RL), GAIL takes advantage of demonstration data by experts(e.g., human), and learns both the policy and reward function of the unknown ... | SklghBKaYS | Official Blind Review #2 | 6: Weak Accept | Summary: This paper presents theoretical analysis of the generalization properties of GAIL, as well as the local convergence of the traditional minibatch SGD applied to its min-max optimization problem (without assuming convex-concave structure of the game). Specifically, the authors first prove a generalization bound... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
On Computation and Generalization of Generative Adversarial Imitation Learning
### Paper Abstract
Generative Adversarial Imitation Learning (GAIL) is a powerful and prac... | |
rk6cfpRjZ | ICLR.cc/2018/Conference | 2018 | Learning Intrinsic Sparse Structures within Long Short-Term Memory | ["Wei Wen", "Yuxiong He", "Samyam Rajbhandari", "Minjia Zhang", "Wenhan Wang", "Fang Liu", "Bin Hu", "Yiran Chen", "Hai Li"] | Model compression is significant for the wide adoption of Recurrent Neural Networks (RNNs) in both user devices possessing limited resources and business clusters requiring quick responses to large-scale service requests. This work aims to learn structurally-sparse Long Short-Term Memory (LSTM) by reducing the sizes of... | ["Sparsity", "Model Compression", "Acceleration", "LSTMs", "Recurrent Neural Networks", "Structural Learning"] | ABSTRACTModel compression is significant for the wide adoption of Recurrent Neural Net-works (RNNs) in both user devices possessing limited resources and business clus-ters requiring quick responses to large-scale service requests. This work aims tolearn structurally-sparse Long Short-Term Memory (LSTM) by reducing the... | B15mUMdef | the text is verbose but method is simple | 6: Marginally above acceptance threshold | The paper spends lots of (repeated) texts on motivating and explaining ISS. But the algorithm is simple, using group lasso to find components that are can retained to preserve the performance. Thus the novelty is limited.
The experiments results are good.
Sec 3.1 should be made more concise. | 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 Intrinsic Sparse Structures within Long Short-Term Memory
### Paper Abstract
Model compression is significant for the wide adoption of Recurrent Neural Networks... |
nRJ08rN_b17 | ICLR.cc/2021/Conference | 2021 | Vision at A Glance: Interplay between Fine and Coarse Information Processing Pathways | ["Zilong Ji", "Xiaolong Zou", "Tiejun Huang", "Si Wu"] | Object recognition is often viewed as a feedforward, bottom-up process in machine learning, but in real neural systems, object recognition is a complicated process which involves the interplay between two signal pathways. One is the parvocellular pathway (P-pathway), which is slow and extracts fine features of objects;... | ["Fast pathway", "Slow pathway", "Interplay", "Robustness", "Visual backward masking", "Biological visual systems", "Biological inspried model"] | ABSTRACTObject recognition is often viewed as a feedforward, bottom-up process in ma-chine learning, but in real neural systems, object recognition is a complicatedprocess which involves the interplay between two signal pathways. One is theparvocellular pathway (P-pathway), which is slow and extracts fine features ofob... | lO552gAVDrY | Results are not strong, and descriptions are not clear enough | 3: Clear rejection | This paper proposed a two-pathway neural network to mimic the interplay between the parvocellular (slow and fine-grained) and magnocellular (fast and course) pathways in neural systems. The two pathways are named as FineNet and CourseNet. During inference, the FineNet received recurrent feedback signals from the Coarse... | 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
Vision at A Glance: Interplay between Fine and Coarse Information Processing Pathways
### Paper Abstract
Object recognition is often viewed as a feedforward, bottom-up p... |
ByeadyrtPB | ICLR.cc/2020/Conference | 2020 | Learning Deep-Latent Hierarchies by Stacking Wasserstein Autoencoders | ["Benoit Gaujac", "Ilya Feige", "David Barber"] | Probabilistic models with hierarchical-latent-variable structures provide state-of-the-art results amongst non-autoregressive, unsupervised density-based models. However, the most common approach to training such models based on Variational Autoencoders often fails to leverage deep-latent hierarchies; successful approa... | ["Generative modelling", "Optimal Transport"] | ABSTRACTProbabilistic models with hierarchical-latent-variable structures provide state-of-the-art results amongst non-autoregressive, unsupervised density-based models.However, the most common approach to training such models based on VariationalAutoencoders often fails to leverage deep-latent hierarchies; successful ... | rJgzs5lwtr | Official Blind Review #1 | 1: Reject | The paper aims to develop a deep generative model, which -unlike VAEs or GANs- comprises a hierarchy of latent variables rather than a direct map from the stochastic latent manifold to the observation space. To this end, the paper builds a training objective based on nesting the Wasserstein distance between the data di... | <|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-Latent Hierarchies by Stacking Wasserstein Autoencoders
### Paper Abstract
Probabilistic models with hierarchical-latent-variable structures provide state-... | |
Ib-8gIymC1 | NeurIPS.cc/2022/Workshop/SyntheticData4ML | 2022 | Mitigating Health Data Poverty: Generative Approaches versus Resampling for Time-series Clinical Data | ["Raffaele Marchesi", "Nicolo Micheletti", "Giuseppe Jurman", "Venet Osmani"] | Several approaches have been developed to mitigate algorithmic bias stemming from health data poverty, where minority groups are underrepresented in training datasets. Augmenting the minority class using resampling (such as SMOTE) is a widely used approach due to the simplicity of the algorithms. However, these algorit... | ["health data poverty", "generative approaches", "minority class", "smote", "approaches", "authentic data", "real data", "clinical data", "algorithmic bias"] | Mitigating Health Data Poverty: GenerativeApproaches versus Resampling for Time-seriesClinical DataRaffaele Marchesi†,1,2Nicolo Micheletti†,1,3Giuseppe Jurman1Venet Osmani1raffaele.marchesi@studenti.unitn.itnicolo.micheletti@student.manchester.ac.ukjurman@fbk.eu ,vosmani@fbk.eu†Equal contribution1Fondazione Bruno Kessl... | d4S7JzsRHW | The paper presents a methodology for addressing, class imbalance and data augmentation, very relevant problem in healthcare datasets. The work is easy to follow, yet the evaluation could be improved with slight changes. | 9: Top 15% of accepted papers, strong accept | Pros:
1) The authors present a simple yet valuable modification to WGAN-GP*, specifically trained for data augmentation purposes.
2) The work is clearly explained and easy to follow and compares with SMOTE, a popular method used in clinical and epidemiological studies using datasets similar to MIMIC.
3) The authors 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
Mitigating Health Data Poverty: Generative Approaches versus Resampling for Time-series Clinical Data
### Paper Abstract
Several approaches have been developed to mitiga... |
B1gJOoRcYQ | ICLR.cc/2019/Conference | 2019 | S3TA: A Soft, Spatial, Sequential, Top-Down Attention Model | ["Alex Mott", "Daniel Zoran", "Mike Chrzanowski", "Daan Wierstra", "Danilo J. Rezende"] | We present a soft, spatial, sequential, top-down attention model (S3TA). This model uses a soft attention mechanism to bottleneck its view of the input. A recurrent core is used to generate query vectors, which actively select information from the input by correlating the query with input- and space-dependent key maps ... | ["Attention", "RL", "Top-Down", "Interpretability"] | ABSTRACTWe present a soft, spatial, sequential, top-down attention model (S3TA). Thismodel uses a soft attention mechanism to bottleneck its view of the input. Arecurrent core is used to generate query vectors, which actively select informationfrom the input by correlating the query with input- and space-dependent key ... | BJlkSl8ahm | Interesting model, however, the performance on the supervised task is not good. | 5: Marginally below acceptance threshold |
[Summary]
This paper proposed a soft, spatial, sequential, top-down attention model which enable the agent and classifier actively select important, task relative information to generate the appropriate output. Given the observations, the proposed method uses a convLSTM to produce the key and value tensor. 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
S3TA: A Soft, Spatial, Sequential, Top-Down Attention Model
### Paper Abstract
We present a soft, spatial, sequential, top-down attention model (S3TA). This model uses a... |
r1x_DaVKwH | ICLR.cc/2020/Conference | 2020 | Is Deep Reinforcement Learning Really Superhuman on Atari? Leveling the playing field | ["Marin Toromanoff", "Emilie Wirbel", "Fabien Moutarde"] | Consistent and reproducible evaluation of Deep Reinforcement Learning (DRL) is not straightforward. In the Arcade Learning Environment (ALE), small changes in environment parameters such as stochasticity or the maximum allowed play time can lead to very different performance. In this work, we discuss the difficulties o... | ["Reinforcement Learning", "Deep Learning", "Atari benchmark", "Reproducibility"] | ABSTRACTConsistent and reproducible evaluation of Deep Reinforcement Learning (DRL) isnot straightforward. In the Arcade Learning Environment (ALE), small changes inenvironment parameters such as stochasticity or the maximum allowed play timecan lead to very different performance. In this work, we discuss the difficult... | HJxg5_Satr | Official Blind Review #3 | 3: Weak Reject | The paper proposes an extension to the work of Machado et al. (2018) for standardizing training and evaluation procedures in the Arcade Learning Environment (ALE). It then introduces a collection of human world records for each Atari game to refute previous claims of superhuman performance, as well as recommend compari... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Is Deep Reinforcement Learning Really Superhuman on Atari? Leveling the playing field
### Paper Abstract
Consistent and reproducible evaluation of Deep Reinforcement Lea... | |
SygkSkSFDB | ICLR.cc/2020/Conference | 2020 | On the expected running time of nonconvex optimization with early stopping | ["Thomas Flynn", "Kwang Min Yu", "Abid Malik", "Shinjae Yoo", "Nicholas D'Imperio"] | This work examines the convergence of stochastic gradient algorithms that use early stopping based on a validation function, wherein optimization ends when the magnitude of a validation function gradient drops below a threshold. We derive conditions that guarantee this stopping rule is well-defined and analyze the expe... | ["non-convex", "stopping times", "statistics", "gradient descent", "early stopping"] | ABSTRACTThis work examines the convergence of stochastic gradient algorithms that useearly stopping based on a validation function, wherein optimization ends whenthe magnitude of a validation function gradient drops below a threshold. Wederive conditions that guarantee this stopping rule is well-defined and analyze the... | B1ghnfO49B | Official Blind Review #3 | 3: Weak Reject | This paper proposes an optimization approach in which the optimizer computes the gradient on a given function yet uses another to decide a stopping time. Conceptually those functions are empirical errors on train and validation folds in the most common setting, although the authors seem to use other settings later in 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
On the expected running time of nonconvex optimization with early stopping
### Paper Abstract
This work examines the convergence of stochastic gradient algorithms that u... | |
Sye7qoC5FQ | ICLR.cc/2019/Conference | 2019 | Adversarial Attacks on Node Embeddings | ["Aleksandar Bojchevski", "Stephan G\u00fcnnemann"] | The goal of network representation learning is to learn low-dimensional node embeddings that capture the graph structure and are useful for solving downstream tasks. However, despite the proliferation of such methods there is currently no study of their robustness to adversarial attacks. We provide the first adversaria... | ["node embeddings", "adversarial attacks"] | ABSTRACTThe goal of network representation learning is to learn low-dimensional node em-beddings that capture the graph structure and are useful for solving downstreamtasks. However, despite the proliferation of such methods, there is currently nostudy of their robustness to adversarial attacks. We provide the first ad... | SJxrTF8Y3X | a novel adversarial attack on node embedding method based on random walks | 6: Marginally above acceptance threshold | Brief Summary:
The authors present a novel adversarial attack on node embedding method based on random walks. They focus on perturbing the structure of the network. Because the bi-level optimization problem can be highly challenging, they refer to factorize a random walk matrix which is proved equivalent to DeepWalk. 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
Adversarial Attacks on Node Embeddings
### Paper Abstract
The goal of network representation learning is to learn low-dimensional node embeddings that capture the graph ... |
I6NRcao1w-X | ICLR.cc/2021/Conference | 2021 | Robust Reinforcement Learning using Adversarial Populations | ["Eugene Vinitsky", "Yuqing du", "Kanaad V Parvate", "Kathy Jang", "Pieter Abbeel", "Alexandre Bayen"] | Reinforcement Learning (RL) is an effective tool for controller design but can struggle with issues of robustness, failing catastrophically when the underlying system dynamics are perturbed. The Robust RL formulation tackles this by adding worst-case adversarial noise to the dynamics and constructing the noise distribu... | ["Robust Control", "Reinforcement Learning", "Multiagent Systems"] | ABSTRACTReinforcement Learning (RL) is an effective tool for controller design but canstruggle with issues of robustness, failing catastrophically when the underlyingsystem dynamics are perturbed. The Robust RL formulation tackles this by addingworst-case adversarial noise to the dynamics and constructing the noise dis... | 13vwhyz4HtM | A clean formulation with some compelling results | 7: Good paper, accept | #### Summary
The authors present a scheme that can be used to train agents to be robust against a population of adversarial policies, in which adversaries can perturb actions via an additive perturbation. Motivated by the observation that agents trained against a single policy may overfit to that policy and hence wil... | 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
Robust Reinforcement Learning using Adversarial Populations
### Paper Abstract
Reinforcement Learning (RL) is an effective tool for controller design but can struggle wi... |
Bya8fGWAZ | ICLR.cc/2018/Conference | 2018 | Value Propagation Networks | ["Nantas Nardelli", "Gabriel Synnaeve", "Zeming Lin", "Pushmeet Kohli", "Nicolas Usunier"] | We present Value Propagation (VProp), a parameter-efficient differentiable planning module built on Value Iteration which can successfully be trained in a reinforcement learning fashion to solve unseen tasks, has the capability to generalize to larger map sizes, and can learn to navigate in dynamic environments. We eva... | ["Learning to plan", "Reinforcement Learning", "Value Iteration", "Navigation", "Convnets"] | ABSTRACTWe present Value Propagation (VProp), a parameter-efficient differentiable plan-ning module built on Value Iteration which can successfully be trained in a rein-forcement learning fashion to solve unseen tasks, has the capability to generalize tolarger map sizes, and can learn to navigate in dynamic environment... | rJRfJZKxf | An extension of Value Iteration Network; the writing needs to be greatly improved | 5: Marginally below acceptance threshold | The paper introduces two alternatives to value iteration network (VIN) proposed by Tamar et al. VIN was proposed to tackle the task of learning to plan using as inputs a position and an image of the map of the environment. The authors propose two new updates value propagation (VProp) and max propagation (MVProp), which... | 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
Value Propagation Networks
### Paper Abstract
We present Value Propagation (VProp), a parameter-efficient differentiable planning module built on Value Iteration which c... |
SJ6yPD5xg | ICLR.cc/2017/conference | 2017 | Reinforcement Learning with Unsupervised Auxiliary Tasks | ["Max Jaderberg", "Volodymyr Mnih", "Wojciech Marian Czarnecki", "Tom Schaul", "Joel Z Leibo", "David Silver", "Koray Kavukcuoglu"] | Deep reinforcement learning agents have achieved state-of-the-art results by directly maximising cumulative reward. However, environments contain a much wider variety of possible training signals. In this paper, we introduce an agent that also maximises many other pseudo-reward functions simultaneously by reinforcement... | ["agent", "extrinsic rewards", "expert human performance", "labyrinth", "reinforcement", "agents", "results", "cumulative reward"] | ABSTRACTDeep reinforcement learning agents have achieved state-of-the-art results by di-rectly maximising cumulative reward. However, environments contain a muchwider variety of possible training signals. In this paper, we introduce an agent thatalso learns separate policies for maximising many other pseudo-reward func... | ryDtZqDVl | Review | 7: Good paper, accept | This paper proposes a way of adding unsupervised auxiliary tasks to a deep RL agent like A3C. Authors propose a bunch of auxiliary control tasks and auxiliary reward tasks and evaluate the agent in Labyrinth and Atari. Proposed UNREAL agent performs significantly better than A3C and also learns faster. This is definite... | 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
Reinforcement Learning with Unsupervised Auxiliary Tasks
### Paper Abstract
Deep reinforcement learning agents have achieved state-of-the-art results by directly maximis... |
yFPqbprG2Qb | ecmlpkdd.org/ECMLPKDD/2021/Workshop/TeachML | 2021 | Deeper Learning By Doing: Integrating Hands-On Research Projects Into A Machine Learning Course | ["Sebastian Raschka"] | Machine learning has seen a vast increase of interest in recent years, along with an abundance of learning resources. While conventional lectures provide students with important information and knowledge, we also believe that additional project-based learning components can motivate students to engage in topics more de... | ["machine learning", "deep learning", "project-based learning"] | Deeper Learning By Doing: Integrating Hands-On ResearchProjects Into A Machine Learning CourseSebastian Raschka1AbstractMachine learning has seen a vast increase of in-terest in recent years, along with an abundance oflearning resources. While conventional lecturesprovide students with important information andknowledg... | eXe7buGvRB | Very clear method report, perhaps too much publication-focused | 8: Top 50% of accepted papers, clear accept | The paper is very clearly written and follows the submission guidelines exactly.
The authors report on "project-based learning" used in a deep learning course. The authors make use of the known fact that applying knowledge is a more effective learning strategy than only absorbing it. In their approach, the course is s... | 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
Deeper Learning By Doing: Integrating Hands-On Research Projects Into A Machine Learning Course
### Paper Abstract
Machine learning has seen a vast increase of interest ... |
rkg6FgrtPB | ICLR.cc/2020/Conference | 2020 | Biologically Plausible Neural Networks via Evolutionary Dynamics and Dopaminergic Plasticity | ["Sruthi Gorantla", "Anand Louis", "Christos H. Papadimitriou", "Santosh Vempala", "Naganand Yadati"] | Artificial neural networks (ANNs) lack in biological plausibility, chiefly because backpropagation requires a variant of plasticity (precise changes of the synaptic weights informed by neural events that occur downstream in the neural circuit) that is profoundly incompatible with the current understanding of the animal... | ["Biological plausibility", "dopaminergic plasticity", "allele frequency", "neural net evolution"] | ABSTRACTArtificial neural networks (ANNs) lack in biological plausibility, chiefly becausebackpropagation requires a variant of plasticity (precise changes of the synap-tic weights informed by neural events that occur downstream in the neural circuit)that is profoundly incompatible with the current understanding of the... | S1eZ0ExCuH | Official Blind Review #2 | 1: Reject | First of all, I must confess that my knowledge is quite limited to read this paper. Perhap the authors present something that I can not catch up at the present.
I conjecture the paper would like to bring the evolution in genetics and perhap brain cirecuits as well to define a novel neural net model, called NNE by the ... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Biologically Plausible Neural Networks via Evolutionary Dynamics and Dopaminergic Plasticity
### Paper Abstract
Artificial neural networks (ANNs) lack in biological plau... | |
ZSLXyrpQHwl | ICML.cc/2020/Workshop/Artemiss | 2020 | Predicting Feature Imputability in the Absence of Ground Truth | ["Niamh McCombe", "Xuemei Ding", "Girijesh Prasad", "David P Finn", "Stephen Todd", "Paula L McClean", "Kongfatt Wong-Lin"] | Data imputation is the most popular method of dealing with missing values, but in most real life applications, large missing data can occur and it is difficult or impossible to evaluate whether data has been imputed accurately (lack of ground truth). This paper addresses these issues by proposing an effective and simpl... | ["Missing Data", "Data Imputation", "Feature Imputability", "Machine Learning", "PCA", "NIPALS"] | Predicting Feature Imputability in the Absence of Ground TruthNiamh McCombe1Xuemei Ding1Girijesh Prasad1David P. Finn2Stephen Todd3Paula L. McClean4KongFatt Wong-Lin1AbstractData imputation is the most popular method ofdealing with missing values, but in most real lifeapplications, large missing data can occur andit is... | hZSotw2N92w | Nice idea; Methods section could be clearer; What actions can be taken with the output? | 6: Marginally above acceptance threshold | Summary: The authors propose a method to tackle the problem of feature imputability where the goal is to understand which features are most accurately imputable based on all other features in the data. This problem is interesting and aligns with the goals of this workshop. They show that using a PCA based imputation st... | 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
Predicting Feature Imputability in the Absence of Ground Truth
### Paper Abstract
Data imputation is the most popular method of dealing with missing values, but in most ... |
J4XaMT9OcZ | ICLR.cc/2021/Conference | 2021 | Mitigating Deep Double Descent by Concatenating Inputs | ["John Chen", "Qihan Wang", "Anastasios Kyrillidis"] | The double descent curve is one of the most intriguing properties of deep neural networks. It contrasts the classical bias-variance curve with the behavior of modern neural networks, occurring where the number of samples nears the number of parameters. In this work, we explore the connection between the double descent ... | ["deep double descent", "feedforward neural network", "image classificaiton"] | ABSTRACTThe double descent curve is one of the most intriguing properties of deep neuralnetworks. It contrasts the classical bias-variance curve with the behavior of mod-ern neural networks, occurring where the number of samples nears the numberof parameters. In this work, we explore the connection between the double d... | lzsT0KJj67 | The Concatenated input construction is cool but the paper falls short from giving an interesting explanation of the results. | 5: Marginally below acceptance threshold | Summary:
The paper empirically studies the Double Descent phenomenon by using a cool construction that squares the dataset size by concatenation of every pair of inputs and linear interpolation of the labels. The main empirical finding is that this construction mitigates double descent.
++++++
Main Comments:
The con... | 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
Mitigating Deep Double Descent by Concatenating Inputs
### Paper Abstract
The double descent curve is one of the most intriguing properties of deep neural networks. It c... |
HJgK0h4Ywr | ICLR.cc/2020/Conference | 2020 | Theory and Evaluation Metrics for Learning Disentangled Representations | ["Kien Do", "Truyen Tran"] | We make two theoretical contributions to disentanglement learning by (a) defining precise semantics of disentangled representations, and (b) establishing robust metrics for evaluation. First, we characterize the concept “disentangled representations” used in supervised and unsupervised methods along three dimensions–in... | ["disentanglement", "metrics"] | ABSTRACTWe make two theoretical contributions to disentanglement learning by (a) defin-ing precise semantics of disentangled representations, and (b) establishing robustmetrics for evaluation. First, we characterize the concept “disentangled represen-tations” used in supervised and unsupervised methods along three dime... | ryevsg7M5H | Official Blind Review #1 | 6: Weak Accept | Summary:
The paper presents a new set of metrics for evaluating disentangled representations in both supervised and unsupervised settings. Disentangled representations are evaluated along three dimensions: informativeness, separability, and interpretability. While previous work offers metrics for similar dimensions (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
Theory and Evaluation Metrics for Learning Disentangled Representations
### Paper Abstract
We make two theoretical contributions to disentanglement learning by (a) defin... | |
Sk4jFoA9K7 | ICLR.cc/2019/Conference | 2019 | PeerNets: Exploiting Peer Wisdom Against Adversarial Attacks | ["Jan Svoboda", "Jonathan Masci", "Federico Monti", "Michael Bronstein", "Leonidas Guibas"] | Deep learning systems have become ubiquitous in many aspects of our lives. Unfortunately, it has been shown that such systems are vulnerable to adversarial attacks, making them prone to potential unlawful uses.
Designing deep neural networks that are robust to adversarial attacks is a fundamental step in making such s... | ["peernet", "peernets", "graph", "geometric deep learning", "adversarial", "perturbation", "defense", "peer regularization"] | ABSTRACTDeep learning systems have become ubiquitous in many aspects of our lives. Un-fortunately, it has been shown that such systems are vulnerable to adversarialattacks, making them prone to potential unlawful uses. Designing deep neuralnetworks that are robust to adversarial attacks is a fundamental step in making ... | rJxejaXmhQ | Analysis and experimental comparisons are lacking | 6: Marginally above acceptance threshold | After reading the authors' response, I'm revising my score upwards from 5 to 6.
The authors propose a defense against adversarial examples, that is inspired by "non local means filtering". The underlying assumption seems to be that, at feature level, adversarial examples manifest as IID noise in feature maps, which ca... | 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
PeerNets: Exploiting Peer Wisdom Against Adversarial Attacks
### Paper Abstract
Deep learning systems have become ubiquitous in many aspects of our lives. Unfortunately,... |
r1lfF2NYvH | ICLR.cc/2020/Conference | 2020 | InfoGraph: Unsupervised and Semi-supervised Graph-Level Representation Learning via Mutual Information Maximization | ["Fan-Yun Sun", "Jordan Hoffman", "Vikas Verma", "Jian Tang"] | This paper studies learning the representations of whole graphs in both unsupervised and semi-supervised scenarios. Graph-level representations are critical in a variety of real-world applications such as predicting the properties of molecules and community analysis in social networks. Traditional graph kernel based me... | ["graph-level representation learning", "mutual information maximization"] | ABSTRACTThis paper studies learning the representations of whole graphs in both unsuper-vised and semi-supervised scenarios. Graph-level representations are critical in avariety of real-world applications such as predicting the properties of molecules andcommunity analysis in social networks. Traditional graph kernel b... | BJlXq_au9H | Official Blind Review #3 | 6: Weak Accept | The paper presents a new graph representation learning method for the whole graph under both unsupervised and semi-supervised setting. Different from existing ones using graph kernel, or graph2vec, the proposed InfoGraph is able to extract graph-level representation with fixed-length features that are generalized well.... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
InfoGraph: Unsupervised and Semi-supervised Graph-Level Representation Learning via Mutual Information Maximization
### Paper Abstract
This paper studies learning the re... | |
BkXMikqxx | ICLR.cc/2017/conference | 2017 | Cortical-Inspired Open-Bigram Representation for Handwritten Word Recognition | ["Th\u00e9odore Bluche", "Christopher Kermorvant", "Claude Touzet", "Herv\u00e9 Glotin"] | Recent research in the cognitive process of reading hypothesized that we do
not read words by sequentially recognizing letters, but rather by identifing
open-bigrams, i.e. couple of letters that are not necessarily next
to each other.
In this paper, we evaluate an handwritten word recognition method based on original
... | ["representation", "letters", "handwritten word recognition", "cognitive process", "words", "couple", "next", "original", "long"] | ABSTRACTRecent research in the cognitive process of reading hypothesized that we do notread words by sequentially recognizing letters, but rather by identifing open-bigrams, i.e.couple of letters that are not necessarily next to each other. In thispaper, we evaluate an handwritten word recognition method based on origi... | BJk3aarEg | Review | 4: Ok but not good enough - rejection | This paper uses an LSTM model to predict what it calls "open bigrams" (bigrams of characters that may or may not have letters inbetween) from handwriting data. These open bigrams are subsequently used to predict the written word in a decoding step. The experiments indicate that the system does slightly better than a ba... | 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
Cortical-Inspired Open-Bigram Representation for Handwritten Word Recognition
### Paper Abstract
Recent research in the cognitive process of reading hypothesized that we... |
B1eWOJHKvB | ICLR.cc/2020/Conference | 2020 | Kernel of CycleGAN as a principal homogeneous space | ["Nikita Moriakov", "Jonas Adler", "Jonas Teuwen"] | Unpaired image-to-image translation has attracted significant interest due to the invention of CycleGAN, a method which utilizes a combination of adversarial and cycle consistency losses to avoid the need for paired data. It is known that the CycleGAN problem might admit multiple solutions, and our goal in this paper i... | ["Generative models", "CycleGAN"] | ABSTRACTUnpaired image-to-image translation has attracted significant interest due to theinvention of CycleGAN, a method which utilizes a combination of adversarial andcycle consistency losses to avoid the need for paired data. It is known that theCycleGAN problem might admit multiple solutions, and our goal in this pa... | rylknOkRFr | Official Blind Review #1 | 6: Weak Accept | This paper focuses on CycleGAN method to show theoretically when the exact solution space is invariant with respect to authomorphisms of the underlying probability spaces for unpaired image-to-image translation.
- The paper provides interesting theoretical results on identifying conditions under which CycleGAN admits... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Kernel of CycleGAN as a principal homogeneous space
### Paper Abstract
Unpaired image-to-image translation has attracted significant interest due to the invention of Cyc... | |
JENSKEEzsoU | aclweb.org/ACL/2020/Workshop/NLP-COVID | 2020 | COVID-QA: A Question Answering Dataset for COVID-19 | ["Timo M\u00f6ller", "Anthony Reina", "Raghavan Jayakumar", "Malte Pietsch"] | We present COVID-QA, a Question Answering dataset consisting of 2,019 question/answer pairs annotated by volunteer biomedical experts on scientific articles related to COVID-19. To evaluate the dataset we compared a RoBERTa base model fine-tuned on SQuAD with the same model trained on SQuAD and our COVID-QA dataset. We... | ["NLP", "Question Answering", "Dataset", "COVID-19", "CORD-19"] | COVID-QA: A Question Answering Dataset for COVID-19Timo M ̈ollerdeepset GmbHBerlin, GermanyG Anthony ReinaIntel CorporationSanta Clara, CA, USARaghavan JayakumarLawrence LivermoreNational Laboratory(retired)Livermore, CA, USAMalte Pietschdeepset GmbHBerlin, GermanyAbstractWe present COVID-QA, a Question An-swering data... | sCoDR2-fOYk | Development of a QA dataset for COVID-19; a work in progress that requires revision | 5: Marginally below acceptance threshold | Authors have developed a QA dataset based on scientific articles about coronaviruses released as part of the COVID-19 Open Research Dataset (CORD-19). A small set of 450 articles were selected based on keyword matching, random sampling, and expert search, and around 2K Q-A pairs were developed against these articles, w... | 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
COVID-QA: A Question Answering Dataset for COVID-19
### Paper Abstract
We present COVID-QA, a Question Answering dataset consisting of 2,019 question/answer pairs annota... |
rI3RMgDkZqJ | ICLR.cc/2021/Conference | 2021 | A Primal Approach to Constrained Policy Optimization: Global Optimality and Finite-Time Analysis | ["Tengyu Xu", "Yingbin Liang", "Guanghui Lan"] | Safe reinforcement learning (SRL) problems are typically modeled as constrained Markov Decision Process (CMDP), in which an agent explores the environment to maximize the expected total reward and meanwhile avoids violating certain constraints on a number of expected total costs. In general, such SRL problems have nonc... | ["safe reinforcement learning", "constrained markov decision process", "policy optimization"] | ABSTRACTSafe reinforcement learning (SRL) problems are typically modeled as constrainedMarkov Decision Process (CMDP), in which an agent explores the environmentto maximize the expected total reward and meanwhile avoids violating certainconstraints on a number of expected total costs. In general, such SRL problemshave ... | Wg-sT2wlEAG | Comments | 5: Marginally below acceptance threshold | This paper proposes an algorithm called CRPO to solve the constraint reinforcement learning. Different from the literature, where the primal-dual formulations are widely used, this paper introduces a pure primal algorithm. The motivation of this paper is clear, since in practice the primal-dual formulation is hard to t... | 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 Primal Approach to Constrained Policy Optimization: Global Optimality and Finite-Time Analysis
### Paper Abstract
Safe reinforcement learning (SRL) problems are typica... |
ByxCrerKvS | ICLR.cc/2020/Conference | 2020 | Set Functions for Time Series | ["Max Horn", "Michael Moor", "Christian Bock", "Bastian Rieck", "Karsten Borgwardt"] | Despite the eminent successes of deep neural networks, many architectures are often hard to transfer to irregularly-sampled and asynchronous time series that occur in many real-world datasets, such as healthcare applications. This paper proposes a novel framework for classifying irregularly sampled time series with una... | ["Time Series", "Set functions", "Irregularly sampling", "Medical Time series", "Dynamical Systems", "Time series classification"] | ABSTRACTDespite the eminent successes of deep neural networks, many architectures areoften hard to transfer to irregularly-sampled and asynchronous time series thatoccur in many real-world datasets, such as healthcare applications. This paperproposes a novel framework for classifying irregularly sampled time series wit... | BJeBhVSAYS | Official Blind Review #3 | 6: Weak Accept | Summary:
The work is focused on classification of irregularly sampled and unaligned multi-modal time series. Prior work has primarily focused on imputation methods, either end-to-end or otherwise. This paper approaches the problem as a set function mapping between the time-series tuples to the class label. The proposed... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Set Functions for Time Series
### Paper Abstract
Despite the eminent successes of deep neural networks, many architectures are often hard to transfer to irregularly-samp... | |
fgpXAu8puGj | ICLR.cc/2021/Conference | 2021 | NAHAS: Neural Architecture and Hardware Accelerator Search | ["Yanqi Zhou", "Xuanyi Dong", "Daiyi Peng", "Ethan Zhu", "Amir Yazdanbakhsh", "Berkin Akin", "Mingxing Tan", "James Laudon"] | Neural architectures and hardware accelerators have been two driving forces for the rapid progress in deep learning.
Although previous works have optimized either neural architectures given fixed hardware, or hardware given fixed neural architectures, none has considered optimizing them jointly. In this paper, we study... | ["neural architecture search", "systems", "hardware"] | ABSTRACTNeural architectures and hardware accelerators have been two driving forces forthe rapid progress in deep learning. Although previous works have optimizedeither neural architectures given fixed hardware, or hardware given fixed neuralarchitectures, none has considered optimizing them jointly. In this paper, we ... | zcIcN0z4HuE | Good exploration with weak technical contributions. | 5: Marginally below acceptance threshold | ##########################################################################
Summary:
The paper proposes an algorithm to search for hardware designs and neural architectures jointly.
The results show that joint optimization improves accuracy and latency, reduces the search samples.
####################################... | 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
NAHAS: Neural Architecture and Hardware Accelerator Search
### Paper Abstract
Neural architectures and hardware accelerators have been two driving forces for the rapid p... |
6cppShZBHg | ICML.cc/2020/Workshop/Artemiss | 2020 | Multi-output prediction of global vegetation distribution with incomplete data | ["Rita Beigaite", "Jesse Read", "Indre Zliobaite"] | As the climate is changing, large changes in vegetation distribution are already taking place and expected to happen in the future. Our goal is to explore the links between climate and vegetation, and build a predictive model mapping climatic conditions to vegetation cover, based on global remote sensing data. The main... | ["multi-output prediction", "incomplete data", "global vegetation modeling"] | Multi-output prediction of global vegetation distribution with incomplete dataRita Beigaite1Jesse Read2Indre Zliobaite1AbstractAs the climate is changing, large changes in veg-etation distribution are already taking place andexpected to happen in the future. Our goal is toexplore the links between climate and vegetatio... | oGIirywSOWI | Multi-output feed-forward networks for the prediction of vegetation type distribution with incomplete data | 6: Marginally above acceptance threshold | The authors use feed-forward neural networks to predict the distribution of vegetation cover in areas described by the climatic features. Due to the human impact, some parts of these areas cannot be assigned to any of the 13 vegetation types. This issue is formulated as an incomplete data problem, which is similar in 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
Multi-output prediction of global vegetation distribution with incomplete data
### Paper Abstract
As the climate is changing, large changes in vegetation distribution ar... |
O_PZRnYcUCm | ICLR.cc/2021/Conference | 2021 | Blank | ["Moitreya Chatterjee", "Anoop Cherian", "Narendra Ahuja"] | Blank | ["Blank"] | J5vPwMmzFO0 | Review for Learning to Generate Videos Using Neural Uncertainty Priors | 4: Ok but not good enough - rejection | #########################
SUMMARY
The authors propose a modified training objective for VAE-based video prediction models. In particular, the goal of this formulation is to not penalize predictions that might be plausible but deviate from the given ground truth future sequence during training, which the authors identi... | 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
Blank
### Paper Abstract
Blank
### Paper Keywords
["Blank"]
### Paper Content
<|im_end|>
<|im_start|>assistant
### Review Title
Review for Learning to Generate Videos ... | |
9uvhpyQwzM_ | ICLR.cc/2021/Conference | 2021 | Evaluation of Similarity-based Explanations | ["Kazuaki Hanawa", "Sho Yokoi", "Satoshi Hara", "Kentaro Inui"] | Explaining the predictions made by complex machine learning models helps users to understand and accept the predicted outputs with confidence. One promising way is to use similarity-based explanation that provides similar instances as evidence to support model predictions. Several relevance metrics are used for this pu... | ["Interpretability", "Explainability"] | ABSTRACTExplaining the predictions made by complex machine learning models helps usersto understand and accept the predicted outputs with confidence. One promising wayis to use similarity-based explanation that provides similar instances as evidenceto support model predictions. Several relevance metrics are used for th... | 6fXmxnb2wCH | Empirical comparative study of metrics for explanations with simple requirements. | 6: Marginally above acceptance threshold | The authors investigate which relevance metrics are desirable for explanations that are based on extracting similar instances as evidence to support a model prediction. They establish three minimal requirements for those metrics: one adapted from the state-of the art: (1) model randomization to ensure the measure is m... | 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
Evaluation of Similarity-based Explanations
### Paper Abstract
Explaining the predictions made by complex machine learning models helps users to understand and accept th... |
5bUrbPuZCcy | NeurIPS.cc/2022/Workshop/Federated_Learning | 2022 | Adaptive Sparse Federated Learning in Large Output Spaces via Hashing | ["Zhaozhuo Xu", "Luyang Liu", "Zheng Xu", "Anshumali Shrivastava"] | This paper focuses on the on-device training efficiency of federated learning (FL), and demonstrates it is feasible to exploit sparsity in the client to save both computation and memory for deep neural networks with large output space. To this end, we propose a sparse FL scheme using hash-based adaptive sampling algori... | ["hashing", "federated learning"] | Adaptive Sparse Federated Learning in Large OutputSpaces via HashingAnonymous Author(s)AffiliationAddressemailAbstractThis paper focuses on the on-device training efficiency of federated learning (FL), 1and demonstrates it is feasible to exploit sparsity in the client to save both compu- 2tation and memory for deep neu... | UfEEwevi06s | This paper propose a sparse federated learning (FL) scheme which utilize the hash-based sampling method to achieve the sparsification of the model in the federated environments. However, technical novelty and experimental evaluation are both limited. | Pros:
1. This paper focus on the sparse FL for deep neural networks in large output spaces (LOS) and only does forward and backward computation on the selected neurons.
2. The experimental results show that it can maintain the same model accuracy with 30% parameters in the output layer.
Cons:
1. Lack of comparisons ... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Adaptive Sparse Federated Learning in Large Output Spaces via Hashing
### Paper Abstract
This paper focuses on the on-device training efficiency of federated learning (F... | ||
B1rQtwJDG | ICLR.cc/2018/Workshop | 2018 | Rethinking Style and Content Disentanglement in Variational Autoencoders | ["Rui Shu", "Shengjia Zhao", "Mykel J. Kochenderfer"] | A common test for whether a generative model learns disentangled representations is its ability to learn style and content as independent factors of variation on digit datasets. To achieve such disentanglement with variational autoencoders, the label information is often provided in either a fully-supervised or semi-su... | ["disentangled representation", "variational autoencoders", "deep representation prior"] | ABSTRACTA common test for whether a generative model learns disentangled representa-tions is its ability to learn style and content as independent factors of variation ondigit datasets. To achieve such disentanglement with variational autoencoders, thelabel information is often provided in either a fully-supervised or ... | rJkw-HbYG | Review | 7: Good paper, accept | The paper addresses the problem of learning representations that disentangle style from label information. A simple but important observation is made: the loss function used for training VAEs does not promote disentanglement, despite the obtained results. The authors empirically show the importance of the inductive bia... | 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
Rethinking Style and Content Disentanglement in Variational Autoencoders
### Paper Abstract
A common test for whether a generative model learns disentangled representati... |
H1edV6VKvS | ICLR.cc/2020/Conference | 2020 | ProxNet: End-to-End Learning of Structured Representation by Proximal Mapping | ["Mao Li", "Yingyi Ma", "Xinhua Zhang"] | Underpinning the success of deep learning is the effective regularization that allows a broad range of structures in data to be compactly modeled in a deep architecture. Examples include transformation invariances, robustness to adversarial/random perturbations, and correlations between multiple modalities. However, m... | ["representation learning", "multiview learning"] | ABSTRACTUnderpinning the success of deep learning is the effective regularization that allowsa broad range of structures in data to be compactly modeled in a deep architecture.Examples include transformation invariances, robustness to adversarial/randomperturbations, and correlations between multiple modalities. Howeve... | BylPwUTaYr | Official Blind Review #1 | 1: Reject | The paper proposes looking at many existing and some new layers on neural networks as proximal operations (where a proximal operator is defined as P_f(z) = argmin_{x\in C} f(x) + 1/2 ||x-z||^2). Trivially many existing layers can be written in this way, but the model can be applied to novel fairly interesting setups.... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
ProxNet: End-to-End Learning of Structured Representation by Proximal Mapping
### Paper Abstract
Underpinning the success of deep learning is the effective regularizati... | |
SyfdsjA9FX | ICLR.cc/2019/Conference | 2019 | Live Face De-Identification in Video | ["Oran Gafni", "Lior Wolf", "Yaniv Taigman"] | We propose a method for face de-identification that enables fully automatic video modification at high frame rates. The goal is to maximally decorrelate the identity, while having the perception (pose, illumination and expression) fixed. We achieve this by a novel feed forward encoder-decoder network architecture that ... | ["video", "identity", "live face", "video live face", "face", "automatic video modification", "high frame rates", "goal", "perception", "pose"] | ABSTRACTWe propose a method for face de-identification that enables fully automatic videomodification at high frame rates. The goal is to maximally decorrelate the identity,while having the perception (pose, illumination and expression) fixed. We achievethis by a novel feed forward encoder-decoder network architecture ... | BJxLJU3chQ | Non-trivial results on an interesting problem | 6: Marginally above acceptance threshold | The submission proposes a method for face de-identification.
A network is trained to transform an input face image to simultaneously stay close to the low- and mid-level representations while maximising the distance of the high-level representation of the input image in a face recognition network. Additionally an adv... | 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
Live Face De-Identification in Video
### Paper Abstract
We propose a method for face de-identification that enables fully automatic video modification at high frame rate... |
HkHrmM1PM | ICLR.cc/2018/Workshop | 2018 | Deep Convolutional Malware Classifiers Can Learn from Raw Executables and Labels Only | ["Marek Kr\u010d\u00e1l", "Ond\u0159ej \u0160vec", "Martin B\u00e1lek", "Otakar Ja\u0161ek"] | We propose and evaluate a simple convolutional deep neural network architecture detecting malicious \emph{Portable Executables} (Windows executable files) by learning from their raw sequences of bytes and labels only, that is, without any domain-specific feature extraction nor preprocessing. On a dataset of 20 million... | ["deep learning", "convolution", "malware detection", "windows executable files", "portable executable", "end-to-end", "representation learning"] | ABSTRACTWe propose and evaluate a simple convolutional deep neural network architecturedetecting malicious Portable Executables (Windows executable files) by learningfrom their raw sequences of bytes and labels only, that is, without any domain-specific feature extraction nor preprocessing. On a dataset of 20 million u... | HJ_I_WyFG | interesting application of CNNs | 6: Marginally above acceptance threshold | This paper proposes to use a convolutional neural network, taking program binaries as inputs, to detect malware. The results are on par with a network that takes in hand-crafted features, and the hidden layers are shown to be complementary to the hand-crafted features.
The paper is well written and easy to follow. The... | 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
Deep Convolutional Malware Classifiers Can Learn from Raw Executables and Labels Only
### Paper Abstract
We propose and evaluate a simple convolutional deep neural netw... |
BJgy-n0cK7 | ICLR.cc/2019/Conference | 2019 | Inter-BMV: Interpolation with Block Motion Vectors for Fast Semantic Segmentation on Video | ["Samvit Jain", "Joseph Gonzalez"] | Models optimized for accuracy on single images are often prohibitively slow to
run on each frame in a video, especially on challenging dense prediction tasks,
such as semantic segmentation. Recent work exploits the use of optical flow to
warp image features forward from select keyframes, as a means to conserve computat... | ["semantic segmentation", "video", "efficient inference", "video segmentation", "video compression"] | ABSTRACTModels optimized for accuracy on single images are often prohibitively slow torun on each frame in a video. Recent work exploits the use of optical flow to warpimage features forward from select keyframes, as a means to conserve computa-tion on video. This approach, however, achieves only limited speedup, even ... | rJgToa76nQ | The paper presents a feature interpolation strategy that has limited novelty | 3: Clear rejection | This paper presents a feature interpolation strategy for fast semantic segmentation in videos. They first compute features of keyframes, then interpolate intermediate frames based on block-motion vectors (BMV), and finally fuse the interpolated features as input to the prediction network. The experiments show that the ... | 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
Inter-BMV: Interpolation with Block Motion Vectors for Fast Semantic Segmentation on Video
### Paper Abstract
Models optimized for accuracy on single images are often pr... |
HyezBa4tPB | ICLR.cc/2020/Conference | 2020 | Dirichlet Wrapper to Quantify Classification Uncertainty in Black-Box Systems | ["Jos\u00e9 Mena Rold\u00e1n", "Oriol Pujol Vila", "Jordi Vitri\u00e0 Marca"] | Nowadays, machine learning models are becoming a utility in many sectors. AI companies deliver pre-trained encapsulated models as application programming interfaces (APIs) that developers can combine with third party components, their models, and proprietary data, to create complex data products. This complexity and th... | ["uncertainty", "black-box classifiers", "rejection", "deep learning", "NLP", "CV"] | ABSTRACTNowadays, machine learning models are becoming a utility in many sectors. AIcompanies deliver pre-trained encapsulated models as application programminginterfaces (APIs) that developers can combine with third party components, theirmodels, and proprietary data, to create complex data products. This complexityan... | SkgWFpePqH | Official Blind Review #3 | 1: Reject | This manuscript proposes to train a wrapper to assess the confidence of a black box classifier decision on new samples. The resulting uncertainty prediction is used to reject decision with low confidence, such that the accuracy of the prediction on retained samples remains high. The idea, although a bit incremental, is... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Dirichlet Wrapper to Quantify Classification Uncertainty in Black-Box Systems
### Paper Abstract
Nowadays, machine learning models are becoming a utility in many sectors... | |
uMDbGsVjCS4 | ICLR.cc/2021/Conference | 2021 | Frequency-aware Interface Dynamics with Generative Adversarial Networks | ["Lukas Prantl", "Tassilo Kugelstadt", "Jan Bender", "Nils Thuerey"] | We present a new method for reconstructing and refining complex surfaces based on physical simulations. Taking a roughly approximated simulation as input, our method infers corresponding spatial details while taking into account how they evolve over time. We consider this problem in terms of spatial and temporal freq... | ["physical simulations", "spatio-temporal dynamics", "generative adversarial networks", "fluids", "elasto-plasticity"] | ABSTRACTWe present a new method for reconstructing and refining complex surfaces basedon physical simulations. Taking a roughly approximated simulation as input, ourmethod infers corresponding spatial details while taking into account how theyevolve over time. We consider this problem in terms of spatial and temporal f... | 9hiF4UAdyJr | Generation of complex surfaces with spatial and temporal discriminators under cycle-consistance supervision of down-sampling reconstruction loss | 5: Marginally below acceptance threshold | This paper introduces the adversarial generative network into a physical simulation task, i.e., reconstruction and refining of complex surfaces. Several losses are applied in the proposed framework, such as MSE, adversarial losses, low-resolution reconstruction loss. Experimental results seem good.
Pros:
+ As far as 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
Frequency-aware Interface Dynamics with Generative Adversarial Networks
### Paper Abstract
We present a new method for reconstructing and refining complex surfaces based... |
hZklxCWmUAm | MICCAI.org/2021/Workshop/COMPAY | 2021 | A Multi-scale Graph Network with Multi-head Attention for Histopathology Image Diagnosis | ["Xiaodan Xing", "Yixin Ma", "Lei Jin", "Tianyang Sun", "Zhong Xue", "Feng Shi", "Jinsong Wu", "Dinggang Shen"] | Hematoxylin-eosin (H&E) staining plays an essential role in brain glioma diagnosis, but reading pathologic images and generating diagnostic reports can be a tedious and laborious work. Pathologists need to combine and navigate extremely large images with different scales and to quantify different aspects for subtyping.... | ["Whole slide image classification", "Graph convolution", "Attention."] | A Multi-scale Graph Network with Multi-headAttention for Histopathology Image DiagnosisSubmission ID 1135No Institute GivenAbstract. Hematoxylin-eosin (H&E) staining plays an essential role inbrain glioma diagnosis, but reading pathologic images and generating di-agnostic reports can be a tedious and laborious work. Pa... | YkEnspYVuPX | A multi-resolution graph is constructed from H&E images and a graph neural network with multi-headed attention is used for glioma subtyping | 6: Marginally above acceptance threshold | The use of multi-resolution graphs to incorporate spatial information is an attractive but the authors do not explain or explore how the multi-resolution nature of the graph affects the graph convolutions. Edges on the graph are now inherently directed as depending on the direction of traversal, the node is either incr... | 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 Multi-scale Graph Network with Multi-head Attention for Histopathology Image Diagnosis
### Paper Abstract
Hematoxylin-eosin (H&E) staining plays an essential role in b... |
uDN8pRAdsoC | ICLR.cc/2021/Conference | 2021 | Hard Masking for Explaining Graph Neural Networks | ["Thorben Funke", "Megha Khosla", "Avishek Anand"] | Graph Neural Networks (GNNs) are a flexible and powerful family of models that build nodes' representations on irregular graph-structured data. This paper focuses on explaining or interpreting the rationale underlying a given prediction of already trained graph neural networks for the node classification task. Existing... | ["Interpretability", "Graph Neural Networks", "Hard Masks"] | ABSTRACTGraph Neural Networks (GNNs) are a flexible and powerful family of modelsthat build nodes’ representations on irregular graph-structured data. This paperfocuses on explaining or interpreting the rationale underlying a given predictionof already trained graph neural networks for the node classification task. Exi... | KaUa_fxo27w | Method for explaining GNN behaviour with room for improvement | 5: Marginally below acceptance threshold | The authors address the problem of explaining the behaviour of graph neural networks (which operate on a computation graph based on their k-hop neighbourhood) such as a graph convolutional network (GCN).
The core idea is to identify, for each node v in the graph, the nodes and features of the graph most relevant to 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
Hard Masking for Explaining Graph Neural Networks
### Paper Abstract
Graph Neural Networks (GNNs) are a flexible and powerful family of models that build nodes' represen... |
rkcya1ZAW | ICLR.cc/2018/Conference | 2018 | Continuous-Time Flows for Efficient Inference and Density Estimation | ["Changyou Chen", "Chunyuan Li", "Liqun Chen", "Wenlin Wang", "Yunchen Pu", "Lawrence Carin"] | Two fundamental problems in unsupervised learning are efficient inference for latent-variable models and robust density estimation based on large amounts of unlabeled data. For efficient inference, normalizing flows have been recently developed to approximate a target distribution arbitrarily well. In practice, however... | ["continuous-time flows", "efficient inference", "density estimation", "deep generative models"] | ABSTRACTTwo fundamental problems in unsupervised learning are efficient inference forlatent-variable models and robust density estimation based on large amounts ofunlabeled data. For efficient inference, normalizing flows have been recently de-veloped to approximate a target distribution arbitrarily well. In practice, ... | HkCNqISxM | Not convinced! | 3: Clear rejection | The authors try to use continuous time generalizations of normalizing flows for improving upon VAE-like models or for standard density estimation problems.
Clarity: the text is mathematically very sloppy / hand-wavy.
1. I do not understand proposition (1). I do not think that the proof is correct (e.g. the generator ... | 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
Continuous-Time Flows for Efficient Inference and Density Estimation
### Paper Abstract
Two fundamental problems in unsupervised learning are efficient inference for lat... |
SJg5J6NtDr | ICLR.cc/2020/Conference | 2020 | Watch, Try, Learn: Meta-Learning from Demonstrations and Rewards | ["Allan Zhou", "Eric Jang", "Daniel Kappler", "Alex Herzog", "Mohi Khansari", "Paul Wohlhart", "Yunfei Bai", "Mrinal Kalakrishnan", "Sergey Levine", "Chelsea Finn"] | Imitation learning allows agents to learn complex behaviors from demonstrations. However, learning a complex vision-based task may require an impractical number of demonstrations. Meta-imitation learning is a promising approach towards enabling agents to learn a new task from one or a few demonstrations by leveraging e... | ["meta-learning", "reinforcement learning", "imitation learning"] | ABSTRACTImitation learning allows agents to learn complex behaviors from demonstrations.However, learning a complex vision-based task may require an impractical num-ber of demonstrations. Meta-imitation learning is a promising approach towardsenabling agents to learn a new task from one or a few demonstrations by lever... | rJe17q5Z9H | Official Blind Review #3 | 8: Accept | The paper introduces a meta-learning approach that can learn from demonstrations and subsequent reinforcement learning trials. The approach works in two phases. First, multiple demonstrations are collected for each task and a meta-learning policy is obtained by embedding the demonstrations of a single task in a context... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Watch, Try, Learn: Meta-Learning from Demonstrations and Rewards
### Paper Abstract
Imitation learning allows agents to learn complex behaviors from demonstrations. Howe... | |
SJexHkSFPS | ICLR.cc/2020/Conference | 2020 | Thinking While Moving: Deep Reinforcement Learning with Concurrent Control | ["Ted Xiao", "Eric Jang", "Dmitry Kalashnikov", "Sergey Levine", "Julian Ibarz", "Karol Hausman", "Alexander Herzog"] | We study reinforcement learning in settings where sampling an action from the policy must be done concurrently with the time evolution of the controlled system, such as when a robot must decide on the next action while still performing the previous action. Much like a person or an animal, the robot must think and move ... | ["deep reinforcement learning", "continuous-time", "robotics"] | ABSTRACTWe study reinforcement learning in settings where sampling an action from thepolicy must be done concurrently with the time evolution of the controlled system,such as when a robot must decide on the next action while still performing the pre-vious action. Much like a person or an animal, the robot must think an... | ByxoiRtUcB | Official Blind Review #2 | 6: Weak Accept | This paper proposes a first step in the direction of 'real-world relevant RL approaches' in the sense of considering environments that don't halt their execution until an agent has finished its optimal action computation and execution but actually just go on being an environment. For this, the notion of a concurrent ac... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Thinking While Moving: Deep Reinforcement Learning with Concurrent Control
### Paper Abstract
We study reinforcement learning in settings where sampling an action from t... | |
04ArenGOz3 | ICLR.cc/2021/Conference | 2021 | Set Prediction without Imposing Structure as Conditional Density Estimation | ["David W Zhang", "Gertjan J. Burghouts", "Cees G. M. Snoek"] | Set prediction is about learning to predict a collection of unordered variables with unknown interrelations. Training such models with set losses imposes the structure of a metric space over sets. We focus on stochastic and underdefined cases, where an incorrectly chosen loss function leads to implausible predictions. ... | ["set prediction", "energy based models"] | ABSTRACTSet prediction is about learning to predict a collection of unordered variableswith unknown interrelations. Training such models with set losses imposes thestructure of a metric space over sets. We focus on stochastic and underdefinedcases, where an incorrectly chosen loss function leads to implausible predicti... | Tgwd_d-_Npm | A novel framework for set prediction | 7: Good paper, accept | Authors propose a new method for formulating set prediction tasks. They propose to use a noisy energy-based model with langevin mcmc + noisy startup as their model. The can approximate the gradient of the likelihood function by computing the enery of ground truth pairs and energy of synthesized pairs where the target i... | 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
Set Prediction without Imposing Structure as Conditional Density Estimation
### Paper Abstract
Set prediction is about learning to predict a collection of unordered vari... |
lfJpQn3xPV- | ICLR.cc/2021/Conference | 2021 | Online Learning of Graph Neural Networks: When Can Data Be Permanently Deleted | ["Lukas Paul Achatius Galke", "Benedikt Franke", "Tobias Zielke", "Ansgar Scherp"] | Online learning of graph neural networks (GNNs) faces the challenges of distribution shift and ever growing and changing training data, when temporal graphs evolve over time. This makes it inefficient to train over the complete graph whenever new data arrives. Deleting old data at some point in time may be preferable t... | ["graph neural networks", "online learning"] | ABSTRACTOnline learning of graph neural networks (GNNs) faces the challenges of distribu-tion shift and ever gbv rowing and changing training data, when temporal graphsevolve over time. This makes it inefficient to train over the complete graph when-ever new data arrives. Deleting old data at some point in time may be ... | MNWzxywcNVF | Empirical work. Results are expected but not exciting. | 5: Marginally below acceptance threshold | This work empirically evaluates the sliding-window strategy for training GNNs with temporal graphs. One may cast the temporal nature of the graph data in an online setting, under which the change of the graph structure as well as the variation of the classes cause distribution shift. The authors conduct a series of exp... | 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
Online Learning of Graph Neural Networks: When Can Data Be Permanently Deleted
### Paper Abstract
Online learning of graph neural networks (GNNs) faces the challenges of... |
8yTS_nAILxt | robot-learning.org/CoRL/2023/Conference | 2023 | REFLECT: Summarizing Robot Experiences for Failure Explanation and Correction | ["Zeyi Liu", "Arpit Bahety", "Shuran Song"] | The ability to detect and analyze failed executions automatically is crucial for an explainable and robust robotic system. Recently, Large Language Models (LLMs) have demonstrated strong reasoning abilities on textual inputs. To leverage the power of LLMs for robot failure explanation, we introduce REFLECT, a framework... | ["Large Language Model", "Explainable AI", "Task Planning"] | REFLECT: Summarizing RobotExperiences for FaiLure Explanation and Correc TionZeyi Liu∗Arpit Bahety∗Shuran SongColumbia University, New Y ork, NY , United Stateshttps://robot-reflect.github.io/Abstract: The ability to detect and analyze failed executions automatically is crucial foran explainable and robust robotic syst... | ugHTULZQ6kT | **Strengths**
* The pipeline(s) for logging robot data and related events is impressive
* The prompt generation scheme used by the authors makes sense
* The ablation studies and the results adequately test the papers' claims on the systems contributed by the authors
* The authors leverage the LLM's ability to localize... | 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
REFLECT: Summarizing Robot Experiences for Failure Explanation and Correction
### Paper Abstract
The ability to detect and analyze failed executions automatically is cru... | ||
ryxB2lBtvH | ICLR.cc/2020/Conference | 2020 | Learning to Coordinate Manipulation Skills via Skill Behavior Diversification | ["Youngwoon Lee", "Jingyun Yang", "Joseph J. Lim"] | When mastering a complex manipulation task, humans often decompose the task into sub-skills of their body parts, practice the sub-skills independently, and then execute the sub-skills together. Similarly, a robot with multiple end-effectors can perform complex tasks by coordinating sub-skills of each end-effector. To r... | ["reinforcement learning", "hierarchical reinforcement learning", "modular framework", "skill coordination", "bimanual manipulation"] | ABSTRACTWhen mastering a complex manipulation task, humans often decompose the taskinto sub-skills of their body parts, practice the sub-skills independently, and thenexecute the sub-skills together. Similarly, a robot with multiple end-effectors canperform complex tasks by coordinating sub-skills of each end-effector.... | r1eDu0Dk5S | Official Blind Review #1 | 6: Weak Accept | The paper presents a hierarchical reinforcement learning method for coordination of multiple cooperative agents with pre-learned adaptable skills. These skills are learned via a maximum entropy objective where diversity of behaviour given each skill is maximised and controlled via a latent conditioning vector. This all... | <|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 to Coordinate Manipulation Skills via Skill Behavior Diversification
### Paper Abstract
When mastering a complex manipulation task, humans often decompose the t... | |
dlEJsyHGeaL | ICLR.cc/2021/Conference | 2021 | Graph Edit Networks | ["Benjamin Paassen", "Daniele Grattarola", "Daniele Zambon", "Cesare Alippi", "Barbara Eva Hammer"] | While graph neural networks have made impressive progress in classification and regression, few approaches to date perform time series prediction on graphs, and those that do are mostly limited to edge changes. We suggest that graph edits are a more natural interface for graph-to-graph learning. In particular, graph e... | ["graph neural networks", "graph edit distance", "time series prediction", "structured prediction"] | ABSTRACTWhile graph neural networks have made impressive progress in classification andregression, few approaches to date perform time series prediction on graphs, andthose that do are mostly limited to edge changes. We suggest that graph edits are amore natural interface for graph-to-graph learning. In particular, gra... | Sg922s85khx | Official Blind Review | 6: Marginally above acceptance threshold | Summary
-------
This paper proposes a model that, given one graph, predicts a sequence of edits that transforms this graph to the next one in a sequence of evolving graphs. To this end, it proposes the graph edit network (GEN), which is a linear output layer that transforms node embeddings to a set of scores that are t... | 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 Edit Networks
### Paper Abstract
While graph neural networks have made impressive progress in classification and regression, few approaches to date perform time se... |
rJl31TNYPr | ICLR.cc/2020/Conference | 2020 | Fooling Detection Alone is Not Enough: Adversarial Attack against Multiple Object Tracking | ["Yunhan Jia", "Yantao Lu", "Junjie Shen", "Qi Alfred Chen", "Hao Chen", "Zhenyu Zhong", "Tao Wei"] | Recent work in adversarial machine learning started to focus on the visual perception in autonomous driving and studied Adversarial Examples (AEs) for object detection models. However, in such visual perception pipeline the detected objects must also be tracked, in a process called Multiple Object Tracking (MOT), to bu... | ["Adversarial examples", "object detection", "object tracking", "security", "autonomous vehicle", "deep learning"] | ABSTRACTRecent work in adversarial machine learning started to focus on the visual percep-tion in autonomous driving and studied Adversarial Examples (AEs) for objectdetection models. However, in such visual perception pipeline the detected objectsmust also be tracked, in a process called Multiple Object Tracking (MOT)... | Hke0M2aaKr | Official Blind Review #2 | 6: Weak Accept | The paper addresses adversarial attacks against visual perception pipelines in autonomous driving. Both subprocesses in the visual perception pipeline, object detection and multiple object tracking (MOT), are considered. The paper proposes a novel approach in adversarial attacks, the tracking hijacking, which can fool ... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Fooling Detection Alone is Not Enough: Adversarial Attack against Multiple Object Tracking
### Paper Abstract
Recent work in adversarial machine learning started to focu... | |
ByxLBMZCb | ICLR.cc/2018/Conference | 2018 | Learning Deep Models: Critical Points and Local Openness | ["Maher Nouiehed", "Meisam Razaviyayn"] | With the increasing interest in deeper understanding of the loss surface of many non-convex deep models, this paper presents a unifying framework to study the local/global optima equivalence of the optimization problems arising from training of such non-convex models. Using the "local openness" property of the underl... | ["Training Deep Models", "Non-convex Optimization", "Local and Global Equivalence", "Local Openness"] | ABSTRACTWith the increasing interest in deeper understanding of the loss surface of many non-convex deepmodels , this paper presents a unifying framework to study the local/global optima equivalence of theoptimization problems arising from training of such non-convex models. Using the local opennessproperty of the unde... | SJtc2C4bz | Theory building work for linear deep networks, and some nonlinear networks. Redresses and unifies some existing results in a clean way. | 6: Marginally above acceptance threshold | The paper studies the local optima of certain types of deep networks. It uses the notion of a locally open map to draw equivalences between local optima and global optima. The basic idea is that for fitting nonlinear models with a convex loss, if the mapping from the weights to the outputs is open, then every local opt... | 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 Deep Models: Critical Points and Local Openness
### Paper Abstract
With the increasing interest in deeper understanding of the loss surface of many non-convex ... |
BkeMXR4KvS | ICLR.cc/2020/Conference | 2020 | DASGrad: Double Adaptive Stochastic Gradient | ["Kin Gutierrez", "Cristian Challu", "Jin Li", "Artur Dubrawski"] | Adaptive moment methods have been remarkably successful for optimization under the presence of high dimensional or sparse gradients, in parallel to this, adaptive sampling probabilities for SGD have allowed optimizers to improve convergence rates by prioritizing examples to learn efficiently. Numerous applications in t... | ["stochastic convex optimization", "adaptivity", "online learning", "transfer learning"] | ABSTRACTAdaptive moment methods have been remarkably successful for optimization un-der the presence of high dimensional or sparse gradients, in parallel to this, adap-tive sampling probabilities for SGD have allowed optimizers to improve conver-gence rates by prioritizing examples to learn efficiently. Numerous applic... | HklOgsSRFr | Official Blind Review #1 | 3: Weak Reject | In the paper, the authors propose a general form that covers most stochastic gradient methods so far, e.g. stochastic gradient descent, Adam, or adaptive probabilities methods. Then, they also provide a convergence analysis of convex problems. In the experiments, they compared the proposed DASGRAD method with Adam, 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
DASGrad: Double Adaptive Stochastic Gradient
### Paper Abstract
Adaptive moment methods have been remarkably successful for optimization under the presence of high dimen... | |
KYPz4YsCPj | ICLR.cc/2021/Conference | 2021 | Inductive Representation Learning in Temporal Networks via Causal Anonymous Walks | ["Yanbang Wang", "Yen-Yu Chang", "Yunyu Liu", "Jure Leskovec", "Pan Li"] | Temporal networks serve as abstractions of many real-world dynamic systems. These networks typically evolve according to certain laws, such as the law of triadic closure, which is universal in social networks. Inductive representation learning of temporal networks should be able to capture such laws and further be appl... | ["temporal networks", "inductive representation learning", "anonymous walk", "network motif"] | ABSTRACTTemporal networks serve as abstractions of many real-world dynamic systems.These networks typically evolve according to certain laws, such as the law of triadicclosure, which is universal in social networks. Inductive representation learningof temporal networks should be able to capture such laws and further be... | 8oL77rt82e9 | INDUCTIVE REPRESENTATION LEARNING IN TEMPO- RAL NETWORKS VIA CAUSAL ANONYMOUS WALKS | 5: Marginally below acceptance threshold | This paper proposes Causal Anonymous Walks (CAWs) that are extracted by temporal random walks and work as automatic retrieval of temporal network motifs to represent network dynamics while avoiding the time-consuming selection and counting of those motifs. CAWs adopt an anonymization strategy that replaces node identit... | 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
Inductive Representation Learning in Temporal Networks via Causal Anonymous Walks
### Paper Abstract
Temporal networks serve as abstractions of many real-world dynamic s... |
rkbvIvyvM | ICLR.cc/2018/Workshop | 2018 | Prototype Matching Networks for Large-Scale Multi-Label Classification | ["Jack Lanchantin", "Arshdeep Sekhon", "Ritambhara Singh", "Yanjun Qi"] | One of the fundamental tasks in understanding genomics is the problem of predicting Transcription Factor Binding Sites (TFBSs). With more than hundreds of Transcription Factors (TFs) as labels, genomic-sequence based TFBS prediction is a challenging multi-label classification task. There are two major biological mechan... | ["multi-label classification", "genomics", "memory-augmented", "deep learning"] | ABSTRACTOne of the fundamental tasks in understanding genomics is the problem of pre-dicting Transcription Factor Binding Sites (TFBSs). With more than hundreds ofTranscription Factors (TFs) as labels, genomic-sequence based TFBS prediction isa challenging multi-label classification task. There are two major biological... | rkoUGHJKf | Little ML contribution, more suited for a genomics/computational biology conference. | 4: Ok but not good enough - rejection | The paper maps a pair (DNA sequence, transcription factor set) into a binary vector that indicate whether each transcription factor binds. The paper might have an application impact that I cannot judge but as far as machine learning is concerned, the contribution is not significant. I would recommend withdrawing from 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
Prototype Matching Networks for Large-Scale Multi-Label Classification
### Paper Abstract
One of the fundamental tasks in understanding genomics is the problem of predic... |
TIV7eEY8qY | ICLR.cc/2023/Workshop/RRL | 2023 | Chain-of-Thought Predictive Control | ["Zhiwei Jia", "Fangchen Liu", "Vineet Thumuluri", "Linghao Chen", "Zhiao Huang", "Hao Su"] | We study generalizable policy learning from demonstrations for complex low-level control tasks (e.g., contact-rich object manipulations). We propose an imitation learning method that incorporates the idea of temporal abstraction and the planning capabilities from Hierarchical RL (HRL) in a novel and effective manner. A... | ["Behavior Cloning", "Hierarchical Imitation Learning", "Generalizable Policy Learning", "Object Manipulation"] | Chain-of-Thought Predictive ControlZhiwei Jia1Fangchen Liu2Vineet Thumuluri1Linghao Chen3Zhiao Huang1Hao Su1AbstractWe study generalizable policy learning fromdemonstrations for complex low-level controltasks (e.g., contact-rich object manipulations).We propose an imitation learning method thatincorporates the idea of ... | zPfiii-GXy | The paper proposes a Transformer based method for Imitation Learning which shows very good performance on complex Control problems. The experimental results are very good but there are some caveats with the proposed method and the baselines used. | 3: Accept | This paper proposes a Transformer-based imitation learning method for solving complex robotic control tasks. The paper is clearly written and understandable. There are several illustrations that can help with understanding the proposed method. The proposed method named CoTPC outperforms several other BC methods by sign... | 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
Chain-of-Thought Predictive Control
### Paper Abstract
We study generalizable policy learning from demonstrations for complex low-level control tasks (e.g., contact-rich... |
S1x0CnEtvB | ICLR.cc/2020/Conference | 2020 | AutoGrow: Automatic Layer Growing in Deep Convolutional Networks | ["Wei Wen", "Feng Yan", "Hai Li"] | Depth is a key component of Deep Neural Networks (DNNs), however, designing depth is heuristic and requires many human efforts. We propose AutoGrow to automate depth discovery in DNNs: starting from a shallow seed architecture, AutoGrow grows new layers if the growth improves the accuracy; otherwise, stops growing and ... | ["Growing", "depth", "neural networks", "automation"] | ABSTRACTDepth is a key component of Deep Neural Networks (DNNs), however, design-ing depth is heuristic and requires many human efforts. We propose AutoGrowto automate depth discovery in DNNs: starting from a shallow seed architecture,AutoGrow grows new layers if the growth improves the accuracy; otherwise, stopsgrowin... | SkeAVHaRYS | Official Blind Review #3 | 3: Weak Reject | This paper's contribution is a method for automatically growing the depth of a neural network during training. It compares several heuristics that may be used to successfully achieve this goal and identifies a set of choices that work well together on multiple datasets.
The paper focuses on CNNs that conform to a popu... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
AutoGrow: Automatic Layer Growing in Deep Convolutional Networks
### Paper Abstract
Depth is a key component of Deep Neural Networks (DNNs), however, designing depth is ... | |
ni_nys-C9D6 | ICLR.cc/2021/Conference | 2021 | Differentiate Everything with a Reversible Domain-Specific Language | ["JinGuo Liu", "Taine Zhao"] | Reverse-mode automatic differentiation (AD) suffers from the issue of having too much space overhead to trace back intermediate computational states for backpropagation.
The traditional method to trace back states is called checkpointing that stores intermediate states into a global stack and restore state through eith... | ["Reversible computing", "automatic differentiation", "Julia"] | Under review as a conference paper at ICLR 2021Differentiate Everything with a Reversible EmbededDomain -Specific LanguageAnonymous authorsPaper under double-blind reviewAbstractReverse-mode automatic di erentiation (AD) su ers from the issue of having toomuch space overhead to trace back intermediate computational sta... | -lBKDkNq67 | Review | 5: Marginally below acceptance threshold | # Changes after rebuttal
Thanks to the authors for their answers to the questions and their revisions to improve the manuscript. It is useful to have further descriptions of reversible computing for an audience that may be unfamiliar with the topic. I would encourage the authors to make further revisions to more concis... | 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
Differentiate Everything with a Reversible Domain-Specific Language
### Paper Abstract
Reverse-mode automatic differentiation (AD) suffers from the issue of having too m... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.