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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Px7xIKHjmMS | ICLR.cc/2021/Conference | 2021 | Beyond GNNs: A Sample Efficient Architecture for Graph Problems | ["Pranjal Awasthi", "Abhimanyu Das", "Sreenivas Gollapudi"] | Despite their popularity in learning problems over graph structured data, existing Graph Neural Networks (GNNs) have inherent limitations for fundamental graph problems such as shortest paths, $k$-connectivity, minimum spanning tree and minimum cuts. In all these instances, it is known that one needs GNNs of high depth... | ["Graph Neural Networks", "Deep Learning Theory", "Graph Connectivity", "Minimum Spanning Trees"] | ABSTRACTDespite their popularity in learning problems over graph structured data, exist-ingGraph Neural Networks (GNNs) have inherent limitations for fundamentalgraph problems such as shortest paths, k-connectivity, minimum spanning tree andminimum cuts. In all these instances, it is known that one needs GNNs of highde... | 1ejc1mWKI2r | Interesting paper | 8: Top 50% of accepted papers, clear accept | The main question this paper tackles is: can one develop sample efficient architectures for graph problems while retaining the simplicity of the message passing framework?
While combining message passing with GNNs has shown to have positive empirical results, we do not know of a general neural architecture that is v... | 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
Beyond GNNs: A Sample Efficient Architecture for Graph Problems
### Paper Abstract
Despite their popularity in learning problems over graph structured data, existing Gra... |
kUk79nBY__2 | ICLR.cc/2023/BlogPosts | 2023 | A Hitchhiker's Guide to Momentum | ["Fabian Pedregosa"] | Polyak momentum is one of the most iconic methods in optimization. Despite it's simplicit, it features rich dynamics that depend both on the step-size and momentum parameter. In this blog post we identify the different regions of the parameter space and discuss their convergence properties using the theory of Chebyshev... | ["optimization", "momentum", "heavy ball"] | ZuwjsdeP8UT | Excellent blog post on convergence behavior of the gradient descent with momentum | 9: Top 15% of accepted papers, strong accept | The article gives a really nice description of the convergence of the gradient descent with momentum in the deterministic case (deterministic gradients, no additive noise etc.). It resembles classical numerical analysis study of convergence (as mentioned, also reference Rutishauser 1959 given), where the asymptotic 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
A Hitchhiker's Guide to Momentum
### Paper Abstract
Polyak momentum is one of the most iconic methods in optimization. Despite it's simplicit, it features rich dynamics ... | |
BJ78bJZCZ | ICLR.cc/2018/Conference | 2018 | Efficiently applying attention to sequential data with the Recurrent Discounted Attention unit | ["Brendan Maginnis", "Pierre Richemond"] | Recurrent Neural Networks architectures excel at processing sequences by
modelling dependencies over different timescales. The recently introduced
Recurrent Weighted Average (RWA) unit captures long term dependencies
far better than an LSTM on several challenging tasks. The RWA achieves
this by applying attention to ea... | ["RNNs"] | ABSTRACTRecurrent Neural Networks architectures excel at processing sequences by mod-elling dependencies over different timescales. The recently introduced RecurrentWeighted Average (RWA) unit captures long term dependencies far better thanan LSTM on several challenging tasks. The RWA achieves this by applying at-tenti... | BkEYMCPlG | Improvement upon a rarely used technique still not showing great results yet. | 4: Ok but not good enough - rejection | The authors present RDA, the Recurrent Discounted Attention unit, that improves upon RWA, the earlier introduced Recurrent Weighted Average unit, by adding a discount factor. While the RWA was an interesting idea with bad results (far worse than the standard GRU or LSTM with standard attention except for hand-picked ta... | 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
Efficiently applying attention to sequential data with the Recurrent Discounted Attention unit
### Paper Abstract
Recurrent Neural Networks architectures excel at proces... |
BJgRsyBtPB | ICLR.cc/2020/Conference | 2020 | A Greedy Approach to Max-Sliced Wasserstein GANs | ["Andr\u00e1s Horv\u00e1th"] | Generative Adversarial Networks have made data generation possible in various use cases, but in case of complex, high-dimensional distributions it can be difficult to train them, because of convergence problems and the appearance of mode collapse.
Sliced Wasserstein GANs and especially the application of the Max-Sliced... | ["GEnerative Adversarial Networks", "GANs", "Wasserstein distances", "Sliced Wasserstein Distance", "Max-sliced Wasserstein distance"] | ABSTRACTGenerative Adversarial Networks have made data generation possible in varioususe cases, but in case of complex, high-dimensional distributions it can be difficultto train them, because of convergence problems and the appearance of mode col-lapse. Sliced Wasserstein GANs and especially the application of the Max... | rygE9UQgqS | Official Blind Review #2 | 1: Reject | This paper proposes two alternative approaches to max-sliced Wasserstein GANs. They are based on the authors’ claim that there is a “flaw” in the Wasserstein-1 distance between probability distributions on the one-dimensional space. Briefly, the authors’ argument says that the “flaw” is that the optimal transport may n... | <|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 Greedy Approach to Max-Sliced Wasserstein GANs
### Paper Abstract
Generative Adversarial Networks have made data generation possible in various use cases, but in case ... | |
_V3kUd_BhGD | MICCAI.org/2022/Challenge/FLARE | 2022 | Self-pretrained V-net based on PCRL for Abdominal Organ Segmentation | ["Jiapeng Zhang"] | Abdomen organ segmentation has many important clinical applications. However, the manual annotating process is time-consuming and labor-intensive. In the "Fast and Low-resource semi-supervised Abdominal oRgan sEgmentation in CT" challenge, the organizer provide massive unlabeled CT images. To effectively utilize unlabe... | ["Self-supervised learning", "Self-transfer learning", "Organ Segmentation"] | Self-pretrained V-net based on PCRL forAbdominal Organ SegmentationJiapeng ZhangIDUniversity of Shanghai for Science and Technology, Shanghai, ChinaAbstract. Abdomen organ segmentation has many important clinicalapplications. However, the manual annotating process is time-consumingand labor-intensive. In the "Fast and ... | HXyQgRvlm9f | More weaknesses in the paper itself | 4: Ok but not good enough - rejection | The authors designed a preservational contrastive representation of learning (PCRL) using different argumentations on unlabeled data to get the self-pre-trained network weight.
In Sections Experiments, PCRL has only improved segmentation performance on the right adrenal gland (RAG) and left adrenal gland (LAG).
There... | 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
Self-pretrained V-net based on PCRL for Abdominal Organ Segmentation
### Paper Abstract
Abdomen organ segmentation has many important clinical applications. However, the... |
BC4UYzbLRZ | MIDL.io/2023/Short_Paper_Track | 2023 | 3D Supervised Contrastive-Learning Network for Classification of Ovarian Neoplasms | ["Tarun Kanti Roy", "Suely Oliveira", "Jesus Gonzalez Bosquet", "Xiaodong Wu"] | Ovarian cancer ranks the $5^{th}$ in cancer deaths among women, accounting for more deaths than any other cancer of the female reproductive system.
We propose a 3D contrastive learning based predictive model to discriminate benign from malignant masses in abdominal CT scans for ovarian cancer patients. We used fully s... | ["Supervised contrastive learning", "ovarian cancer", "classification", "deep learning", "feature encoder", "densenet", "resnet", "cross validation", "tumors", "CT scan"] | Medical Imaging with Deep Learning – Under Review 2023 Short Paper – MIDL 2023 submission3D Supervised Contrastive-Learning Network forClassification of Ovarian NeoplasmsTarun Roy tarunkanti-roy@uiowa.eduJesus Gonzalez Bosquet jsus-gonzalezbosquet@uiowa.eduSuely Oliveira suely-oliveira@uiowa.eduXiaodong Wu xiaodong-wu@... | RaOamhoAs0D | Constrastive learning of benign/malignant ovarian masses | 6: Marginally above acceptance threshold | The paper proposes an approach for ovarian mass classification based on contrastive learning, which outperforms a traditional training setup on a private dataset of 331 patients.
Pros:
* Training done in 3D
* Several baselines compared
Cons:
* Abstract talks about early detection, but it is unclear how advanced 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
3D Supervised Contrastive-Learning Network for Classification of Ovarian Neoplasms
### Paper Abstract
Ovarian cancer ranks the $5^{th}$ in cancer deaths among women, acc... |
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... | SkG_8jUVe | Review | 5: Marginally below acceptance threshold | This paper explores the use of Open Bigrams as a target representation of words, for application to handwriting image recognition.
Pros:
- The use of OBs is novel and interesting.
- Clearly written and explained.
Cons:
- No comparison to previous state of the art, only with author-generated results.
- More ablation... | 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
Cortical-Inspired Open-Bigram Representation for Handwritten Word Recognition
### Paper Abstract
Recent research in the cognitive process of reading hypothesized that we... |
rkHywl-A- | ICLR.cc/2018/Conference | 2018 | Learning Robust Rewards with Adverserial Inverse Reinforcement Learning | ["Justin Fu", "Katie Luo", "Sergey Levine"] | Reinforcement learning provides a powerful and general framework for decision
making and control, but its application in practice is often hindered by the need
for extensive feature and reward engineering. Deep reinforcement learning methods
can remove the need for explicit engineering of policy or value features, but
... | ["inverse reinforcement learning", "deep reinforcement learning"] | ABSTRACTReinforcement learning provides a powerful and general framework for decisionmaking and control, but its application in practice is often hindered by the needfor extensive feature and reward engineering. Deep reinforcement learning meth-ods can remove the need for explicit engineering of policy or value feature... | ryZzenclz | Using the deterministic-MDP formulation of MaxEnt IRL is a concern | 6: Marginally above acceptance threshold | SUMMARY:
This paper considers the Inverse Reinforcement Learning (IRL) problem, and particularly suggests a method that obtains a reward function that is robust to the change of dynamics of the MDP.
It starts from formulating the problem within the MaxEnt IRL framework of Ziebart et al. (2008). The challenge of MaxEnt... | 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
Learning Robust Rewards with Adverserial Inverse Reinforcement Learning
### Paper Abstract
Reinforcement learning provides a powerful and general framework for decision ... |
HkGGfhC5Y7 | ICLR.cc/2019/Conference | 2019 | Towards a better understanding of Vector Quantized Autoencoders | ["Aurko Roy", "Ashish Vaswani", "Niki Parmar", "Arvind Neelakantan"] | Deep neural networks with discrete latent variables offer the promise of better symbolic reasoning, and learning abstractions that are more useful to new tasks. There has been a surge in interest in discrete latent variable models, however, despite several recent improvements, the training of discrete latent variable... | ["machine translation", "vector quantized autoencoders", "non-autoregressive", "NMT"] | Under review as a conference paper at ICLR 2019Towards a better understanding of VectorQuantized AutoencodersAnonymous authorsPaper under double-blind reviewAbstractDeep neural networks with discrete latent variables offer the promise ofbetter symbolic reasoning, and learning abstractions that are more usefulto new tas... | H1xD-S0dh7 | Training procedure for VQ-VAE is equivalent to the EM algorithm | 7: Good paper, accept | General:
The paper presents an alternative view on the training procedure for the VQ-VAE. The authors have noticed that there is a close connection between the original training algorithm and the well-known EM algorithm. Then, they proposed to use the soft EM algorithm. In the experiments the authors showed that the so... | 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
Towards a better understanding of Vector Quantized Autoencoders
### Paper Abstract
Deep neural networks with discrete latent variables offer the promise of better symbol... |
D1E1h-K3jso | ICLR.cc/2021/Conference | 2021 | Learning from Noisy Data with Robust Representation Learning | ["Junnan Li", "Caiming Xiong", "Steven Hoi"] | Learning from noisy data has attracted much attention, where most methods focus on label noise. In this work, we propose a new framework which simultaneously addresses three types of noise commonly seen in real-world data: label noise, out-of-distribution input, and input corruption. In contrast to most existing method... | ["label noise", "out-of-distribution noise", "contrastive learning"] | ABSTRACTLearning from noisy data has attracted much attention, where most methods focuson label noise. In this work, we propose a new framework which simultaneouslyaddresses three types of noise commonly seen in real-world data: label noise, out-of-distribution input, and input corruption. In contrast to most existing ... | HwPVdjHOvMc | An extention of contrastive learning for training noise-robust deep networks. | 6: Marginally above acceptance threshold | #######################################################################
Summary:
The paper proposes noise-robust contrastive learning to combat label noise, out-of-distribution input and input corruption simultaneously. In particular, this paper embeds images into low-dimensional representations by training an autoe... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Learning from Noisy Data with Robust Representation Learning
### Paper Abstract
Learning from noisy data has attracted much attention, where most methods focus on label ... |
SJdCUMZAW | ICLR.cc/2018/Conference | 2018 | Data-efficient Deep Reinforcement Learning for Dexterous Manipulation | ["Ivo Popov", "Nicolas Heess", "Timothy P. Lillicrap", "Roland Hafner", "Gabriel Barth-Maron", "Matej Vecerik", "Thomas Lampe", "Tom Erez", "Yuval Tassa", "Martin Riedmiller"] | Grasping an object and precisely stacking it on another is a difficult task for traditional robotic control or hand-engineered approaches. Here we examine the problem in simulation and provide techniques aimed at solving it via deep reinforcement learning. We introduce two straightforward extensions to the Deep Determi... | ["Reinforcement learning", "robotics", "dexterous manipulation", "off-policy learning"] | ABSTRACTGrasping an object and precisely stacking it on another is a difficult task for tra-ditional robotic control or hand-engineered approaches. Here we examine theproblem in simulation and provide techniques aimed at solving it via deep rein-forcement learning. We introduce two straightforward extensions to the Dee... | SkHZuZqxf | Multiple minor extensions | 3: Clear rejection | The title is too generic and even a bit misleading. Dexterous manipulation usually refers to more complex skills, like in-hand manipulation or using the fingers to turn an object, and not simple pick and place tasks. Reinforcement learning methods are generally aiming to be data-efficient, and the method does not seem ... | 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
Data-efficient Deep Reinforcement Learning for Dexterous Manipulation
### Paper Abstract
Grasping an object and precisely stacking it on another is a difficult task for ... |
ryMxXPFex | ICLR.cc/2017/conference | 2017 | Discrete Variational Autoencoders | ["Jason Tyler Rolfe"] | Probabilistic models with discrete latent variables naturally capture datasets composed of discrete classes. However, they are difficult to train efficiently, since backpropagation through discrete variables is generally not possible. We present a novel method to train a class of probabilistic models with discrete late... | ["Deep learning", "Unsupervised Learning"] | ABSTRACTProbabilistic models with discrete latent variables naturally capture datasets com-posed of discrete classes. However, they are difficult to train efficiently, sincebackpropagation through discrete variables is generally not possible. We presenta novel method to train a class of probabilistic models with discre... | r16qB307e | Rich set of ideas on how to make VAEs work better. | 8: Top 50% of accepted papers, clear accept | This is an interesting paper on how to handle reparameterization in VAEs when you have discrete variables. The idea is to introduce a smoothing transformation that is shared between the generative model and the recognition model (leading to cancellations).
A second contribution is to introduce an RBM as the prior mode... | 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
Discrete Variational Autoencoders
### Paper Abstract
Probabilistic models with discrete latent variables naturally capture datasets composed of discrete classes. However... |
J3OUycKwz- | ICLR.cc/2021/Conference | 2021 | Mapping the Timescale Organization of Neural Language Models | ["Hsiang-Yun Sherry Chien", "Jinhan Zhang", "Christopher Honey"] | In the human brain, sequences of language input are processed within a distributed and hierarchical architecture, in which higher stages of processing encode contextual information over longer timescales. In contrast, in recurrent neural networks which perform natural language processing, we know little about how the m... | ["natural language processing", "LSTM", "timescale", "hierarchy", "temporal context"] | ABSTRACTIn the human brain, sequences of language input are processed within a dis-tributed and hierarchical architecture, in which higher stages of processing en-code contextual information over longer timescales. In contrast, in recurrent neu-ral networks which perform natural language processing, we know little abou... | QxnC9gJQbz2 | Great to see some methods from neuroscience applied to interpretability research for a relevant question, results and setup could be improved | 3: Clear rejection | _**Update after author response**_: I think this is a very promising paper, and I am really excited about seeing techniques from neuroscience employed to answer questions about neural network models. The authors have further conducted several additional experiments after reviewer comments, which I appreciate. However, ... | 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
Mapping the Timescale Organization of Neural Language Models
### Paper Abstract
In the human brain, sequences of language input are processed within a distributed and hi... |
QAMjwGiU9pk | CUHK.edu.hk/2021/Course/IERG5350 | 2021 | Using Enhanced Gaussian Cross-Entropy in Imitation Learning to Digging the First Diamond in Minecraft | ["Yingjie CAI", "Xiao Zhang"] | Although state-ofthe-art reinforcement learning (RL) systems has led to breakthroughs in many difficult tasks, the sample inefficiency of standard reinforcement learning methods still precludes their application to more extremely complex tasks. Such limitation will make many reinforcement learning systems cannot be app... | ["reinforcement learning ObtainDiamond"] | Using Enhanced Gaussian Cross-Entropy in ImitationLearning to Digging the First Diamond in MinecraftYingjie Cai & Xiao ZhangDepartment of Electronic EngineeringThe Chinese University of Hong KongShatin, Hong Kong{caiyingjie,xzhang9411}@link.cuhk.edu.hkAbstractAlthough state-ofthe-art reinforcement learning (RL) systems... | RzdKbEWWN0s | Summary: This paper focuses on improving the performance of imitation learning problem. The main contribution of this paper is to use an enhanced Gaussian cross entropy loss for imitation learning problem based on the MineRL environment. | 7: Good paper, accept | Originality: This paper modifies cross entropy loss function and turns the imitation learning problem into a classification problem.
Quality: The paper's technical quality is good. Rewards are clearly defined and the environment settings are mentioned.
Clarity: This paper is clearly written and the sections are organi... | 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
Using Enhanced Gaussian Cross-Entropy in Imitation Learning to Digging the First Diamond in Minecraft
### Paper Abstract
Although state-ofthe-art reinforcement learning ... |
rkeeoeHYvr | ICLR.cc/2020/Conference | 2020 | AdvCodec: Towards A Unified Framework for Adversarial Text Generation | ["Boxin Wang", "Hengzhi Pei", "Han Liu", "Bo Li"] | Machine learning (ML) especially deep neural networks (DNNs) have been widely applied to real-world applications. However, recent studies show that DNNs are vulnerable to carefully crafted \emph{adversarial examples} which only deviate from the original data by a small magnitude of perturbation.
While there has been ... | ["adversarial text generation", "tree-autoencoder", "human evaluation"] | ABSTRACTWhile there has been great interest on generating imperceptible adversarial ex-amples in continuous data domain (e.g. image and audio) to explore the modelvulnerabilities, generating adversarial text in the discrete domain is still challeng-ing. The main contribution of this paper is to propose a general target... | HyeN5T9AtS | Official Blind Review #3 | 6: Weak Accept | This paper proposes a new attack framework AdvCodec for adversarial text generation. The main idea is to use a tree-based autoencoder to embed text data into the continuous vector space and then optimize to find the adversarial perturbation in the vector space. The authors consider two types of attacks: concat attack 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
AdvCodec: Towards A Unified Framework for Adversarial Text Generation
### Paper Abstract
Machine learning (ML) especially deep neural networks (DNNs) have been widely ap... | |
GANphlMvsa | NoDaLiDa/2023/Conference | 2023 | Is Part-of-Speech Tagging a Solved Problem for Icelandic? | ["\u00d6rvar K\u00e1rason", "Hrafn Loftsson"] | We train and evaluate four Part-of-Speech tagging models for Icelandic. Three are older models that obtained the highest accuracy for Icelandic when they were introduced. The fourth model is of a type that currently reaches state-of-the-art accuracy. We use the most recent version of the MIM-GOLD training/testing corpu... | ["Part-of-Speech Tagging", "Icelandic", "Transformer", "ConvBERT", "error analysis", "annotator disagreement", "annotation errors"] | Is Part-of-Speech Tagging a Solved Problem for Icelandic?Örvar KárasonDepartment of Computer ScienceReykjavik UniversityIcelandorvark13@ru.isHrafn LoftssonDepartment of Computer ScienceReykjavik UniversityIcelandhrafn@ru.isAbstractWe train and evaluate four Part-of-Speechtagging models for Icelandic. Three areolder m... | mTENbntPeRK | The authors of this paper train and test four data-driven POS-tagging models for Icelandic and conclude that, with very little additional effort, POS tagging could be called a solved problem for Icelandic. However, the text types of the gold standard are not described and so it is impossible to say, is the POS tagging ... | 7: Good paper, accept | The authors of this paper have trained and tested four data-driven POS taggers (TriTagger, IceStagger, ABLTagger and ConvBert) on a new large gold-standard corpus (ca 1 million tokens). The best model, ConvBERT-base achieves 97.79% token accuracy and 73.43% sentence accuracy. Basing on error analysis, the authors infer... | 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
Is Part-of-Speech Tagging a Solved Problem for Icelandic?
### Paper Abstract
We train and evaluate four Part-of-Speech tagging models for Icelandic. Three are older mode... |
6T45-4TFqaX | eswc-conferences.org/ESWC/2021/Conference/Research_Track | 2021 | Convolutional Complex Knowledge Graph Embeddings | ["Caglar Demir", "Axel-Cyrille Ngonga Ngomo"] | We investigate the problem of learning continuous vector representations of knowledge graphs for predicting missing links. Recent results suggest that using a Hermitian inner product on complex-valued embeddings or convolutions on real-valued embeddings can be effective means for predicting missing links. We bring thes... | ["Knowledge graph embeddings", "Convolutions", "Complex numbers", "Hermitian inner product"] | Convolutional Complex Knowledge GraphEmbeddingsCaglar Demir and Axel-Cyrille Ngonga NgomoData Science Research Group, Paderborn UniversityAbstract. We investigate the problem of learning continuous vectorrepresentations of knowledge graphs for predicting missing links. Recentresults suggest that using a Hermitian inner... | VzlSNxwfz6t | Novel extension to ComplEx with excellent presentation and evaluation | 2: Accept | The work presents an extension to the ComplEx graph embedding method that improves the shortcomings of ComplEx in terms of transitive relations.
The paper does an excellent job at summarizing related work and highlighting the differences with the proposed method. The motivation and design choices are clearly presented... | 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
Convolutional Complex Knowledge Graph Embeddings
### Paper Abstract
We investigate the problem of learning continuous vector representations of knowledge graphs for pred... |
k9GoaycDeio | ICLR.cc/2021/Conference | 2021 | Improving Local Effectiveness for Global Robustness Training | ["JINGYUE LU", "M. Pawan Kumar"] | Despite its increasing popularity, deep neural networks are easily fooled. To alleviate this deficiency, researchers are actively developing new training strategies, which encourage models that are robust to small input perturbations. Several successful robust training methods have been proposed. However, many of them ... | ["strong adversaries", "local effectiveness", "adversaries", "local balls", "mnist", "global robustness", "global robustness training", "popularity", "deep neural networks", "deficiency"] | ABSTRACTDespite its popularity, deep neural networks are easily fooled. To alleviate thisdeficiency, researchers are actively developing new training strategies, whichencourage models that are robust to small input perturbations. Several successfulrobust training methods have been proposed. However, many of them rely o... | fZjF5Wtz7QP | Good empirical results but there are concerns about the novelty and clarity of the provided justifications | 5: Marginally below acceptance threshold | **Summary:**
The paper proposes a new way to combine existing techniques for improving adversarial robustness: adversarial training, Jacobian regularization and TRADES consistency loss. The obtained results on three datasets are better than the baselines which rely on adversarial training, Jacobian regularization or TR... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Improving Local Effectiveness for Global Robustness Training
### Paper Abstract
Despite its increasing popularity, deep neural networks are easily fooled. To alleviate t... |
V69LGwJ0lIN | ICLR.cc/2021/Conference | 2021 | OPAL: Offline Primitive Discovery for Accelerating Offline Reinforcement Learning | ["Anurag Ajay", "Aviral Kumar", "Pulkit Agrawal", "Sergey Levine", "Ofir Nachum"] | Reinforcement learning (RL) has achieved impressive performance in a variety of online settings in which an agent’s ability to query the environment for transitions and rewards is effectively unlimited. However, in many practical applications, the situation is reversed: an agent may have access to large amounts of und... | ["Offline Reinforcement Learning", "Primitive Discovery", "Unsupervised Learning"] | ABSTRACTReinforcement learning (RL) has achieved impressive performance in a variety ofonline settings in which an agent’s ability to query the environment for transitionsand rewards is effectively unlimited. However, in many practical applications, thesituation is reversed: an agent may have access to large amounts of... | bDV_GCn84xx | Interesting research question but motivation of the method unclear | 6: Marginally above acceptance threshold | In the RL setting, this paper tackles the case where an agent may have access to large amounts of offline experience data. The objective of the work is to find an effective way to leverage this data for finding temporally extended primitive behaviors. The paper provides results that show how performing offline primitiv... | 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
OPAL: Offline Primitive Discovery for Accelerating Offline Reinforcement Learning
### Paper Abstract
Reinforcement learning (RL) has achieved impressive performance in a... |
8VUywK1AT7d | thecvf.com/ECCV/2022/Workshop/VIPriors | 2022 | GraphVid: It Only Takes a Few Nodes to Understand a Video | ["Eitan Kosman", "Dotan Di Castro"] | We propose a concise representation of videos that encode perceptually meaningful features into graphs. With this representation, we aim to leverage the large amount of redundancies in videos and save computations. First, we construct superpixel-based graph representations of videos by considering superpixels as graph ... | ["videos", "nodes", "video", "representation", "graphvid", "video graphvid", "concise representation", "meaningful features", "graphs", "large amount"] | 000001002003004005006007008009010011012013014015016017018019020021022023024025026027028029030031032033034035036037038039040041042043044000001002003004005006007008009010011012013014015016017018019020021022023024025026027028029030031032033034035036037038039040041042043044ECCV#4861ECCV#4861GraphVid : It Only Takes a Few N... | -AYhRTcWDe5 | The idea presented in the paper is simple but can effectively speed up action recognition, therefore the paper should be accepted. | 6: Marginally above acceptance threshold | Summary:
- The authors propose an efficient graph video representation, GraphVid, that can be used for action recognition with reduced time and memory requirements. GraphVid results in a large efficiency gain without decreasing the performance.
Positive points:
+ The idea presented in the paper is interesting and ... | 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
GraphVid: It Only Takes a Few Nodes to Understand a Video
### Paper Abstract
We propose a concise representation of videos that encode perceptually meaningful features i... |
BVPowUU1cR | ICLR.cc/2021/Conference | 2021 | Assisting the Adversary to Improve GAN Training | ["Andreas Munk", "William Harvey", "Frank Wood"] | Some of the most popular methods for improving the stability and performance of GANs involve constraining or regularizing the discriminator. In this paper we consider a largely overlooked regularization technique which we refer to as the Adversary's Assistant (AdvAs). We motivate this using a different perspective to t... | ["Generative Adversarial Networks", "GANs"] | ABSTRACTSome of the most popular methods for improving the stability and performanceof GANs involve constraining or regularizing the discriminator. In this paper weconsider a largely overlooked regularization technique which we refer to as theAdversary’s Assistant (AdvAs). We motivate this using a different perspective... | l83gld8C01 | review | 6: Marginally above acceptance threshold | This paper proposes a new regularizer to improve GAN training. By noticing that the discriminator does not always reach optimum at each iteration, this paper proposes Adversary's Assistant (AdvAs) for helping the discriminator to satisfy this condition. Interestingly, compared to the previous methods for improving GAN ... | 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
Assisting the Adversary to Improve GAN Training
### Paper Abstract
Some of the most popular methods for improving the stability and performance of GANs involve constrain... |
Sym_tDJwM | ICLR.cc/2018/Workshop | 2018 | TVAE: Triplet-Based Variational Autoencoder using Metric Learning | ["Haque Ishfaq", "Assaf Hoogi", "Daniel Rubin"] | Deep metric learning has been demonstrated to be highly effective in learning semantic representation and encoding information that can be used to measure data similarity, by relying on the embedding learned from metric learning. At the same time, variational autoencoder (VAE) has widely been used to approximate infere... | ["Metric Learning", "Variational Autoencoder", "Representation Learning", "Deep learning", "Semi-Supervised Learning"] | ABSTRACTDeep metric learning has been demonstrated to be highly effective in learningsemantic representation and encoding information that can be used to measuredata similarity, by relying on the embedding learned from metric learning. Atthe same time, variational autoencoder (V AE) has widely been used to approxi-mate... | B1hgWa-Fz | A possibly interesting addition to VAE. Experiments are insufficient. | 3: Clear rejection | The paper proposes adding a supervised signal to VAE signal which is meant to improve the usability of the latent space vectors. Specifically, the proposal is a triplet-based loss first used by Frome et al. (2007), which has since been widely used in metric learning (example Weinberger & Saul 2009). This loss relies o... | 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
TVAE: Triplet-Based Variational Autoencoder using Metric Learning
### Paper Abstract
Deep metric learning has been demonstrated to be highly effective in learning semant... |
B1gd0nEFwS | ICLR.cc/2020/Conference | 2020 | Universal Source-Free Domain Adaptation | ["Jogendra Nath Kundu", "Naveen Venkat", "Rahul M V", "R. Venkatesh Babu"] | There is a strong incentive to develop versatile learning techniques that can transfer the knowledge of class-separability from a labeled source domain to an unlabeled target domain in the presence of a domain-shift. Existing domain adaptation (DA) approaches are not equipped for practical DA scenarios as a result of t... | ["unsupervised domain adaptation", "knowledge transfer", "source-free adaptation"] | ABSTRACTThere is a strong incentive to develop versatile learning techniques that can transferthe knowledge of class-separability from a labeled source domain to an unlabeledtarget domain in the presence of a domain-shift. Existing domain adaptation(DA) approaches are not equipped for practical DA scenarios as a result... | B1lOc2F5OS | Official Blind Review #3 | 3: Weak Reject | The paper proposes a domain adaptation problem with the source domain not available in addition to previous universal domain adaptation. The paper employs a two-stage training procedure to solve the problem.
In the first procurement stage, the paper aims to extend the source classification boundary to out-of-domain da... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Universal Source-Free Domain Adaptation
### Paper Abstract
There is a strong incentive to develop versatile learning techniques that can transfer the knowledge of class-... | |
ByJHuTgA- | ICLR.cc/2018/Conference | 2018 | On the State of the Art of Evaluation in Neural Language Models | ["G\u00e1bor Melis", "Chris Dyer", "Phil Blunsom"] | Ongoing innovations in recurrent neural network architectures have provided a steady influx of apparently state-of-the-art results on language modelling benchmarks. However, these have been evaluated using differing codebases and limited computational resources, which represent uncontrolled sources of experimental vari... | ["rnn", "language modelling"] | ABSTRACTOngoing innovations in recurrent neural network architectures have provided asteady influx of apparently state-of-the-art results on language modelling bench-marks. However, these have been evaluated using differing codebases and limitedcomputational resources, which represent uncontrolled sources of experiment... | HkGW8A2gG | Important big-picture work in a fast-moving field | 8: Top 50% of accepted papers, clear accept | The authors perform a comprehensive validation of LSTM-based word and character language models, establishing that recent claims that other structures can consistently outperform the older stacked LSTM architecture result from failure to fully explore the hyperparameter space. Instead, with more thorough hyperparameter... | 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
On the State of the Art of Evaluation in Neural Language Models
### Paper Abstract
Ongoing innovations in recurrent neural network architectures have provided a steady i... |
GHCu1utcBvX | ICLR.cc/2021/Conference | 2021 | Transferability of Compositionality | ["Yuanpeng Li", "Liang Zhao", "Joel Hestness", "Ka Yee Lun", "Kenneth Church", "Mohamed Elhoseiny"] | Compositional generalization is the algebraic capacity to understand and produce large amount of novel combinations from known components. It is a key element of human intelligence for out-of-distribution generalization. To equip neural networks with such ability, many algorithms have been proposed to extract compositi... | ["Compositionality"] | ABSTRACTCompositional generalization is the algebraic capacity to understand and producelarge amount of novel combinations from known components. It is a key element ofhuman intelligence for out-of-distribution generalization. To equip neural networkswith such ability, many algorithms have been proposed to extract comp... | xGHcvStyzNG | the paper considers an interesting general problem, but the concrete supervised learning instantiation is problematic | 4: Ok but not good enough - rejection | The paper introduces a “transferability of compositionality” problem and proposes an approach to alleviate it. The said problem may arise when one trains neural models to produce “compositional” representations of the input. In the paper “compositional representations” consist of multiple vectors which are supposed to ... | 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
Transferability of Compositionality
### Paper Abstract
Compositional generalization is the algebraic capacity to understand and produce large amount of novel combination... |
NTEz-6wysdb | ICLR.cc/2021/Conference | 2021 | Distilling Knowledge from Reader to Retriever for Question Answering | ["Gautier Izacard", "Edouard Grave"] | The task of information retrieval is an important component of many natural language processing systems, such as open domain question answering. While traditional methods were based on hand-crafted features, continuous representations based on neural networks recently obtained competitive results. A challenge of using ... | ["question answering", "information retrieval"] | ABSTRACTThe task of information retrieval is an important component of many natural lan-guage processing systems, such as open domain question answering. While tra-ditional methods were based on hand-crafted features, continuous representationsbased on neural networks recently obtained competitive results. A challenge ... | PWS7IuRysr | Interesting technique | 6: Marginally above acceptance threshold | The authors propose a training technique for information retrieval models in the context of (open domain) question answering. Assuming the existence of some reader model, the idea is to use internal information of that model as a training signal for a retriever. Specifically, they use the attention activations over the... | 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
Distilling Knowledge from Reader to Retriever for Question Answering
### Paper Abstract
The task of information retrieval is an important component of many natural langu... |
PcUprce4TM2 | ICLR.cc/2021/Conference | 2021 | CAFE: Catastrophic Data Leakage in Federated Learning | ["Xiao Jin", "Ruijie Du", "Pin-Yu Chen", "Tianyi Chen"] | Private training data can be leaked through the gradient sharing mechanism deployed in machine learning systems, such as federated learning (FL).
Increasing batch size is often viewed as a promising defense strategy against data leakage. In this paper, we revisit this defense premise and propose an advanced data leakag... | ["cafe", "catastrophic data leakage", "aggregated gradients", "federated", "gradient", "mechanism", "machine", "systems"] | ABSTRACTPrivate training data can be leaked through the gradient sharing mechanism de-ployed in machine learning systems, such as federated learning (FL). Increasingbatch size is often viewed as a promising defense strategy against data leakage.In this paper, we revisit this defense premise and propose an advanced data... | -b7PWDT8np | Simple, but unclear. | 4: Ok but not good enough - rejection | This work introduces CAFE, a novel training algorithm to leak training data in a federated learning setup. Extending from "deep leakage from gradient" fake images are optimised with respect to the difference observed from the client gradients (i.e. with the real images) and the one observed with the current version of ... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
CAFE: Catastrophic Data Leakage in Federated Learning
### Paper Abstract
Private training data can be leaked through the gradient sharing mechanism deployed in machine l... |
Skltqh4KvB | ICLR.cc/2020/Conference | 2020 | Are there any 'object detectors' in the hidden layers of CNNs trained to identify objects or scenes? | ["Ella M. Gale", "Nicholas Martin", "Ryan Blything", "Anh Nguyen", "Jeffrey S. Bowers"] | Various methods of measuring unit selectivity have been developed with the aim of better understanding how neural networks work. But the different measures provide divergent estimates of selectivity, and this has led to different conclusions regarding the conditions in which selective object representations are learne... | ["neural networks", "localist coding", "selectivity", "object detectors", "CCMAS", "CNNs", "activation maximisation", "information representation", "network dissection", "interpretabillity", "signal detection"] | ABSTRACTVarious methods of measuring unit selectivity have been developed with the aimof better understanding how neural networks work. But the different measuresprovide divergent estimates of selectivity, and this has led to different conclusionsregarding the conditions in which selective object representations are le... | B1e6QO3SKH | Official Blind Review #2 | 3: Weak Reject | The paper empirically studies the category selectivity of individual cells in hidden units of CNNs. It is a sort of "meta-study" and comparison of different metrics proposed to identify cells with a preference for a specific target category. The claimed finding is that there are no cells that are "sufficiently" selecti... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Are there any 'object detectors' in the hidden layers of CNNs trained to identify objects or scenes?
### Paper Abstract
Various methods of measuring unit selectivity hav... | |
HygQro05KX | ICLR.cc/2019/Conference | 2019 | $A^*$ sampling with probability matching | ["Yichi Zhou", "Jun Zhu"] | Probabilistic methods often need to draw samples from a nontrivial distribution. $A^*$ sampling is a nice algorithm by building upon a top-down construction of a Gumbel process, where a large state space is divided into subsets and at each round $A^*$ sampling selects a subset to process. However, the selection rule de... | ["probability", "subsets", "probabilistic methods", "samples", "nontrivial distribution", "sampling", "nice algorithm", "construction", "gumbel process", "large state space"] | ABSTRACTProbabilistic methods often need to draw samples from a nontrivial distribution.Asampling is a nice algorithm by building upon a top-down construction of aGumbel process, where a large state space is divided into subsets and at each roundAsampling selects a subset to process. However, the selection rule depends... | HyxV1doKhX | Strong contribution to family of A* sampling algorithms, but lacks clarity. | 5: Marginally below acceptance threshold | Summary: This paper introduces a probability matching approach for optimizing Gumbel processes, i.e. the extension of the Gumbel-Max trick to more general measure spaces. The basic idea is to use a more refined subset selection mechanism as compared to A* Sampling, but at the cost of being able to guarantee an exact sa... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
$A^*$ sampling with probability matching
### Paper Abstract
Probabilistic methods often need to draw samples from a nontrivial distribution. $A^*$ sampling is a nice alg... |
0l3buSHSJj0 | IEEE.org/2022/Workshop/altVIS | 2022 | Maybe, Maybe Not: A Survey on Uncertainty in Visualization | ["Anonymous"] | Understanding and evaluating uncertainty play a key role in decision-making. When a viewer studies a visualization that demands inference, it is necessary that uncertainty is portrayed in it. This paper showcases the importance of representing uncertainty in visualizations. It provides an overview of uncertainty visual... | ["Uncertainty", "Data Visualization"] | ABSTRACTUnderstanding and evaluating uncertainty play a key role in decision-making. When a viewer studies a visualization that demands infer-ence, it is necessary that uncertainty is portrayed in it. This papershowcases the importance of representing uncertainty in visualiza-tions. It provides an overview of uncertain... | CUoynUKxtEj | This paper does a good job of outlining the field of uncertainty visualization and explaining the challenges of the field. It doesn't, however, feel provocative or alternative in any manner. Its only "joke" value comes from the author using xkcd strips to illustrate the points. | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Maybe, Maybe Not: A Survey on Uncertainty in Visualization
### Paper Abstract
Understanding and evaluating uncertainty play a key role in decision-making. When a viewer ... | |||
MZPGZsuwqr | NeurIPS.cc/2021/Workshop/DLDE | 2021 | A neural multilevel method for high-dimensional parametric PDEs | ["Cosmas Hei\u00df", "Ingo G\u00fchring", "Martin Eigel"] | In scientific machine learning, neural networks recently have become a popular tool for learning the solutions of differential equations.
However, practical results often conflict the existing theoretical predictions in that observed convergence stagnates early. A substantial improvement can be achieved by the presente... | ["scientific machine learning", "deep learning", "partial differential equations", "multigrid methods", "uncertainty quantification"] | A neural multilevel method for high-dimensionalparametric PDEsCosmas HeißTechnical University of Berlincosmas.heiss@gmail.comIngo GühringTechnical University of Berlinguehring@math.tu-berlin.deMartin EigelWeierstraß Institutemartin.eigel@wias-berlin.deAbstractIn scientific machine learning, neural networks recently ha... | Xd187ZLOMc | Solid works | I think this article is written well. It contains both the empirical
study and the theoretical analysis, which are quite solid. The proposed method used neural
network to learn the basic coefficients in the multi-level scheme,
which works well for a stationary parametric diffusion problems.
To further demonstrate the f... | 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 neural multilevel method for high-dimensional parametric PDEs
### Paper Abstract
In scientific machine learning, neural networks recently have become a popular tool fo... | |
Byl1e1BFPH | ICLR.cc/2020/Conference | 2020 | Dual Sequential Monte Carlo: Tunneling Filtering and Planning in Continuous POMDPs | ["Yunbo Wang", "Bo Liu", "Jiajun Wu", "Yuke Zhu", "Simon Shaolei Du", "Li Fei-Fei", "Joshua B. Tenenbaum"] | We present the DualSMC network that solves continuous POMDPs by learning belief representations and then leveraging them for planning. It is based on the fact that filtering, i.e. state estimation, and planning can be viewed as two related sequential Monte Carlo processes, with one in the belief space and the other in ... | ["planning", "tunneling filtering", "continuous pomdps", "belief representations", "dualsmc network", "fact", "state estimation"] | ABSTRACTWe present the DualSMC network that solves continuous POMDPs by learningbelief representations and then leveraging them for planning. It is based on thefact that filtering, i.e. state estimation, and planning can be viewed as two relatedsequential Monte Carlo processes, with one in the belief space and the othe... | H1gXKDdpKS | Official Blind Review #3 | 1: Reject | The contribution of the paper is a method of filtering and planning as joint probabilistic inference. For that, the duality of inference and control is used, such that an adversarial particle filtering approch can be employed. Several experiments that support the claim are presented.
I recommend to reject the paper fr... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Dual Sequential Monte Carlo: Tunneling Filtering and Planning in Continuous POMDPs
### Paper Abstract
We present the DualSMC network that solves continuous POMDPs by lea... | |
BTzgpgtNaGq | graphicsinterface.org/Graphics_Interface/2022/Conference | 2022 | Comparison of a VR Stylus with a Controller, Hand Tracking and a Mouse for Object Manipulation and Medical Marking Tasks in Virtual Reality | ["Anonymous"] | For medical surgery planning, virtual reality (VR) provides a new kind of user experience, where 3D images of the operation area can be utilized. Using VR, it is possible to view the 3D models in a more realistic 3D environment, which would reduce the perception problems and increase spatial understanding. In the prese... | ["3D visualization", "virtual reality", "hand tracking", "controller interaction", "mouse interaction", "3D imaging"] | ABSTRACTFor medical surgery planning, virtual reality (VR) provides a newkind of user experience, where 3D images of the operation area canbe utilized. Using VR, it is possible to view the 3D models in amore realistic 3D environment, which would reduce the perceptionproblems and increase spatial understanding. In the p... | hC-4PGvv9Oq | some weaknesses with the study | 5: Marginally below acceptance threshold | I reviewed an earlier version of this submission. It has been improved by adding a video, which is very helpful to understand the interaction techniques, and also by citing more related work. There is some limited justification given in section 1 (paragraphs 4 and 5) for the choice of interaction methods. I don't fi... | 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
Comparison of a VR Stylus with a Controller, Hand Tracking and a Mouse for Object Manipulation and Medical Marking Tasks in Virtual Reality
### Paper Abstract
For medica... |
TOzBiasXxyX | eswc-conferences.org/ESWC/2021/Conference/Resources_Track | 2021 | CSKG: The CommonSense Knowledge Graph | ["Filip Ilievski", "Pedro Szekely", "Bin Zhang"] | Sources of commonsense knowledge aim to support applications in natural language understanding, computer vision, and knowledge graphs. These sources contain complementary knowledge to each other, which makes their integration desired. Yet, such integration is not trivial because of their different foci, modeling approa... | ["Commonsense Knowledge", "Knowledge graphs", "Embeddings"] | CSKG: The CommonSense Knowledge GraphFilip Ilievski, Pedro Szekely, and Bin ZhangInformation Sciences Institute, University of Southern Californiafilievski,pszekely,binzhang g@isi.eduAbstract. Sources of commonsense knowledge support applications innatural language understanding, computer vision, and knowledge graphs.G... | vHZTInipu39 | Paper 37 review - Updated after rebuttal | 1: Weak Accept | The paper describes CSKG, the first integrated knowledge graph that connects several common sense knowledge graphs. I think the paper takes an interesting research direction and the work has the potential to benefit the community. However, the paper has several weakness and does not fully articulate the benefits of the... | 2: The reviewer is willing to defend the evaluation but not sufficiently familiar with the state of the art or the specific topic 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
CSKG: The CommonSense Knowledge Graph
### Paper Abstract
Sources of commonsense knowledge aim to support applications in natural language understanding, computer vision,... |
eom0IUrF__F | ICLR.cc/2021/Conference | 2021 | CoCo: Controllable Counterfactuals for Evaluating Dialogue State Trackers | ["SHIYANG LI", "Semih Yavuz", "Kazuma Hashimoto", "Jia Li", "Tong Niu", "Nazneen Rajani", "Xifeng Yan", "Yingbo Zhou", "Caiming Xiong"] | Dialogue state trackers have made significant progress on benchmark datasets, but their generalization capability to novel and realistic scenarios beyond the held- out conversations is less understood. We propose controllable counterfactuals (COCO) to bridge this gap and evaluate dialogue state tracking (DST) models on... | ["task-oriented dialogue", "dialogue state tracking", "robustness", "dst", "evaluation"] | ABSTRACTDialogue state trackers have made significant progress on benchmark datasets, buttheir generalization capability to novel and realistic scenarios beyond the held-out conversations is less understood. We propose controllable counterfactuals(COCO) to bridge this gap and evaluate dialogue state tracking (DST) mode... | CvZ6d0LROU | Creating challenging examples for DST systems: interesting results but may need more clarification in the methods section | 6: Marginally above acceptance threshold | Update after reading author response:
Thank you for the thoughtful response. I appreciate that you have added extra implementation details and am changing my score to a 6.
Regarding the limitations in scope: My apologies if my review was confusingly worded. I just wanted to clarify that I had meant that the coun... | 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
CoCo: Controllable Counterfactuals for Evaluating Dialogue State Trackers
### Paper Abstract
Dialogue state trackers have made significant progress on benchmark datasets... |
SkgRW64twr | ICLR.cc/2020/Conference | 2020 | Deep Multi-View Learning via Task-Optimal CCA | ["Heather D. Couture", "Roland Kwitt", "J.S. Marron", "Melissa Troester", "Charles M. Perou", "Marc Niethammer"] | Canonical Correlation Analysis (CCA) is widely used for multimodal data analysis and, more recently, for discriminative tasks such as multi-view learning; however, it makes no use of class labels. Recent CCA methods have started to address this weakness but are limited in that they do not simultaneously optimize the C... | ["multi-view", "components analysis", "CCA", "representation learning", "deep learning"] | ABSTRACTMulti-view learning seeks to form better models by making use of multiple featuresets representing the same samples. Exploiting feature correlations during train-ing can enable better models. The traditional method for computing correlationsbetween feature sets is Canonical Correlation Analysis (CCA), which fin... | SJl61Mt2FS | Official Blind Review #2 | 3: Weak Reject | Originality:
CCA is a generative model that learns a shared subspace based on two (or multi) views of the data. Being generative, it might not have strong discriminative power for some downstream classification tasks. Previous approaches to infuse discriminative power into the shared subspace estimated by CCA are lin... | <|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 Multi-View Learning via Task-Optimal CCA
### Paper Abstract
Canonical Correlation Analysis (CCA) is widely used for multimodal data analysis and, more recently, for... | |
H1kjdOYlx | ICLR.cc/2017/conference | 2017 | Modular Multitask Reinforcement Learning with Policy Sketches | ["Jacob Andreas", "Dan Klein", "Sergey Levine"] | We describe a framework for multitask deep reinforcement learning guided by
policy sketches. Sketches annotate each task with a sequence of named subtasks,
providing high-level structural relationships among tasks, but not providing the
detailed guidance required by previous work on learning policy abstractions for
RL ... | ["Reinforcement Learning", "Transfer Learning"] | ABSTRACTWe describe a framework for multitask deep reinforcement learning guided bypolicy sketches . Sketches annotate each task with a sequence of named subtasks,providing high-level structural relationships among tasks, but notproviding thedetailed guidance required by previous work on learning policy abstractions fo... | SyJDbsb4e | Review | 3: Clear rejection | The paper proposes a new RL architecture that aims at learning policies from sketches i.e sequence of high-level operations to execute for solving a particular task. The model relies on a hierarchical structure where the sub-policy is chosen depending on the current operation to execute in the sketch . The learning alg... | 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
Modular Multitask Reinforcement Learning with Policy Sketches
### Paper Abstract
We describe a framework for multitask deep reinforcement learning guided by policy sketc... |
r1Chut9xl | ICLR.cc/2017/conference | 2017 | Inference and Introspection in Deep Generative Models of Sparse Data | ["Rahul G. Krishnan", "Matthew Hoffman"] | Deep generative models such as deep latent Gaussian models (DLGMs) are powerful and popular density estimators. However, they have been applied almost exclusively to dense data such as images; DLGMs are rarely applied to sparse, high-dimensional integer data such as word counts or product ratings. One reason is that th... | ["Unsupervised Learning", "Deep learning"] | ABSTRACTDeep generative models such as deep latent Gaussian models (DLGMs) are pow-erful and popular density estimators. However, they have been applied almostexclusively to dense data such as images; DLGMs are rarely applied to sparse,high-dimensional integer data such as word counts or product ratings. One reasonis t... | H1Dc01zNe | Decent paper, but lacking novelty | 5: Marginally below acceptance threshold | This paper introduces three tricks for training deep latent variable models on sparse discrete data:
1) tf-idf weighting
2) Iteratively optimizing variational parameters after initializing them with an inference network
3) A technique for improving the interpretability of the deep model
The first idea is sensible but ... | 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
Inference and Introspection in Deep Generative Models of Sparse Data
### Paper Abstract
Deep generative models such as deep latent Gaussian models (DLGMs) are powerful a... |
BJyy3a0Ez | ICLR.cc/2018/Conference | 2018 | Continuous Propagation: Layer-Parallel Training | ["Michael James", "Devansh Arpit", "Herman Sahota", "Ilya Sharapov"] | Continuous propagation is a parallel technique for training deep neural networks with batch size one at full utilization of a multiprocessor system. It enables spatially distributed computations on emerging deep learning hardware accelerators that do not impose programming limitations of contemporary GPUs. The algorith... | ["Deep Learning", "Model parallelism", "Learning theory"] | ABSTRACTContinuous propagation is a parallel technique for training deep neural networkswith batch size one at full utilization of a multiprocessor system. It enables spa-tially distributed computations on emerging deep learning hardware acceleratorsthat do not impose programming limitations of contemporary GPUs. The a... | H1YbqHUBG | Promising idea, but needs more demonstration of practicality | 5: Marginally below acceptance threshold | Pros:
* asynchronous model-parallel training of deep learning models would potentially help in further scaling of deep learning models
* the paper is clearly written and easy to understand
Cons:
* weak experiments: performance of algorithms are not analyzed in terms of wall-clock, and important baselines are not compa... | 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
Continuous Propagation: Layer-Parallel Training
### Paper Abstract
Continuous propagation is a parallel technique for training deep neural networks with batch size one a... |
SR2L__h9q9p | ICML.cc/2020/Workshop/SAS | 2020 | Investigating Self-supervised Pre-training for End-to-end Speech Translation | ["Ha Nguyen", "Fethi Bougares", "Natalia Tomashenko", "Yannick Est\u00e8ve", "laurent besacier"] | Self-supervised learning from raw speech has been proven beneficial to improve automatic speech recognition (ASR). We investigate here its impact on end-to-end automatic speech translation (AST) performance. We use a contrastive predictive coding (CPC) model pre-trained from unlabeled speech as a feature extractor for ... | ["self-supervised learning from speech", "automatic speech translation", "end-to-end models", "low resource settings."] | Investigating Self-supervised Pre-training for End-to-end Speech TranslationHa Nguyen1 2Fethi Bougares3Natalia Tomashenko2Yannick Est `eve2Laurent Besacier1AbstractSelf-supervised learning from raw speech hasbeen proven beneficial to improve automaticspeech recognition (ASR). We investigate hereits impact on end-to-end... | JpqRX8tdMGK | Good paper and detailed analysis on the use of CPC model in end2end speech translation | 8: Top 50% of accepted papers, clear accept | This paper investigates the application of self-supervised pre-training to end-to-end speech translation. In particular, it presents an application of the contrastive predictive coding (CPC) model that gives the possibility of leveraging massive audio data without the need of the textual transcriptions. An extensive se... | 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
Investigating Self-supervised Pre-training for End-to-end Speech Translation
### Paper Abstract
Self-supervised learning from raw speech has been proven beneficial to im... |
zM6fevLxIhI | ICLR.cc/2021/Conference | 2021 | Variational Structured Attention Networks for Dense Pixel-Wise Prediction | ["Guanglei Yang", "Paolo Rota", "Xavier Alameda-Pineda", "Dan Xu", "Mingli Ding", "Elisa Ricci"] | State-of-the-art performances in dense pixel-wise prediction tasks are obtained with specifically designed convolutional networks. These models often benefit from attention mechanisms that allow better learning of deep representations. Recent works showed the importance of estimating both spatial- and channel-wise atte... | ["attention network", "pixel-wise prediction"] | ABSTRACTState-of-the-art performances in dense pixel-wise prediction tasks are obtainedwith specifically designed convolutional networks. These models often benefitfrom attention mechanisms that allow better learning of deep representations. Re-cent works showed the importance of estimating both spatial- and channel-wi... | S01pt94K-nR | Recommendation to Reject | 5: Marginally below acceptance threshold | ##########################################################################
Summary:
This paper proposes the VarIational STructured Attention networks (VISTA-Net), which improves pervious SOTA models for dense pixel-wise prediction tasks. The proposed VISTA-Net is featured by two aspects: 1) A new structured attention... | 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
Variational Structured Attention Networks for Dense Pixel-Wise Prediction
### Paper Abstract
State-of-the-art performances in dense pixel-wise prediction tasks are obtai... |
HyxPx3R9tm | ICLR.cc/2019/Conference | 2019 | Variational Discriminator Bottleneck: Improving Imitation Learning, Inverse RL, and GANs by Constraining Information Flow | ["Xue Bin Peng", "Angjoo Kanazawa", "Sam Toyer", "Pieter Abbeel", "Sergey Levine"] | Adversarial learning methods have been proposed for a wide range of applications, but the training of adversarial models can be notoriously unstable. Effectively balancing the performance of the generator and discriminator is critical, since a discriminator that achieves very high accuracy will produce relatively uninf... | ["reinforcement learning", "generative adversarial networks", "imitation learning", "inverse reinforcement learning", "information bottleneck"] | ABSTRACTAdversarial learning methods have been proposed for a wide range of applications,but the training of adversarial models can be notoriously unstable. Effectively bal-ancing the performance of the generator and discriminator is critical, since a dis-criminator that achieves very high accuracy will produce relativ... | Byl41tz9nX | a constraint on the discriminator of GAN model to maintain informative gradients | 6: Marginally above acceptance threshold | This paper proposed a constraint on the discriminator of GAN model to maintain informative gradients. It is completed by control the mutual information between the observations and the discriminator’s internal representation to be no bigger than a predefined value. The idea is interesting and the discussions of applic... | 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
Variational Discriminator Bottleneck: Improving Imitation Learning, Inverse RL, and GANs by Constraining Information Flow
### Paper Abstract
Adversarial learning methods... |
3YdNZD5dMxI | ICLR.cc/2021/Conference | 2021 | Unconditional Synthesis of Complex Scenes Using a Semantic Bottleneck | ["Samaneh Azadi", "Michael Tschannen", "Eric Tzeng", "Sylvain Gelly", "Trevor Darrell", "Mario Lucic"] | Coupling the high-fidelity generation capabilities of label-conditional image synthesis methods with the flexibility of unconditional generative models, we propose a semantic bottleneck GAN model for unconditional synthesis of complex scenes. We assume pixel-wise segmentation labels are available during training and us... | ["Unconditional Image Synthesis", "Complex Scene", "GAN", "Semantic Bottleneck"] | ABSTRACTCoupling the high-fidelity generation capabilities of label-conditional image syn-thesis methods with the flexibility of unconditional generative models, we proposea semantic bottleneck GAN model for unconditional synthesis of complex scenes.We assume pixel-wise segmentation labels are available during training... | aF-dixAE0H | Reasonable approach but validation is unconvincing | 4: Ok but not good enough - rejection | The paper suggests a new approach for unconditional generation of complex scenes. The approach performs generation in two steps: first a semantic map is generated from noise using a conventional generator architecture, then the semantic image is turner into an RGB image by SPADE translator.
The paper has several stren... | 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
Unconditional Synthesis of Complex Scenes Using a Semantic Bottleneck
### Paper Abstract
Coupling the high-fidelity generation capabilities of label-conditional image sy... |
B1xFVhActm | ICLR.cc/2019/Conference | 2019 | Fake Sentence Detection as a Training Task for Sentence Encoding | ["Viresh Ranjan", "Heeyoung Kwon", "Niranjan Balasubramanian", "Minh Hoai"] | Sentence encoders are typically trained on generative language modeling tasks with large unlabeled datasets. While these encoders achieve strong results on many sentence-level tasks, they are difficult to train with long training cycles.
We introduce fake sentence detection as a new discriminative training tas... | ["fake sentence detection", "sentence encoders", "training task", "sentence", "tasks", "encoders", "fake sentences", "task", "generative language", "large unlabeled datasets"] | ABSTRACTSentence encoders are typically trained on generative language modeling taskswith large unlabeled datasets. While these encoders achieve strong results onmany sentence-level tasks, they are difficult to train with long training cycles. Weintroduce fake sentence detection as a new discriminative training task fo... | B1e6e0PInX | Method description confusing; empirical comparison against previous work is lacking | 3: Clear rejection | This paper proposes a method for learning sentences encoders using artificially generated (fake) sentences. While the idea is interesting, the paper has the following issues:
- There are other methods that aim at generating artificial training data, e.g.: Z. Zhao, D. Dua, S. Singh. Generating Natural Adversarial Exam... | 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
Fake Sentence Detection as a Training Task for Sentence Encoding
### Paper Abstract
Sentence encoders are typically trained on generative language modeling tasks wit... |
rJZ6RMTIG | ICLR.cc/2018/Workshop | 2018 | An Empirical Study of Weights in Deep Convolutional Neural Networks and Its Application to Training Convergence | ["Haihao Shen", "Jiong Gong", "Jianhui Li", "Xiaoli Liu", "Xinan Lin"] | This paper presents an empirical study of weights in deep neural networks and propose a quantitative metric, Logarithmical Geometric Mean of absolute weight parameter (LoGM), to evaluate the impact of weight on training convergence. We develop an automatic tool to measure LoGM and conduct extensive experiments on Image... | ["Deep Neural Networks", "Quantitative Metric", "Convergence", "Divergence Region"] | ABSTRACTThis paper presents an empirical study of weights in deep neural networks andpropose a quantitative metric, Logarithmical Geometric Mean of absolute weightparameter (LoGM), to evaluate the impact of weight on training convergence. Wedevelop an automatic tool to measure LoGM and conduct extensive experimentson I... | S18Or9lFG | Quantitative metric to detect convergence problems in CNN training | 5: Marginally below acceptance threshold | This paper proposes a metric called LoGM to evaluate the evolution of the CNN weight during the training. The authors claim that LoGM can be used to detect convergence problems in the CNN training.
The paper is confusing and hard to understand. There are different aspects that are unclear. For example:
- Authors sta... | 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
An Empirical Study of Weights in Deep Convolutional Neural Networks and Its Application to Training Convergence
### Paper Abstract
This paper presents an empirical study... |
_ptUyYP19mP | ICLR.cc/2021/Conference | 2021 | BeBold: Exploration Beyond the Boundary of Explored Regions | ["Tianjun Zhang", "Huazhe Xu", "Xiaolong Wang", "Yi Wu", "Kurt Keutzer", "Joseph E. Gonzalez", "Yuandong Tian"] | Efficient exploration under sparse rewards remains a key challenge in deep reinforcement learning. To guide exploration, previous work makes extensive use of intrinsic reward (IR). There are many heuristics for IR, including visitation counts, curiosity, and state-difference. In this paper, we analyze the pros and cons... | ["reinforcement learning", "exploration"] | ABSTRACTEfficient exploration under sparse rewards remains a key challenge in deep rein-forcement learning. To guide exploration, previous work makes extensive use ofintrinsic reward (IR). There are many heuristics for IR, including visitation counts,curiosity, and state-difference. In this paper, we analyze the pros a... | pQkE-Uw2aIn | This paper is overall well-written, motivated and empirically-supported. | 8: Top 50% of accepted papers, clear accept | ------------------------------------
**Summary:**
This paper proposes BeBold, a new definition of intrinsic reward to guide exploration in sparse reward problems. This intrinsic reward combines the ideas behind count-based approaches and state-diff approaches. They demonstrate the success of BeBold by comparin... | 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
BeBold: Exploration Beyond the Boundary of Explored Regions
### Paper Abstract
Efficient exploration under sparse rewards remains a key challenge in deep reinforcement l... |
8YFhXYe1Ps | ICLR.cc/2021/Conference | 2021 | Interpretability Through Invertibility: A Deep Convolutional Network With Ideal Counterfactuals And Isosurfaces | ["Leon Sixt", "Martin Schuessler", "Philipp Wei\u00df", "Tim Landgraf"] | Current state of the art computer vision applications rely on highly complex models. Their interpretability is mostly limited to post-hoc methods which are not guaranteed to be faithful to the model. To elucidate a model’s decision, we present a novel interpretable model based on an invertible deep convolutional networ... | ["Interpretable Machine Learning", "Counterfactuals", "Computer Vision", "Human Evaluation", "User Study"] | ABSTRACTCurrent state of the art computer vision applications rely on highly complex mod-els. Their interpretability is mostly limited to post-hoc methods which are notguaranteed to be faithful to the model. To elucidate a model’s decision, wepresent a novel interpretable model based on an invertible deep convolutional... | ZF17IvHJIus | Use of invertible CNNs to construct counterfactuals and isosurfaces | 5: Marginally below acceptance threshold | This paper describes a computational method to construct ideal counterfactuals and isosurfaces via invertible CNNs, and uses it to reveal biases in three different datasets.
Strengths:
1. The use of directional derivative to construct ideal counterfactuals is interesting.
2. Leveraging PCA to construct isosurfaces is... | 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
Interpretability Through Invertibility: A Deep Convolutional Network With Ideal Counterfactuals And Isosurfaces
### Paper Abstract
Current state of the art computer visi... |
6d5El_LENnf | iscaconf.org/ISCA/2023/Workshop/ASSYST | 2023 | TAP: Efficient Derivation of Tensor Parallel Plans for Large Neural Networks | ["Ziji Shi", "Le Jiang", "Ang Wang", "Jie Zhang", "Xianyan Jia", "Yong Li", "Chencan Wu", "Jialin Li", "Wei Lin"] | Model parallelism is essential to train large language models efficiently. However, determining the optimal model parallel schedule for a given neural network can be slow and inefficient due to the vast choice space. To address this challenge, we propose a tensor model parallelism framework called TAP, which automatica... | ["distributed learning", "machine learning system", "model parallelism"] | Architecture and System Support for Transformer Models (ASSYST), ISCA, 2023TAP: Efficient Derivation of Tensor Parallel Plansfor Large Neural NetworksZiji Shi∗†, Le Jiang†, Jie Zhang†, Xianyan Jia†, Yong Li†, Chencan Wu†, Jialin Li∗, Wei Lin†,∗National University of Singapore†Alibaba GroupAbstract —Model parallelism is... | yNq4OOJjcC | Good improvements but not enough technical detail. | 5: Marginally below acceptance threshold | I enjoyed reading the paper, and found the approach of finding subgraphs in large neural network graphs and exploring parallelism schemes in these subgraphs to reduce the search space sensible. I was a bit disappointed not to see any detail on the GraphNode IR, especially since this is listed as a main contribution | 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
TAP: Efficient Derivation of Tensor Parallel Plans for Large Neural Networks
### Paper Abstract
Model parallelism is essential to train large language models efficiently... |
SJi9WOeRb | ICLR.cc/2018/Conference | 2018 | Gradient Estimators for Implicit Models | ["Yingzhen Li", "Richard E. Turner"] | Implicit models, which allow for the generation of samples but not for point-wise evaluation of probabilities, are omnipresent in real-world problems tackled by machine learning and a hot topic of current research. Some examples include data simulators that are widely used in engineering and scientific research, genera... | ["Implicit Models", "Approximate Inference", "Deep Learning"] | ABSTRACTImplicit models, which allow for the generation of samples but not for point-wiseevaluation of probabilities, are omnipresent in real-world problems tackled by ma-chine learning and a hot topic of current research. Some examples include datasimulators that are widely used in engineering and scientific research,... | rJ8QfICez | an interesting paper | 6: Marginally above acceptance threshold | This paper deals with the estimation of the score function, i.e., the derivative of the log likelihood. Some methods were introduced and a new method using Stein identity was proposed. The setup of the trasnductive learning was introduced to add the prediction power to the proposed method. The method was used to severa... | 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
Gradient Estimators for Implicit Models
### Paper Abstract
Implicit models, which allow for the generation of samples but not for point-wise evaluation of probabilities,... |
rywDjg-RW | ICLR.cc/2018/Conference | 2018 | Neural-Guided Deductive Search for Real-Time Program Synthesis from Examples | ["Ashwin Kalyan", "Abhishek Mohta", "Oleksandr Polozov", "Dhruv Batra", "Prateek Jain", "Sumit Gulwani"] | Synthesizing user-intended programs from a small number of input-output exam-
ples is a challenging problem with several important applications like spreadsheet
manipulation, data wrangling and code refactoring. Existing synthesis systems
either completely rely on deductive logic techniques that are extensively hand-
e... | ["Program synthesis", "deductive search", "deep learning", "program induction", "recurrent neural networks"] | ABSTRACTSynthesizing user-intended programs from a small number of input-output exam-ples is a challenging problem with several important applications like spreadsheetmanipulation, data wrangling and code refactoring. Existing synthesis systemseither completely rely on deductive logic techniques that are extensively ha... | SyFsGdSlM | Although the search method chosen was reasonable, the only real innovation here is to use the LSTM to learn a search heuristic. | 6: Marginally above acceptance threshold | The paper presents a branch-and-bound approach to learn good programs
(consistent with data, expected to generalise well), where an LSTM is
used to predict which branches in the search tree should lead to good
programs (at the leaves of the search tree). The LSTM learns from
inputs of program spec + candidate branch (g... | 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-Guided Deductive Search for Real-Time Program Synthesis from Examples
### Paper Abstract
Synthesizing user-intended programs from a small number of input-output e... |
r1lnigSFDr | ICLR.cc/2020/Conference | 2020 | Improving the Gating Mechanism of Recurrent Neural Networks | ["Albert Gua", "Caglar Gulcehre", "Tom le Paine", "Razvan Pascanu", "Matt Hoffman"] | In this work, we revisit the gating mechanisms widely used in various recurrent and feedforward networks such as LSTMs, GRUs, or highway networks. These gates are meant to control information flow, allowing gradients to better propagate back in time for recurrent models. However, to propagate gradients over very long t... | ["recurrent neural networks", "LSTM", "GRUs", "gating mechanisms", "deep learning", "reinforcement learning"] | ABSTRACTGating mechanisms are widely used in neural network models, where they allowgradients to backpropagate more easily through depth or time. However, their satura-tion property introduces problems of its own. For example, in recurrent models thesegates need to have outputs near 1to propagate information over long ... | S1llO2HCKH | Official Blind Review #1 | 6: Weak Accept |
This paper introduces and studies modifications to gating mechanisms in RNNs.
The first modification is uniform gate initialization. Here the biases in the forget and input gate bias are sampled such that after the application of the sigmoid the values are in the range (1/d, 1-1/d) where d is the dimensionality of 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
Improving the Gating Mechanism of Recurrent Neural Networks
### Paper Abstract
In this work, we revisit the gating mechanisms widely used in various recurrent and feedfo... | |
IXvfIex0mX6f | logconference.io/LOG/2022/Conference | 2022 | DiffWire: Inductive Graph Rewiring via the Lovász Bound | ["Adri\u00e1n Arnaiz-Rodr\u00edguez", "Ahmed Begga", "Francisco Escolano", "Nuria M Oliver"] | Graph Neural Networks (GNNs) have been shown to achieve competitive results to tackle graph-related tasks, such as node and graph classification, link prediction and node and graph clustering in a variety of domains. Most GNNs use a message passing framework and hence are called MPNNs. Despite their promising results, ... | ["GNN", "graph neural networks", "Geometric deep learning", "MPNNs", "graph rewiring", "over-smoothing", "over-squashing", "Lov\u00e1sz bound", "spectral gap", "graph diffusion", "commute times"] | DiffWire: Inductive Graph Rewiring via the Lovász BoundAdrian Arnaiz-RodriguezELLIS Alicanteadrian@ellisalicante.orgAhmed BeggaUniversity of AlicanteFrancisco EscolanoELLIS Alicantesco@ellisalicante.orgNuria OliverELLIS Alicantenuria@ellisalicante.orgAbstractGraph Neural Networks (GNNs) have been shown to achieve comp... | R3wNU5H7MX | This paper proposes DiffWire, a fully differentiable, inductive, and parameter-free graph rewiring algorithm based on the Lovász bound. DiffWire uses the commute times as a relevance function for edge re-weighting, and proposes two types of new layers to either learn the commute times (CT-Layer) or adds a layer that op... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
DiffWire: Inductive Graph Rewiring via the Lovász Bound
### Paper Abstract
Graph Neural Networks (GNNs) have been shown to achieve competitive results to tackle graph-re... | |||
_VoSnnpUDkC | graphicsinterface.org/Graphics_Interface/2021/Conference | 2021 | Generating Adversarial Examples for Robust Deception against Image Transfer and Reloading | ["Chuan Zhou", "Duohe Ma", "Tianwei Zhang", "Liming Wang"] | Adversarial examples play an irreplaceable role in evaluating deep learning models' security and robustness. It is necessary and important to understand the effectiveness of adversarial examples to utilize them for model improvement. In this paper, we explore the impact of input transformation on adversarial examples. ... | ["Adversarial Examples", "Robustness", "Reloading"] | ABSTRACTAdversarial examples play an irreplaceable role in evaluating deeplearning models’ security and robustness. It is necessary and impor-tant to understand the effectiveness of adversarial examples to utilizethem for model improvement. In this paper, we explore the impactof input transformation on adversarial exam... | QN2XZZ3DIMv | Generating Adversarial Examples for Robust Deception against Image Transfer and Reloading | 3: Clear rejection | The authors explore how storing and transmitting images affects the attack performance of adversarial examples. They propose a novel metric called "Confidence Iteration" to generate adversarial examples that are robust to said effects of storage and transmission.
Comments:
- Although the objective is clear, there is n... | 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
Generating Adversarial Examples for Robust Deception against Image Transfer and Reloading
### Paper Abstract
Adversarial examples play an irreplaceable role in evaluatin... |
BylRVjC9K7 | ICLR.cc/2019/Conference | 2019 | Explaining Adversarial Examples with Knowledge Representation | ["Xingyu Zhou", "Tengyu Ma", "Huahong Zhang"] | Adversarial examples are modified samples that preserve original image structures but deviate classifiers. Researchers have put efforts into developing methods for generating adversarial examples and finding out origins. Past research put much attention on decision boundary changes caused by these methods. This paper, ... | ["adversarial example", "knowledge representation", "distribution imitation"] | ABSTRACTAdversarial examples are modified samples that preserve original image structuresbut deviate classifiers. Researchers have put efforts into developing methods forgenerating adversarial examples and finding out origins. Past research put muchattention on decision boundary changes caused by these methods. This pa... | SJlH5W753Q | Very hard to read | 3: Clear rejection | **First of all, this paper uses 11 pages**
Submission instruction is "There will be a strict upper limit of 10 pages."
The readability of the manuscript should be improved.
I'm not convinced why Chapter 2 motivates Chapter 3. I think Ch. 2 and Ch. 3 are different stories.
| 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
Explaining Adversarial Examples with Knowledge Representation
### Paper Abstract
Adversarial examples are modified samples that preserve original image structures but de... |
HygiDTVKPr | ICLR.cc/2020/Conference | 2020 | A Mention-Pair Model of Annotation with Nonparametric User Communities | ["Silviu Paun", "Juntao Yu", "Jon Chamberlain", "Udo Kruschwitz", "Massimo Poesio"] | The availability of large datasets is essential for progress in coreference and other areas of NLP. Crowdsourcing has proven a viable alternative to expert annotation, offering similar quality for better scalability. However, crowdsourcing require adjudication, and most models of annotation focus on classification task... | ["model of annotation", "coreference resolution", "anaphoric annotation", "mention pair model", "bayesian nonparametrics"] | ABSTRACTThe availability of large datasets is essential for progress in coreference and otherareas of NLP. Crowdsourcing has proven a viable alternative to expert annotation,offering similar quality for better scalability. However, crowdsourcing requireadjudication, and most models of annotation focus on classification... | Byec-4kCYr | Official Blind Review #2 | 3: Weak Reject | This paper extends the unpooled mention pair model of annotation (MPA) (Paun et al., 2018b) with the hierarchical priors (e.g., mean and variance) on the ability of the annotators. The proposed method was evaluated on Phrase Detectives 2 corpus, which was annotated by players in a game with a purpose setting for corefe... | <|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 Mention-Pair Model of Annotation with Nonparametric User Communities
### Paper Abstract
The availability of large datasets is essential for progress in coreference and... | |
uQfOy7LrlTR | ICLR.cc/2021/Conference | 2021 | Scaling the Convex Barrier with Active Sets | ["Alessandro De Palma", "Harkirat Behl", "Rudy R Bunel", "Philip Torr", "M. Pawan Kumar"] | Tight and efficient neural network bounding is of critical importance for the scaling of neural network verification systems. A number of efficient specialised dual solvers for neural network bounds have been presented recently, but they are often too loose to verify more challenging properties. This lack of tightness ... | ["Neural Network Verification", "Neural Network Bounding", "Optimisation for Deep Learning"] | ABSTRACTTight and efficient neural network bounding is of critical importance for the scal-ing of neural network verification systems. A number of efficient specialised dualsolvers for neural network bounds have been presented recently, but they are oftentoo loose to verify more challenging properties. This lack of tig... | Ue-TsCMawlK | Review for "Scaling the Convex Barrier with Active Sets" | 7: Good paper, accept | The authors present a custom solver for verifying properties of neural networks (such as robustness properties). Prior work for neural network verification relies on generating bounds by solving convex relaxations ("convex barrier"). The authors describe a sparse dual solver for a new relaxation which is tighter (but h... | 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
Scaling the Convex Barrier with Active Sets
### Paper Abstract
Tight and efficient neural network bounding is of critical importance for the scaling of neural network ve... |
ByetnC4FwS | ICLR.cc/2020/Conference | 2020 | Recurrent Chunking Mechanisms for Conversational Machine Reading Comprehension | ["Hongyu Gong", "Yelong Shen", "Dian Yu", "Jianshu Chen", "Dong Yu"] | In this paper, we focus on the conversational machine reading comprehension (MRC) problem, where the input to a model could be a lengthy document and a series of interconnected questions.
To deal with long inputs, previous approaches usually chunk them into equally-spaced segments and predict answers based on each chun... | ["Recurrent Chunking Policy", "Machine Reading Comprehension", "Reinforcement Learning"] | ABSTRACTIn this paper, we focus on the conversational machine reading comprehension(MRC) problem, where the input to a model could be a lengthy document and aseries of interconnected questions. To deal with long inputs, previous approachesusually chunk them into equally-spaced segments and predict answers based oneach ... | rJedWjNpYB | Official Blind Review #3 | 6: Weak Accept | The paper considers the task of extractive QA where the document is longer than the document encoder's size limit. (The conversational part does not play a big role in the method.) One solution is to look at a chunk of document tokens per time step. The paper proposes (1) a way to propagate information between time ste... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Recurrent Chunking Mechanisms for Conversational Machine Reading Comprehension
### Paper Abstract
In this paper, we focus on the conversational machine reading comprehen... | |
rkYgAJWCZ | ICLR.cc/2018/Conference | 2018 | One-shot and few-shot learning of word embeddings | ["Andrew Kyle Lampinen", "James Lloyd McClelland"] | Standard deep learning systems require thousands or millions of examples to learn a concept, and cannot integrate new concepts easily. By contrast, humans have an incredible ability to do one-shot or few-shot learning. For instance, from just hearing a word used in a sentence, humans can infer a great deal about it, by... | ["One-shot learning", "embeddings", "word embeddings", "natural language processing", "NLP"] | ABSTRACTStandard deep learning systems require thousands or millions of examples tolearn a concept, and cannot integrate new concepts easily. By contrast, humanshave an incredible ability to do one-shot or few-shot learning. For instance, fromjust hearing a word used in a sentence, humans can infer a great deal about i... | HJnrNAtlG | Review | 3: Clear rejection | The paper proposes a technique for exploiting prior knowledge to learn embedding representations for new words with minimal data. The authors provide a good motivation for the task and it is also a nice step in the general direction of learning deep nets and other systems with minimal supervision.
The problem is usef... | 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
One-shot and few-shot learning of word embeddings
### Paper Abstract
Standard deep learning systems require thousands or millions of examples to learn a concept, and can... |
Sklyn6EYvH | ICLR.cc/2020/Conference | 2020 | Disentangled Representation Learning with Sequential Residual Variational Autoencoder | ["Nanxiang Li", "Shabnam Ghaffarzadegan", "Liu Ren"] | Recent advancements in unsupervised disentangled representation learning focus on extending the variational autoencoder (VAE) with an augmented objective function to balance the trade-off between disentanglement and reconstruction. We propose Sequential Residual Variational Autoencoder (SR-VAE) that defines a "Residual... | ["Disentangled Representation Learning", "Variational Autoencoder", "Residual Learning"] | ABSTRACTRecent advancements in unsupervised disentangled representation learning focuson extending the variational autoencoder (V AE) with an augmented objective func-tion to balance the trade-off between disentanglement and reconstruction. Wepropose Sequential Residual Variational Autoencoder (SR-V AE) that defines a“... | rylCNq2htS | Official Blind Review #2 | 8: Accept | The authors of this paper present a novel that for unsupervised disentangled representation learning. The model, named sequential residual VAE (SR-VAE), gradually activates individual latent variables to reconstruct residuals. Quantitative and qualitative experiments show that the proposed model outperforms beta-VAE an... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Disentangled Representation Learning with Sequential Residual Variational Autoencoder
### Paper Abstract
Recent advancements in unsupervised disentangled representation ... | |
6VPl9khIMz | ICLR.cc/2021/Conference | 2021 | Adaptive Stacked Graph Filter | ["Hoang NT", "Takanori Maehara", "Tsuyoshi Murata"] | We study Graph Convolutional Networks (GCN) from the graph signal processing viewpoint by addressing a difference between learning graph filters with fully-connected weights versus trainable polynomial coefficients. We find that by stacking graph filters with learnable polynomial parameters, we can build a highly adapt... | ["Graph Convolutional Network", "vertex classification", "graph signal processing", "adaptive graph filter"] | ABSTRACTWe study Graph Convolutional Networks (GCN) from the graph signal processingviewpoint by addressing a difference between learning graph filters with fully-connected weights versus trainable polynomial coefficients. We find that by stack-ing graph filters with learnable polynomial parameters, we can build a high... | VXgr7O_yV5p | The Review | 5: Marginally below acceptance threshold | This paper proposes to stack the graph filters with learnable polynomial parameters to construct the new graph neural network model. Generally, this paper is well organized and easy to read. Here are my concerns.
1.Essentially, this paper argues that the approximation of Chebyshev polynomials in GCN can only capture 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
Adaptive Stacked Graph Filter
### Paper Abstract
We study Graph Convolutional Networks (GCN) from the graph signal processing viewpoint by addressing a difference betwee... |
HklldOe6sV | ICML.cc/2019/Workshop/RL4RealLife | 2019 | Neural Heterogeneous Scheduler | ["Tegg Taekyong Sung", "Valliappa Chockalingam", "Alex Yahja", "Bo Ryu"] | Access to massive computation allows researchers and developers to succeeded in using technology to enhance processes in many applications. However there have been claims as to the tapering of the exponential decrease in the cost of hardware (following Moore's law) due to physical hardware limitations. Next generation ... | ["deep reinforcement learning", "resource allocation", "actor-critic"] | Neural Heterogeneous SchedulerTegg Taekyong Sung1 *Valliappa Chockalingam1 2 *Alex Yahja1Bo Ryu1AbstractAccess to massive computation allows researchersand developers to succeeded in using technologyto enhance processes in many applications. How-ever there have been claims as to the tapering ofthe exponential decrease ... | Hkl_i3qSp4 | Bordeline. Good problem. Missing technical details and references. Limited empirical evaluation. | 3: Borderline | This paper proposes a reinforcement learning scheduler, which is implemented using a neural network. The scheduler decides adaptively which jobs to execute. The job can be executed only if all jobs that this job depends on have been executed. The goal is to minimize the total execution time of a batch of jobs. This pro... | 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 Heterogeneous Scheduler
### Paper Abstract
Access to massive computation allows researchers and developers to succeeded in using technology to enhance processes i... |
1l8xClLl53 | Interspeech.org/2023/Workshop/SSW | 2023 | Importance of Human Factors in Text-To-Speech Evaluations | ["Lev Finkelstein", "Joshua Camp", "Rob Clark"] | Both mean opinion score (MOS) evaluations and preference tests in text-to-speech are often associated with high rating variance. In this paper we investigate two important factors that affect that variance. One factor is that the variance is coming from how raters are picked for a specific test, and another is the dyna... | ["Text-to-Speech", "Subjective evaluations", "MOS", "Preference Tests", "Comparative Tests", "Test Reproducibility"] | Importance of Human Factors in Text-To-Speech EvaluationsLev Finkelstein, Josh Camp, Rob ClarkGoogle{finklev, joshcamp, rajclarck }@google.comAbstractBoth mean opinion score (MOS) evaluations and preferencetests in text-to-speech are often associated with high rating vari-ance. In this paper we investigate two importan... | mkhe0X-a6m | Interesting research on implications of some design aspects of MOS/preference test. | 7: Good paper, accept | This paper analyses how two human factors affect the variance of MOS/preference test evaluations. 1) rater bias -each people may have different views; 2) variation of rates with time (learning vs fatigue). This is a very interesting topic as human evaluation is still dominant in our area and reducing the variance would... | 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
Importance of Human Factors in Text-To-Speech Evaluations
### Paper Abstract
Both mean opinion score (MOS) evaluations and preference tests in text-to-speech are often a... |
-6b4dsHIdW0 | NeurIPS.cc/2022/Workshop/LaReL | 2022 | Language-guided Task Adaptation for Imitation Learning | ["Prasoon Goyal", "Ray Mooney", "Scott Niekum"] | We introduce a novel setting, wherein an agent needs to learn a task from a demonstration of a related task with the difference between the tasks communicated in natural language. The proposed setting allows reusing demonstrations from other tasks, by providing low effort language descriptions, and can also be used to ... | ["language", "imitation learning", "learning from demonstration"] | Language-guided Task Adaptation for ImitationLearningPrasoon Goyal, Raymond J. Mooney, Scott NiekumDepartment of Computer ScienceUniversity of Texas at Austin{pgoyal,mooney,sniekum}@cs.utexas.eduAbstractWe introduce a novel setting, wherein an agent needs to learn a task from ademonstration of a related task with the d... | CkSkFmZvI99 | Interesting paper but evaluation procedure and results should be made clearer | 7: Good paper, accept | ## Summary
This paper proposes a new setting related to imitation learning and instruction following where the agent must learn a target task from a demonstration of a source task and a natural language description of the difference between the source and desired target tasks. To investigate this task adaptation settin... | 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
Language-guided Task Adaptation for Imitation Learning
### Paper Abstract
We introduce a novel setting, wherein an agent needs to learn a task from a demonstration of a ... |
SJ4Z72Rctm | ICLR.cc/2019/Conference | 2019 | Composing Entropic Policies using Divergence Correction | ["Jonathan J Hunt", "Andre Barreto", "Timothy P Lillicrap", "Nicolas Heess"] | Deep reinforcement learning (RL) algorithms have made great strides in recent years. An important remaining challenge is the ability to quickly transfer existing skills to novel tasks, and to combine existing skills with newly acquired ones. In domains where tasks are solved by composing skills this capacity holds the ... | ["maximum entropy RL", "policy composition", "deep rl"] | ABSTRACTDeep reinforcement learning (RL) algorithms have made great strides in recentyears. An important remaining challenge is the ability to quickly transfer exist-ing skills to novel tasks, and to combine existing skills with newly acquired ones.In domains where tasks are solved by composing skills this capacity hol... | r1l45Kv92X | Interesting work, but need further improvement | 5: Marginally below acceptance threshold |
-- Contribution, Originality, and Quality --
This paper has presented two approaches for transfer learning in the reinforcement learning (RL) setting: max-ent GPI (Section 3.1) and DC (Section 3.2). The authors have also established some theoretical results for these two approaches (Theorem 3.1 and 3.2), and also dem... | 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
Composing Entropic Policies using Divergence Correction
### Paper Abstract
Deep reinforcement learning (RL) algorithms have made great strides in recent years. An import... |
BVSM0x3EDK6 | ICLR.cc/2021/Conference | 2021 | Robust and Generalizable Visual Representation Learning via Random Convolutions | ["Zhenlin Xu", "Deyi Liu", "Junlin Yang", "Colin Raffel", "Marc Niethammer"] | While successful for various computer vision tasks, deep neural networks have shown to be vulnerable to texture style shifts and small perturbations to which humans are robust. In this work, we show that the robustness of neural networks can be greatly improved through the use of random convolutions as data augmentatio... | ["domain generalization", "robustness", "representation learning", "data augmentation"] | ABSTRACTWhile successful for various computer vision tasks, deep neural networks haveshown to be vulnerable to texture style shifts and small perturbations to whichhumans are robust. In this work, we show that the robustness of neural networks canbe greatly improved through the use of random convolutions as data augmen... | 4BQoSUinQx_ | Interesting use of Random convolutions for Data-Augmentation | 6: Marginally above acceptance threshold | This paper proposes a simple way to increase the robustness of the learned representations in a network perform a series of object recognition tasks by adding a random convolution layer as a pre-processing stage, thus “filtering the image” and preserving the global shape but altering the local `texture’ of the newly tr... | 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 and Generalizable Visual Representation Learning via Random Convolutions
### Paper Abstract
While successful for various computer vision tasks, deep neural networ... |
g7j2e-zlKFS | eswc-conferences.org/ESWC/2021/Conference/Research_Track | 2021 | Data Reliability and Trustworthiness through Digital Transmission Contracts | ["Simon Mangel", "Lars Gleim", "Jan Pennekamp", "Klaus Wehrle", "Stefan Decker"] | As decision-making is increasingly data-driven, trustworthiness and reliability of the underlying data, e.g., maintained in knowledge graphs or on the Web, are essential requirements for their usability in the industry. However, neither traditional solutions, such as paper-based data curation processes, nor state-of-th... | ["Digital transmission contracts", "Trust", "Data immutability", "Non-repudiation", "Accountability", "Data dynamics", "Linked Data", "Knowledge graphs"] | Data Reliability and Trustworthinessthrough Digital Transmission ContractsSimon Mangel1, Lars Gleim1, Jan Pennekamp2,Klaus Wehrle2, and Stefan Decker1;31Databases and Information Systems, RWTH Aachen University, Germany2Communication and Distributed Systems, RWTH Aachen University, Germany3Fraunhofer FIT, Sankt Augusti... | OmzRIXpEDby | An interesting solution to a pressing problem in sharing of trustworthy (linked) data | 2: Accept | The authors address the pressing issue of sharing (linked) data in a trustworthy and reliable manner by introducing the concept of Digital Transmission Contracts and demontrate its practical applicability and scalability with a PoC called ReShare.
The paper is well written, well structured and easy to read. The motivat... | 2: The reviewer is willing to defend the evaluation but not sufficiently familiar with the state of the art or the specific topic 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
Data Reliability and Trustworthiness through Digital Transmission Contracts
### Paper Abstract
As decision-making is increasingly data-driven, trustworthiness and reliab... |
yuXQOhKRjBr | ICLR.cc/2021/Conference | 2021 | Towards Powerful Graph Neural Networks: Diversity Matters | ["Xu Bingbing", "Huawei Shen", "Qi Cao", "Yuanhao Liu", "Keting Cen", "Xueqi Cheng"] | Graph neural networks (GNNs) offer us an effective framework for graph representation learning via layer-wise neighborhood aggregation. Their success is attributed to their expressive power at learning representation of nodes and graphs. To achieve GNNs with high expressive power, existing methods mainly resort to comp... | ["GNNs", "Expressive power", "Diverse sampling", "Injective"] | ABSTRACTGraph neural networks (GNNs) offer us an effective framework for graph repre-sentation learning via layer-wise neighborhood aggregation. Their success is at-tributed to their expressive power at learning representation of nodes and graphs.To achieve GNNs with high expressive power, existing methods mainly resor... | eB32BJqEXl | This paper introduces a graph sampling method to increase the chance that GNN can differentiate different node representations. The idea makes sense but is straightforward. The novelty is limited. | 4: Ok but not good enough - rejection | In this paper, the authors propose to sample nodes of a given graph multiple times to form a set of K sub-graphs. GNNs are then applied on each sampled graph for learning node representations. For each node, all representations are combined for the downstream tasks. The idea of doing multiple sampling is to increase th... | 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
Towards Powerful Graph Neural Networks: Diversity Matters
### Paper Abstract
Graph neural networks (GNNs) offer us an effective framework for graph representation learni... |
HxWEL2zQ3AK | ML_Reproducibility_Challenge/2021/Fall | 2021 | Comparing Rewinding and Fine-tuning in Neural Network Pruning | ["Szymon Jakub Mikler"] | Scope of Reproducibility
We are reproducing Comparing Rewinding and Fine-tuning in Neural Networks, by Renda et al. In this work the authors compare three different approaches to retraining neural networks after pruning: 1) fine-tuning, 2) rewinding weights as in Frankle et al. and 3) a new, original method involv... | ["pruning", "lottery ticket hypothesis", "finetuning"] | Comparing Rewinding and Fine-tuning in Neural NetworkPruning Reproducibility Challenge 2021Anonymous Author(s)AffiliationemailReproduction Summary 1Scope of Reproducibility 2We are reproducing Comparing Rewinding and Fine-tuning in Neural Networks , by Renda et al. [2020]. In this work 3the authors compare three differ... | HMfgNETCilc | Review | 8: Top 50% of accepted papers, clear accept | This report does an excellent job at a reproducing the original paper. I want to commend the authors at their outstanding tenacity to reproduce the results from the paper description alone. Only a few comments/suggestions for future reproduction efforts, but overall I think this is an excellent report and should defini... | 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
Comparing Rewinding and Fine-tuning in Neural Network Pruning
### Paper Abstract
Scope of Reproducibility We are reproducing Comparing Rewinding and Fine-tuning in ... |
9wHe4F-lpp | ICLR.cc/2021/Conference | 2021 | FTBNN: Rethinking Non-linearity for 1-bit CNNs and Going Beyond | ["Zhuo Su", "Linpu Fang", "Deke Guo", "Dewen Hu", "Matti Pietik\u00e4inen", "Li Liu"] | Binary neural networks (BNNs), where both weights and activations are binarized into 1 bit, have been widely studied in recent years due to its great benefit of highly accelerated computation and substantially reduced memory footprint that appeal to the development of resource constrained devices. In contrast to previo... | ["Binary neural networks", "network quantization", "network compression"] | ABSTRACTBinary neural networks (BNNs), where both weights and activations are binarizedinto 1 bit, have been widely studied in recent years due to its great benefit ofhighly accelerated computation and substantially reduced memory footprint thatappeal to the development of resource constrained devices. In contrast to p... | rXEN3-UyRhI | Official Blind Review #2 | 4: Ok but not good enough - rejection | ---paper summary---:
This paper proposes to improve the BNN’s discriminative ability by introducing additional non-linearities. In addition, the paper exploits the group convolution to enable a wider network, which can strengthen BNN’s representational ability, while keeping the total overhead unchanged.
---Pro... | 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
FTBNN: Rethinking Non-linearity for 1-bit CNNs and Going Beyond
### Paper Abstract
Binary neural networks (BNNs), where both weights and activations are binarized into 1... |
-2FCwDKRREu | ICLR.cc/2021/Conference | 2021 | Learning Invariant Representations for Reinforcement Learning without Reconstruction | ["Amy Zhang", "Rowan Thomas McAllister", "Roberto Calandra", "Yarin Gal", "Sergey Levine"] | We study how representation learning can accelerate reinforcement learning from rich observations, such as images, without relying either on domain knowledge or pixel-reconstruction. Our goal is to learn representations that provide for effective downstream control and invariance to task-irrelevant details. Bisimulatio... | ["rich observations", "bisimulation metrics", "representation learning", "state abstractions"] | ABSTRACTWe study how representation learning can accelerate reinforcement learning fromrich observations, such as images, without relying either on domain knowledge orpixel-reconstruction. Our goal is to learn representations that provide for effectivedownstream control and invariance to task-irrelevant details. Bisimu... | RbVptpSUFSy | The paper is well written and presents a clear approach to a well motivated problem with strong evaluation results. | 7: Good paper, accept | The authors propose an approach to robust representation learning of observations for reinforcement learning by training a model to align the euclidean distance between two observations with bisimulation metrics that quantify how similar the states that generated the observations are in terms of the control problem. 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
Learning Invariant Representations for Reinforcement Learning without Reconstruction
### Paper Abstract
We study how representation learning can accelerate reinforcement... |
KR-IAbeRpDH | NeurIPS.cc/2020/Workshop/SVRHM | 2020 | Assessing The Importance Of Colours For CNNs In Object Recognition | ["Aditya Singh", "Alessandro Bay", "Andrea Mirabile"] | Humans rely heavily on shapes as a primary cue for object recognition. As secondary cues, colours and textures are also beneficial in this regard. Convolutional neural networks (CNNs), an imitation of biological neural networks, have been shown to exhibit conflicting properties. Some studies indicate that CNNs are bias... | ["Image classification", "colour bias", "incongruent evaluation"] | Assessing The Importance Of Colours For CNNs InObject RecognitionAditya Singh, Alessandro Bay, Andrea MirabileZebra AI, Zebra TechnologiesLondon, United Kingdom{firstname.lastname}@zebra.comAbstractHumans rely heavily on shapes as a primary cue for object recognition. As sec-ondary cues, colours and textures are also b... | kMFojYbqyuf | Do humans and deep networks rely on colour information similarly? | 7: Good paper, accept | I find this line of investigation to study the impact of colour and its representation in deep networks very interesting and important for human- and machine-vision community. It would be nice to read more about the shared representation between humans and machines in the conclusion/discussion section linking (more str... | 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
Assessing The Importance Of Colours For CNNs In Object Recognition
### Paper Abstract
Humans rely heavily on shapes as a primary cue for object recognition. As secondary... |
SygT21SFvB | ICLR.cc/2020/Conference | 2020 | Towards Understanding Generalization in Gradient-Based Meta-Learning | ["Simon Guiroy", "Vikas Verma", "Christopher J. Pal"] | In this work we study generalization of neural networks in gradient-based meta-learning by analyzing various properties of the objective landscapes. We experimentally demonstrate that as meta-training progresses, the meta-test solutions obtained by adapting the meta-train solution of the model to new tasks via few step... | ["meta-learning", "objective landscapes"] | ABSTRACTIn this work we study generalization of neural networks in gradient-based meta-learning by analyzing various properties of the objective landscapes. We exper-imentally demonstrate that as meta-training progresses, the meta-test solutionsobtained by adapting the meta-train solution of the model to new tasks via ... | rkeynNmgqB | Official Blind Review #3 | 3: Weak Reject | This paper addresses empirical study of generalization behavior in gradient-based meta-learning. To this end, authors evaluated: (1) flatness of minima in terms of the spectral norm of Hessian matrix; (2) coherence of adaptation trajectories; (3) the average inner product between meta-test gradient vectors. Finally a r... | <|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 Understanding Generalization in Gradient-Based Meta-Learning
### Paper Abstract
In this work we study generalization of neural networks in gradient-based meta-le... | |
BkByNuojz | MIDL.amsterdam/2018/Conference | 2018 | Fully Automatic Segmentation of Sphenoid Sinus in CT Images with 3D Convolutional Neural Networks | ["Kamal Souadih", "Ahror Belaid", "Douraied Ben Salem"] | Today, Deep learning algorithms have quickly become essential in the field of medical image analysis. Compared to the traditional methods, these Deep learning techniques are more efficient in extracting compact information leading towards significant improvement performance of medical image analysis system. We present ... | ["3D Convolutional neural network", "Computed Tomography", "Automatic Segmentation", "Sphenoid Sinus"] | Fully Automatic Segmentation of Sphenoid Sinus inCT Images with 3D Convolutional Neural NetworksKamal SouadihMedical Computing Laboratory (LIMED),University of Abderrahmane Mira,06000, Bejaia, Algeriakamal.souadih@univ-bejaia.dzAhror BelaidMedical Computing Laboratory (LIMED),University of Abderrahmane Mira,06000, Beja... | SJgc5z9af | limited contribution, preliminary validation | 1: Strong rejection | This paper addresses the problem of segmenting sphenoid sinus in CT scans. This is an application paper, where a standard CNN model is applied to perform patch based segmentation, with pre- and post-processing steps. The pre-processing step involves the extraction of a region of interest and a registration step. The po... | 2: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Fully Automatic Segmentation of Sphenoid Sinus in CT Images with 3D Convolutional Neural Networks
### Paper Abstract
Today, Deep learning algorithms have quickly become ... |
c77KhoLYSwF | ICLR.cc/2021/Conference | 2021 | Just How Toxic is Data Poisoning? A Benchmark for Backdoor and Data Poisoning Attacks | ["Avi Schwarzschild", "Micah Goldblum", "Arjun Gupta", "John P Dickerson", "Tom Goldstein"] | Data poisoning and backdoor attacks manipulate training data in order to cause models to fail during inference. A recent survey of industry practitioners found that data poisoning is the number one concern among threats ranging from model stealing to adversarial attacks. However, we find that the impressive performanc... | ["Poisoning", "backdoor", "attack", "benchmark"] | ABSTRACTData poisoning and backdoor attacks manipulate training data in order to causemodels to fail during inference. A recent survey of industry practitioners foundthat data poisoning is the number one concern among threats ranging from modelstealing to adversarial attacks. However, we find that the impressive perfor... | OvP4cdmNanw | Review for Just How Toxic is Data Poisoning? A Benchmark for Backdoor and Data Poisoning Attacks | 4: Ok but not good enough - rejection | Reject
This paper consists of two main parts, essentially inspired from the argumentation of the Baconian method. In particular, the first part criticizes current work in the area of data poisoning attacks (pars destruens), while the second part tries to overcome the limits described in the first half of the paper (pa... | 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
Just How Toxic is Data Poisoning? A Benchmark for Backdoor and Data Poisoning Attacks
### Paper Abstract
Data poisoning and backdoor attacks manipulate training data in... |
B1fysiAqK7 | ICLR.cc/2019/Conference | 2019 | Probabilistic Binary Neural Networks | ["Jorn W.T. Peters", "Tim Genewein", "Max Welling"] | Low bit-width weights and activations are an effective way of combating the increasing need for both memory and compute power of Deep Neural Networks. In this work, we present a probabilistic training method for Neural Network with both binary weights and activations, called PBNet. By embracing stochasticity during tra... | ["binary neural Network", "efficient deep learning", "stochastic training", "discrete neural network", "efficient inference"] | ABSTRACTLow bit-width weights and activations are an effective way of combating theincreasing need for both memory and compute power of Deep Neural Networks. Inthis work, we present a probabilistic training method for Neural Network with bothbinary weights and activations, called PBNet. By embracing stochasticity durin... | H1eCvLlshm | The investigated problem is interesting, but methods used to handle binary networks are not impressive | 5: Marginally below acceptance threshold | To reduce the deep neural networks' reliance on memory and high power consumption, the paper proposed a kind of probabilistic neural networks with both binary hidden node and binary weights. The paper presents a probabilistic way to binarize the activation and weight values. Also, it proposed a random version of batch-... | 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
Probabilistic Binary Neural Networks
### Paper Abstract
Low bit-width weights and activations are an effective way of combating the increasing need for both memory and c... |
W59BHjEDfz | logconference.io/LOG/2022/Conference | 2022 | Graph-based Reinforcement Learning meets Mixed Integer Programs: An application to 3D robot assembly discovery | ["Niklas Funk", "Svenja Menzenbach", "Georgia Chalvatzaki", "Jan Peters"] | Robot assembly discovery is a challenging problem that lives at the intersection of resource allocation and motion planning. The goal is to combine a predefined set of objects to form something new while considering task execution with the robot-in-the-loop. In this work, we tackle the problem of building arbitrary, pr... | ["graph neural networks", "reinforcement learning", "robotics"] | Graph-based Reinforcement Learning meets Mixed IntegerPrograms: An application to 3D robot assembly discoveryNiklas FunkTechnical University of Darmstadtniklas@robot-learning.deSvenja MenzenbachTechnical University of Darmstadtsvenja.menzenbach@stud.tu-darmstadt.deGeorgia ChalvatzakiTechnical University of Darmstadtgeo... | XEq-lQiNrtF | Contributions:
1. The paper focuses on an interesting problem of autonomous robotic assembly discovery (RAD). The goal is that a robotic agent should reason about abstract 3D target shapes that need to be fulfilled given a set of available building block.
2. The problem is challenging to solve .
3. The authors use 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
Graph-based Reinforcement Learning meets Mixed Integer Programs: An application to 3D robot assembly discovery
### Paper Abstract
Robot assembly discovery is a challengi... | |||
SyG1QnRqF7 | ICLR.cc/2019/Conference | 2019 | Towards Resisting Large Data Variations via Introspective Learning | ["Yunhan Zhao", "Ye Tian", "Wei Shen", "Alan Yuille"] | Learning deep networks which can resist large variations between training andtesting data is essential to build accurate and robust image classifiers. Towardsthis end, a typical strategy is to apply data augmentation to enlarge the trainingset. However, standard data augmentation is essentially a brute-force ... | ["Introspective learning", "Large variations resistance", "Image classification", "Generative models"] | ABSTRACTLearning deep networks which can resist large variations between training andtesting data is essential to build accurate and robust image classifiers. Towardsthis end, a typical strategy is to apply data augmentation to enlarge the trainingset. However, standard data augmentation is essentially a brute-force st... | SJlQKiZGT7 | Good discussion, improved comparisons needed | 5: Marginally below acceptance threshold | This paper suggests the use of learned transformation networks, embedded within introspective networks to improve classification performance with synthesized examples.
The authors cite a number of related works, and give a good introduction to both introspective learning, as well as the particular usage for large vari... | 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
Towards Resisting Large Data Variations via Introspective Learning
### Paper Abstract
Learning deep networks which can resist large variations between training andtestin... |
Byg5ZANtvH | ICLR.cc/2020/Conference | 2020 | Short and Sparse Deconvolution --- A Geometric Approach | ["Yenson Lau", "Qing Qu", "Han-Wen Kuo", "Pengcheng Zhou", "Yuqian Zhang", "John Wright"] | Short-and-sparse deconvolution (SaSD) is the problem of extracting localized, recurring motifs in signals with spatial or temporal structure. Variants of this problem arise in applications such as image deblurring, microscopy, neural spike sorting, and more. The problem is challenging in both theory and practice, as na... | ["short", "sparse deconvolution", "geometric", "sasd", "problem", "theory", "deconvolution", "motifs", "signals", "spatial"] | ABSTRACTShort-and-sparse deconvolution (SaSD) is the problem of extracting localized,recurring motifs in signals with spatial or temporal structure. Variants of thisproblem arise in applications such as image deblurring, microscopy, neural spikesorting, and more. The problem is challenging in both theory and practice, ... | rJlh3raLqH | Official Blind Review #2 | 3: Weak Reject | This paper analyzes some of the optimization challenges presented by a particular formulation of "short-and-sparse deconvolution" and proposes a new general purpose algorithm to try to alleviate them. The paper is reasonably clearly written and the experimental results seem impressive (as a non-specialist in this area)... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Short and Sparse Deconvolution --- A Geometric Approach
### Paper Abstract
Short-and-sparse deconvolution (SaSD) is the problem of extracting localized, recurring motifs... | |
YmA86Zo-P_t | ICLR.cc/2021/Conference | 2021 | What they do when in doubt: a study of inductive biases in seq2seq learners | ["Eugene Kharitonov", "Rahma Chaabouni"] | Sequence-to-sequence (seq2seq) learners are widely used, but we still have only limited knowledge about what inductive biases shape the way they generalize. We address that by investigating how popular seq2seq learners generalize in tasks that have high ambiguity in the training data. We use four new tasks to study le... | ["inductive biases", "description length", "sequence-to-sequence models"] | ABSTRACTSequence-to-sequence (seq2seq) learners are widely used, but we still have onlylimited knowledge about what inductive biases shape the way they generalize. Weaddress that by investigating how popular seq2seq learners generalize in tasksthat have high ambiguity in the training data. We use four new tasks to stud... | BaEccdWKLSt | Why is this important? | 4: Ok but not good enough - rejection | The paper introduces a series of new datasets and task and investigates the inductive bias of seq2seq models. For each dataset, (at least) two hidden hypothesis could explain the data. The tasks investigated are count-vs-memorization, add-or-multiply, hierarchical-or-linear, composition-or-memorization. The datasets 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
What they do when in doubt: a study of inductive biases in seq2seq learners
### Paper Abstract
Sequence-to-sequence (seq2seq) learners are widely used, but we still have... |
tf8a4jDRFCv | ICLR.cc/2021/Conference | 2021 | Learning Aggregation Functions | ["Giovanni Pellegrini", "Alessandro Tibo", "Paolo Frasconi", "Andrea Passerini", "Manfred Jaeger"] | Learning on sets is increasingly gaining attention in the machine learning community, due to its widespread applicability. Typically, representations over sets are computed by using fixed aggregation functions such as sum or maximum. However, recent results showed that universal function representation by sum- (or max... | ["Deep learning", "Neural networks", "Relational and structured data", "Aggregation functions"] | ABSTRACTLearning on sets is increasingly gaining attention in the machine learning com-munity, due to its widespread applicability. Typically, representations over setsare computed by using fixed aggregation functions such as sum or maximum.However, recent results showed that universal function representation by sum- (... | ukmCfRiCNgL | A novel method for aggregating the information from sets | 6: Marginally above acceptance threshold | Summary:
Universal function representation guarantee requires either highly discontinuous mappings or a highly dimensional latent space. For this reason the authors propose a new parametric family of aggregation functions, called LAF (for learning aggregation functions). It can be seen as a smooth version of the class... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Learning Aggregation Functions
### Paper Abstract
Learning on sets is increasingly gaining attention in the machine learning community, due to its widespread applicabil... |
0BaWDGvCa5p | ICLR.cc/2021/Conference | 2021 | A Provably Convergent and Practical Algorithm for Min-Max Optimization with Applications to GANs | ["Oren Mangoubi", "Sushant Sachdeva", "Nisheeth K Vishnoi"] | We present a first-order algorithm for nonconvex-nonconcave min-max optimization problems such as those that arise in training GANs. Our algorithm provably converges in $\mathrm{poly}(d,L, b)$ steps for any loss function $f:\mathbb{R}^d \times \mathbb{R}^d \rightarrow \mathbb{R}$ which is $b$-bounded with ${L}$-Lipsch... | ["min-max optimization", "GANs"] | ABSTRACTWe present a first-order algorithm for nonconvex-nonconcave min-max optimiza-tion problems such as those that arise in training GANs. Our algorithm provablyconverges in poly(d;L;b )steps for any loss function f:RdRd!Rwhichisb-bounded with L-Lipschitz gradient. To achieve convergence, we 1) give anovel approxima... | TL5WmAu2YON | A Provably Convergent and Practical Algorithm for Min-Max Optimization with Applications to GANs | 6: Marginally above acceptance threshold | This paper proposes a new stochastic gradient descent-ascent-based method to approximate a stationary point (or local min-max solution) of a nonconvex-nonconcave minimax problem with application in GANs. The method is similar to the one in the GAN original paper, but the authors incorporate it with an acceptance rule a... | 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 Provably Convergent and Practical Algorithm for Min-Max Optimization with Applications to GANs
### Paper Abstract
We present a first-order algorithm for nonconvex-nonc... |
mhEd8uOyNTI | ICLR.cc/2021/Conference | 2021 | Representational correlates of hierarchical phrase structure in deep language models | ["Matteo Alleman", "Jonathan Mamou", "Miguel A Del Rio", "Hanlin Tang", "Yoon Kim", "SueYeon Chung"] | While contextual representations from Transformer-based architectures have set a new standard for many NLP tasks, there is not yet a complete accounting of their inner workings. In particular, it is not entirely clear what aspects of sentence-level syntax are captured by these representations, nor how (if at all) they ... | ["bertology", "interpretability", "computational neuroscience", "population coding"] | ABSTRACTWhile contextual representations from pretrained Transformer models have set anew standard for many NLP tasks, there is not yet a complete accounting of theirinner workings. In particular, it is not entirely clear what aspects of sentence-levelsyntax are captured by these representations, nor how (if at all) th... | WDOZdzxq6xj | Reveal the correlations of hierarchical phrase structure and BERT's output, but lack theoretical explanation. | 6: Marginally above acceptance threshold | --- Overall ---
This paper provides some insights into the relation of BERT's output w.r.t. the parser tree (in terms of constituent phrases) of the input sentence. As some previous work has pointed out, BERT model contains the parsing information (e.g., Hewitt & Manning NAACL’19)). This work can be regarded as a mode... | 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
Representational correlates of hierarchical phrase structure in deep language models
### Paper Abstract
While contextual representations from Transformer-based architect... |
vlcVTDaufN | ICLR.cc/2021/Conference | 2021 | Differentiable Combinatorial Losses through Generalized Gradients of Linear Programs | ["Xi Gao", "Han Zhang", "Aliakbar Panahi", "Tom Arodz"] | Combinatorial problems with linear objective function play a central role in many computer science applications, and efficient algorithms for solving them are well known. However, the solutions to these problems are not differentiable with respect to the parameters specifying the problem instance – for example, shortes... | ["combinatorial optimization", "linear programs", "generalized gradient"] | ABSTRACTCombinatorial problems with linear objective function play a central role in manycomputer science applications, and efficient algorithms for solving them are wellknown. However, the solutions to these problems are not differentiable with re-spect to the parameters specifying the problem instance – for example, ... | MOJj-92FseJ | Differentiating over the objective of a linear (integer) program is not an interesting problem | 3: Clear rejection | The value of the optimal objective as a function of the cost vector $c$ can be written as $z^*(c) = c^T u^*(c)$ where the optimal solution $u^*$ also depends on $c$. The function $u^*(c)$ is piecewise constant -- there are finitely (resp. countably) many feasible solutions; candidates for $u^*$ -- and so the function $... | 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
Differentiable Combinatorial Losses through Generalized Gradients of Linear Programs
### Paper Abstract
Combinatorial problems with linear objective function play a cent... |
wjfD_pio2K | ICLR.cc/2023/BlogPosts | 2023 | Decay No More | ["Fabian Schaipp"] | Weight decay is among the most important tuning parameters to reach high accuracy for large-scale machine learning models. In this blog post, we revisit AdamW, the weight decay version of Adam, summarizing empirical findings as well as theoretical motivations from an optimization perspective. | ["weight decay", "adam", "optimization"] | o2182sxlML | Nicely written and relevant blog | 7: Good paper, accept | This blog discusses the different ways weight decay can be implemented in the Adam optimizer. A special focus is given to the AdamW version. The blog also discusses the current understanding of why weight decay improves generalisation.
The blog is nicely written and provides an easy introduction to this area of the l... | 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
Decay No More
### Paper Abstract
Weight decay is among the most important tuning parameters to reach high accuracy for large-scale machine learning models. In this blog ... | |
ryUprTOv7q0 | ICLR.cc/2021/Conference | 2021 | Quantum Deformed Neural Networks | ["Roberto Bondesan", "Max Welling"] | We develop a new quantum neural network layer designed to run efficiently on a quantum computer but that can be simulated on a classical computer when restricted in the way it entangles input states. We first ask how a classical neural network architecture, both fully connected or convolutional, can be executed on a qu... | ["Quantum machine learning", "Binary neural networks", "Bayesian deep learning"] | ABSTRACTWe develop a new quantum neural network layer designed to run efficiently ona quantum computer but that can be simulated on a classical computer when re-stricted in the way it entangles input states. We first ask how a classical neuralnetwork architecture, both fully connected or convolutional, can be executed ... | AvY1h9NHpD | too many crucial details are deferred to the supplementary material | 6: Marginally above acceptance threshold | The fundamental idea in this paper is to endow classical signals with a complex Hilberspace structure so as to harness the probabilistic nature of quantum mechaniscs for pattern recognition based on quantum computing principles. Following this idea, the authors consider (probabiistic) binary neural networks and develop... | 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
Quantum Deformed Neural Networks
### Paper Abstract
We develop a new quantum neural network layer designed to run efficiently on a quantum computer but that can be simul... |
XsQLS6Ls5- | NeurIPS.cc/2022/Workshop/Offline_RL | 2022 | Model-based Trajectory Stitching for Improved Offline Reinforcement Learning | ["Charles Alexander Hepburn", "Giovanni Montana"] | In many real-world applications, collecting large and high-quality datasets may be too costly or impractical. Offline reinforcement learning (RL) aims to infer an optimal decision-making policy from a fixed set of data. Getting the most information from historical data is then vital for good performance once the policy... | ["policy", "trajectory", "improved offline reinforcement", "data augmentation strategy", "many", "applications", "large", "datasets", "costly"] | Model-based Trajectory Stitching for ImprovedOffline Reinforcement LearningCharles A. Hepburn1Giovanni Montana1,21University of Warwick2Alan Turing Institute{Charlie.Hepburn,g.montana}@warwick.ac.ukAbstractIn many real-world applications, collecting large and high-quality datasets maybe too costly or impractical. Offli... | -Yl5KMi9GgX | Review | 5: Marginally below acceptance threshold | This paper presents a data augmentation scheme for BC in the offline setting. The authors propose to produce trajectory stitching based on the likelihood of the next state, next action, and high values. The authors train a forward dynamics model, an inverse dynamics model, a value function with a reward generator and a... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Model-based Trajectory Stitching for Improved Offline Reinforcement Learning
### Paper Abstract
In many real-world applications, collecting large and high-quality datase... |
OEgDatKuz2O | ICLR.cc/2021/Conference | 2021 | EMTL: A Generative Domain Adaptation Approach | ["Jianfeng Zhang", "Illyyne Saffar", "Aladin Virmaux", "Bal\u00e1zs K\u00e9gl"] | We propose an unsupervised domain adaptation approach based on generative models. We show that when the source probability density function can be learned, one-step Expectation–Maximization iteration plus an additional marginal density function constraint will produce a proper mediator probability density function to b... | ["unsupervised domain adaptation", "EM", "generative model", "density estimation", "deep learning", "transfer learning"] | ABSTRACTWe propose an unsupervised domain adaptation approach based on generativemodels. We show that when the source probability density function can be learned,one-step Expectation–Maximization iteration plus an additional marginal densityfunction constraint will produce a proper mediator probability density function... | 97XgRYd7qC | An interesting motivation, but more work would be needed. | 5: Marginally below acceptance threshold | This paper proposes a novel method for Unsupervised Domain Adaptation (UDA) when the source domain's privacy should be preserved. The authors propose EMTL, which is a generative method using multivariate densities using RNADE (Uria et al., 2013) and a mediator joint density function bridging both source and target doma... | 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
EMTL: A Generative Domain Adaptation Approach
### Paper Abstract
We propose an unsupervised domain adaptation approach based on generative models. We show that when the ... |
HygXkJHtvB | ICLR.cc/2020/Conference | 2020 | Using Objective Bayesian Methods to Determine the Optimal Degree of Curvature within the Loss Landscape | ["Devon Jarvis", "Richard Klein", "Benjamin Rosman"] | The efficacy of the width of the basin of attraction surrounding a minimum in parameter space as an indicator for the generalizability of a model parametrization is a point of contention surrounding the training of artificial neural networks, with the dominant view being that wider areas in the landscape reflect better... | ["Objective Bayes", "Information Geometry", "Artificial Neural Networks"] | ABSTRACTThe efficacy of the width of the basin of attraction surrounding a minimum inparameter space as an indicator for the generalizability of a model parametriza-tion is a point of contention surrounding the training of artificial neural networks,with the dominant view being that wider areas in the landscape reflect... | SJxwSVtcKr | Official Blind Review #1 | 1: Reject | This paper targets at a deep learning theory contribution based on information geometry. This contribution is tightly based on Zhang et al. (2018) and explains the generalization of deep learning from a Bayesian perspective. The main contribution the authors claimed is an optimal degree of curvature exist which gives 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
Using Objective Bayesian Methods to Determine the Optimal Degree of Curvature within the Loss Landscape
### Paper Abstract
The efficacy of the width of the basin of attr... | |
MJAqnaC2vO1 | ICLR.cc/2021/Conference | 2021 | Auto Seg-Loss: Searching Metric Surrogates for Semantic Segmentation | ["Hao Li", "Chenxin Tao", "Xizhou Zhu", "Xiaogang Wang", "Gao Huang", "Jifeng Dai"] | Designing proper loss functions is essential in training deep networks. Especially in the field of semantic segmentation, various evaluation metrics have been proposed for diverse scenarios. Despite the success of the widely adopted cross-entropy loss and its variants, the mis-alignment between the loss functions and e... | ["Loss Function Search", "Metric Surrogate", "Semantic Segmentation"] | ABSTRACTDesigning proper loss functions is essential in training deep networks. Especiallyin the field of semantic segmentation, various evaluation metrics have been pro-posed for diverse scenarios. Despite the success of the widely adopted cross-entropy loss and its variants, the mis-alignment between the loss functio... | Mjs8kwRjcof | Careful study of semantic segmentation proxy loss functions | 7: Good paper, accept | This paper looks at loss functions for semantic segmentation. Typical metrics are not easily differentiable w.r.t. the outputs of the DNNs that generate the labels. Instead, a proxy/surrogate loss function is learned jointly with the network in a two-level optimization.
Pros:
i) Good accuracy results. Code will allow... | 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
Auto Seg-Loss: Searching Metric Surrogates for Semantic Segmentation
### Paper Abstract
Designing proper loss functions is essential in training deep networks. Especiall... |
8Ln-Bq0mZcy | ICLR.cc/2021/Conference | 2021 | On the Critical Role of Conventions in Adaptive Human-AI Collaboration | ["Andy Shih", "Arjun Sawhney", "Jovana Kondic", "Stefano Ermon", "Dorsa Sadigh"] | Humans can quickly adapt to new partners in collaborative tasks (e.g. playing basketball), because they understand which fundamental skills of the task (e.g. how to dribble, how to shoot) carry over across new partners. Humans can also quickly adapt to similar tasks with the same partners by carrying over conventions t... | ["Multi-agent games", "emergent behavior", "transfer learning", "human-AI collaboration"] | ABSTRACTHumans can quickly adapt to new partners in collaborative tasks (e.g. playingbasketball), because they understand which fundamental skills of the task (e.g.how to dribble, how to shoot) carry over across new partners. Humans can alsoquickly adapt to similar tasks with the same partners by carrying over conventi... | GbUX4VGLOgn | Review | 7: Good paper, accept | This paper makes the observation that when performing cooperative tasks with partners, there are two components to learn: how to perform the task, and how to coordinate with the partner according to conventions. Therefore, it proposes to separate these two components via a modular architecture, which learns a task-spec... | 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 the Critical Role of Conventions in Adaptive Human-AI Collaboration
### Paper Abstract
Humans can quickly adapt to new partners in collaborative tasks (e.g. playing b... |
H1_EDpogx | ICLR.cc/2017/conference | 2017 | Near-Data Processing for Machine Learning | ["Hyeokjun Choe", "Seil Lee", "Hyunha Nam", "Seongsik Park", "Seijoon Kim", "Eui-Young Chung", "Sungroh Yoon"] | In computer architecture, near-data processing (NDP) refers to augmenting the memory or the storage with processing power so that it can process the data stored therein. By offloading the computational burden of CPU and saving the need for transferring raw data in its entirety, NDP exhibits a great potential for accele... | ["processing", "ndp", "isp", "sgd", "machine learning", "memory", "data", "potential", "ssd"] | ABSTRACTIn computer architecture, near-data processing (NDP) refers to augmenting thememory or the storage with processing power so that it can process the data storedtherein. By offloading the computational burden of CPU and saving the need fortransferring raw data in its entirety, NDP exhibits a great potential for a... | B1dxnA-El | Near-Data Processing for Machine Learning | 5: Marginally below acceptance threshold | While the idea of moving the processing for machine learning into silicon contained within the (SSD) data storage devices is intriguing and offers the potential for low-power efficient computation, it is a rather specialized topic, so I don't feel it will be of especially wide interest to the ICLR audience. The paper d... | 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
Near-Data Processing for Machine Learning
### Paper Abstract
In computer architecture, near-data processing (NDP) refers to augmenting the memory or the storage with pro... |
HklUCCVKDB | ICLR.cc/2020/Conference | 2020 | Uncertainty-guided Continual Learning with Bayesian Neural Networks | ["Sayna Ebrahimi", "Mohamed Elhoseiny", "Trevor Darrell", "Marcus Rohrbach"] | Continual learning aims to learn new tasks without forgetting previously learned ones. This is especially challenging when one cannot access data from previous tasks and when the model has a fixed capacity. Current regularization-based continual learning algorithms need an external representation and extra computation... | ["continual learning", "catastrophic forgetting"] | ABSTRACTContinual learning aims to learn new tasks without forgetting previously learnedones. This is especially challenging when one cannot access data from previoustasks and when the model has a fixed capacity. Current regularization-basedcontinual learning algorithms need an external representation and extra computa... | S1ewYgHaKr | Official Blind Review #2 | 6: Weak Accept | **** Post Rebuttal ****
I have read the author's response and other reviewers' comments. In light of comments by other reviewers, I am increasing the score. The paper reports decent empirical results in some challenging settings which might be useful to the continual learning community.
**** End ****
The paper pres... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Uncertainty-guided Continual Learning with Bayesian Neural Networks
### Paper Abstract
Continual learning aims to learn new tasks without forgetting previously learned o... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.