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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
SklEs2EYvS | ICLR.cc/2020/Conference | 2020 | Structural Multi-agent Learning | ["Kaiqian Han", "Liangliang Ren", "Jiwen Lu", "Jie Zhou"] | In this paper, we propose a multi-agent learning framework to model communication in complex multi-agent systems. Most existing multi-agent reinforcement learning methods require agents to exchange information with the environment or global manager to achieve effective and efficient interaction. We model the multi-agen... | ["Multi-agent Learning", "Communication", "Graph Network"] | ABSTRACTIn this paper, we propose a multi-agent learning framework to model communi-cation in complex multi-agent systems. Most existing multi-agent reinforcementlearning methods require agents to exchange information with the environmentor global manager to achieve effective and efficient interaction. We model themult... | SJgTq70y5S | Official Blind Review #3 | 6: Weak Accept | In this paper authors propose a framework for learning in multi-agent systems. Main contribution of the proposed framework is that it learns the communication structure between the agents, which improves the perfomance. In order to impose a structure authors use a graph neural network, which enables the agents to share... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Structural Multi-agent Learning
### Paper Abstract
In this paper, we propose a multi-agent learning framework to model communication in complex multi-agent systems. Most... | |
yeK_9wxRDbA | eswc-conferences.org/ESWC/2021/Conference/Resources_Track | 2021 | kgbench: A Collection of Knowledge Graph Datasets for Evaluating Relational and Multimodal Machine Learning | ["Peter Bloem", "Xander Wilcke", "Lucas van Berkel", "Victor de Boer"] | Graph neural networks and other machine learning models offer a promising direction for interpretable machine learning on relational and multimodal data. Until now, however, progress in this area is difficult to gauge. This is primarily due to a limited number of datasets with (a) a high enough number of labeled nodes ... | ["Machine Learning", "Graph Learning", "Benchmark"] | kgbench : A Collection of Knowledge GraphDatasets for Evaluating Relational andMultimodal Machine LearningPeter Bloem?[0000 000201895817], Xander Wilcke?[0000 000324158438],Lucas van Berkel?[0000 000225241279], and Victor de Boer[0000 00019079039X]Informatics institute, Vrije Universiteit Amsterdam, the Netherlands{p.b... | v2voV20if-J | A valuable collection of datasets for node classification and link prediction in RDF graphs. | 1: Weak Accept | The paper describes a resource called "kgbench" that is a collection of evaluation datasets primarily meant to be used for node labeling tasks on RDF graphs, although it can also be used for link prediction.
The datasets and the code to load them are all available on GitHub (https://github.com/pbloem/kgbench). The Git... | 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
kgbench: A Collection of Knowledge Graph Datasets for Evaluating Relational and Multimodal Machine Learning
### Paper Abstract
Graph neural networks and other machine le... |
HJ94fqApW | ICLR.cc/2018/Conference | 2018 | Rethinking the Smaller-Norm-Less-Informative Assumption in Channel Pruning of Convolution Layers | ["Jianbo Ye", "Xin Lu", "Zhe Lin", "James Z. Wang"] | Model pruning has become a useful technique that improves the computational efficiency of deep learning, making it possible to deploy solutions in resource-limited scenarios. A widely-used practice in relevant work assumes that a smaller-norm parameter or feature plays a less informative role at the inference time. In ... | ["model pruning", "batch normalization", "convolutional neural network", "ISTA"] | ABSTRACTModel pruning has become a useful technique that improves the computationalefficiency of deep learning, making it possible to deploy solutions in resource-limited scenarios. A widely-used practice in relevant work assumes that a smaller-norm parameter or feature plays a less informative role at the inference ti... | B1KcBUqlz | A pruning approach based on the batch normalization layer. The algorithm is easy to reproduce and seems to obtain interesting results. Sensibility analysis would be nice | 6: Marginally above acceptance threshold | This paper proposes an interesting approach to prune a deep model from a computational point of view. The idea is quite simple as pruning using the connection in the batch norm layer. It is interesting to add the memory cost per channel into the optimization process.
The paper suggests normal pruning does not necess... | 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
Rethinking the Smaller-Norm-Less-Informative Assumption in Channel Pruning of Convolution Layers
### Paper Abstract
Model pruning has become a useful technique that impr... |
fhxHhXTnHc | KDD.org/2023/Workshop/epiDAMIK | 2023 | Accurate Measures of Vaccination and Concerns of Vaccine Holdouts from Web Search Logs | ["Serina Chang", "Adam Fourney", "Eric Horvitz"] | To design effective vaccine policies, policymakers need detailed data about who has been vaccinated, who is holding out, and why. However, existing data in the US are insufficient: reported vaccination rates are often delayed or missing, and surveys of vaccine hesitancy are limited by high-level questions and self-repo... | ["COVID-19", "vaccination", "health behaviors", "misinformation", "search logs", "graph machine learning"] | ABSTRACTTo design effective vaccine policies, policymakers need detaileddata about who has been vaccinated, who is holding out, and why.However, existing data in the US are insufficient: reported vacci-nation rates are often delayed or missing, and surveys of vaccinehesitancy are limited by high-level questions and sel... | ZvkFh9VHIQ | The authors did an in-depth analysis of the search logs related to the vaccines to detect an individual’s vaccine intent and further discovered insights on the behavioral difference of (i) early vaccine adaptors and (ii) vaccine-resistant groups. Overall, the paper is well written, is original, and would help the commu... | 5: Top 50% of accepted papers, clear accept | Summary (Long)
- The authors did an in-depth analysis of the search logs related to the vaccines to detect an individual’s vaccine intent and further discovered insights on the behavioral difference of (i) early vaccine adaptors and (ii) vaccine-resistant groups. Their pipeline of the vaccine intent classifier includes... | 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
Accurate Measures of Vaccination and Concerns of Vaccine Holdouts from Web Search Logs
### Paper Abstract
To design effective vaccine policies, policymakers need detaile... |
xQx1O7WXSA | robot-learning.org/CoRL/2023/Conference | 2023 | Expansive Latent Planning for Sparse Reward Offline Reinforcement Learning | ["Robert Gieselmann", "Florian T. Pokorny"] | Sampling-based motion planning algorithms excel at searching global solution paths in geometrically complex settings. However, classical approaches, such as RRT, are difficult to scale beyond low-dimensional search spaces and rely on privileged knowledge e.g. about collision detection and underlying state distances. In... | ["model-based reinforcement learning", "planning", "robot manipulation"] | Expansive Latent Planning for Sparse RewardOffline Reinforcement LearningRobert GieselmannKTH Royal Institute of TechnologySwedenrobgie@kth.seFlorian T. PokornyKTH Royal Institute of TechnologySwedenfpokorny@kth.seAbstract: Sampling-based motion planning algorithms excel at searching globalsolution paths in geometrical... | 0DYRt1sa9k | **Clarity**:
Although the writing itself is of good quality, the presentation is frequently confusing. In many places, information feels scattered, unclear, or missing throughout the paper. For instance, on line 194, it's not clear whether hindsight goal relabelling is used to train both policies, or just one, and whi... | 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
Expansive Latent Planning for Sparse Reward Offline Reinforcement Learning
### Paper Abstract
Sampling-based motion planning algorithms excel at searching global solutio... | ||
AHOs7Sm5H7R | ICLR.cc/2021/Conference | 2021 | Towards Resolving the Implicit Bias of Gradient Descent for Matrix Factorization: Greedy Low-Rank Learning | ["Zhiyuan Li", "Yuping Luo", "Kaifeng Lyu"] | Matrix factorization is a simple and natural test-bed to investigate the implicit regularization of gradient descent. Gunasekar et al. (2017) conjectured that gradient flow with infinitesimal initialization converges to the solution that minimizes the nuclear norm, but a series of recent papers argued that the language... | ["matrix factorization", "gradient descent", "implicit regularization", "implicit bias"] | ABSTRACTMatrix factorization is a simple and natural test-bed to investigate the implicitregularization of gradient descent. Gunasekar et al. (2017) conjectured that Gradi-ent Flow with infinitesimal initialization converges to the solution that minimizesthe nuclear norm, but a series of recent papers argued that the l... | a2n8lpZ_tbt | Interesting paper; unfortunately there is a mathematical flaw | 6: Marginally above acceptance threshold | This paper analyses the implicit regularization in matrix factorization. The author suggest that for infinitesimal initialization the implicit regularization can be equivalently understood as iteratively solving matrix factorization with increasing rank until the rank is high enough such that the model can be fitted to... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Towards Resolving the Implicit Bias of Gradient Descent for Matrix Factorization: Greedy Low-Rank Learning
### Paper Abstract
Matrix factorization is a simple and natura... |
HJgFW6EKvH | ICLR.cc/2020/Conference | 2020 | Generating Robust Audio Adversarial Examples using Iterative Proportional Clipping | ["Hongting Zhang", "Qiben Yan", "Pan Zhou"] | Audio adversarial examples, imperceptible to humans, have been constructed to attack automatic speech recognition (ASR) systems. However, the adversarial examples generated by existing approaches usually involve notable noise, especially during the periods of silence and pauses, which may lead to the detection of such ... | ["audio adversarial examples", "attack", "machine learning"] | ABSTRACTAudio adversarial examples, imperceptible to humans, have been constructed toattack automatic speech recognition (ASR) systems. However, the adversarial ex-amples generated by existing approaches usually involve notable noise, especiallyduring the periods of silence and pauses, which may lead to the detection o... | S1ls_R5pKr | Official Blind Review #2 | 3: Weak Reject | This paper introduces an adversarial attack mechanism, Iterative Proportional Clipping (IPC), based on a differentiable MFCC feature extractor. In contrast to other audio attacks, the authors claim that the perturbations are well masked and are much less audible.
They also claim to propose the first attack mechanism 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
Generating Robust Audio Adversarial Examples using Iterative Proportional Clipping
### Paper Abstract
Audio adversarial examples, imperceptible to humans, have been cons... | |
SyxvSiCcFQ | ICLR.cc/2019/Conference | 2019 | Neural Network Cost Landscapes as Quantum States | ["Abdulah Fawaz", "Sebastien Piat", "Paul Klein", "Peter Mountney", "Simone Severini"] | Quantum computers promise significant advantages over classical computers for a number of different applications. We show that the complete loss function landscape of a neural network can be represented as the quantum state output by a quantum computer. We demonstrate this explicitly for a binary neural network and, fu... | ["quantum", "neural networks", "meta-learning"] | ABSTRACTQuantum computers promise significant advantages over classical computers fora number of different applications. We show that the complete loss function land-scape of a neural network can be represented as the quantum state output by aquantum computer. We demonstrate this explicitly for a binary neural networka... | ByxRMTMc3m | More appropriate for quantum computing literature | 3: Clear rejection | Review of "Neural Network Cost Landscapes as Quantum States"
Paper summary:
The paper proposes a new algorithm "quantum amplitude amplification"
for training and model selection in binary neural networks. (in which
both weights and activations are restricted to the set -1, 1)
Section 2 references related work and gi... | 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 Network Cost Landscapes as Quantum States
### Paper Abstract
Quantum computers promise significant advantages over classical computers for a number of different a... |
02X3kfP6W4 | MIDL.io/2020/Conference | 2020 | Assessing the validity of saliency maps for abnormality localization in medical imaging | ["Nishanth Thumbavanam Arun", "Nathan Gaw", "Praveer Singh", "Ken Chang", "Katharina Viktoria Hoebel", "Jay Patel", "Mishka Gidwani", "Jayashree Kalpathy-Cramer"] | Saliency maps have become a widely used method to assess which areas of the input image are most pertinent to the prediction of a trained neural network. However, in the context of medical imaging, there is no study to our knowledge that has examined the efficacy of these techniques and quantified them using overlap w... | ["Saliency maps", "localization", "anomaly detection", "medical imaging", "deep learning."] | Medical Imaging with Deep Learning 2020 Short Paper { MIDL 2020Assessing the validity of saliency maps for abnormalitylocalization in medical imagingNishanth Thumbavanam Arun1Nathan Gaw2Praveer Singh1Ken Chang1Katharina Viktoria Hoebel1Jay Patel1Mishka Gidwani1Jayashree Kalpathy-Cramer1kalpathy@nmr.mgh.harvard.edu1Athi... | IC6D-Qll2 | Assessing saliency map validity | 3: Weak accept | This paper concerns the assessment of saliency map validity. It was shown that GradCAM is superior to other methods in terms of model and parameter randomization. This is a useful results, as the interpretability that saliency mapping enables is becoming more and more important to help visualize why deep networks are ... | 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
Assessing the validity of saliency maps for abnormality localization in medical imaging
### Paper Abstract
Saliency maps have become a widely used method to assess which... |
kMB2WAfisY | ICRA.org/2022/Workshop/Contact-Rich | 2022 | Pathologies and Challenges of Using Differentiable Simulators in Policy Optimization for Contact-Rich Manipulation | ["H.J. Terry Suh", "Max Simchowitz", "Kaiqing Zhang", "Tao Pang", "Russ Tedrake"] | Policy search methods in Reinforcement Learning (RL) have shown impressive results in contact-rich tasks such as dexterous manipulation. However, the high variance of zero-order Monte-Carlo gradient estimates results in slow convergence and a requirement for a high number of samples. By replacing these zero-order gradi... | ["Policy Optimization", "Differentiable Simulation", "Policy Gradients", "Contact-rich manipulation"] | Pathologies and Challenges of Using Differentiable Simulators inPolicy Optimization for Contact-Rich ManipulationH.J. Terry Suh, Max Simchowitz, Kaiqing Zhang, Tao Pang, Russ TedrakeAbstract —Policy search methods in Reinforcement Learning(RL) have shown impressive results in contact-rich tasks such asdexterous manipul... | dPkVIM1LOug | A nice study on using zero- vs. first-order gradients in policy learning that can be improved by adding qualitative evaluations of the resulting policies | 8: Top 50% of accepted papers, clear accept | This paper studies the use of differentiable simulators for gradient computations in policy optimization for contact-rich manipulation tasks.
Such simulators can provide first-order gradients in contrast to classical zero-order gradient estimation using samples.
The authors conduct various case studies to highlight 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
Pathologies and Challenges of Using Differentiable Simulators in Policy Optimization for Contact-Rich Manipulation
### Paper Abstract
Policy search methods in Reinforcem... |
H1ls_eSKPH | ICLR.cc/2020/Conference | 2020 | Overcoming Catastrophic Forgetting via Hessian-free Curvature Estimates | ["Leonid Butyrev", "Georgios Kontes", "Christoffer L\u00f6ffler", "Christopher Mutschler"] | Learning neural networks with gradient descent over a long sequence of tasks is problematic as their fine-tuning to new tasks overwrites the network weights that are important for previous tasks. This leads to a poor performance on old tasks – a phenomenon framed as catastrophic forgetting. While early approaches use ... | ["catastrophic forgetting", "multi-task learning", "continual learning"] | ABSTRACTLearning neural networks with gradient descent over a long sequence of tasks isproblematic as their fine-tuning to new tasks overwrites the network weights thatare important for previous tasks. This leads to a poor performance on old tasks – aphenomenon framed as catastrophic forgetting . While early approaches... | rkelpoAaKS | Official Blind Review #2 | 3: Weak Reject | The paper focuses on alleviating the problem of "catastrophic forgetting", exhibited by neural networks learned with gradient-based algorithms over long sequence of tasks. In such learning scenarios, tuning of parameters over the new tasks lead to degradation of performance over the old tasks as the parameters importan... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Overcoming Catastrophic Forgetting via Hessian-free Curvature Estimates
### Paper Abstract
Learning neural networks with gradient descent over a long sequence of tasks i... | |
SI7fwS_Uqxc | ICLR.cc/2022/Workshop/OSC | 2022 | Iterative Causal Discovery with Active Interventions | ["Anonymous"] | Multi-entity causal discovery is a fundamental problem in machine learning. Understanding the underlying causal relations is important for counterfactual reasoning and robustness. However, the causal structure is only identifiable up to a Markov Equivalence Class with observational data. In real life, it is usually har... | ["causal discovery", "causal inference", "multi-entity prediction", "intervention"] | ABSTRACTMulti-entity causal discovery is a fundamental problem in machine learning. Un-derstanding the underlying causal relations is important for counterfactual rea-soning and robustness. However, the causal structure is only identifiable up to aMarkov Equivalence Class with observational data. In real life, it is us... | Bego8R_0Z5 | Simple algorithm, unclear how it merges interventional data (possibly wrong), does not cite or compare related work | 1: Technically incorrect, off-topic, or other significant issues - rejection | While I think the paper addresses a potentially interesting problem, I have several issues with its current version.
The first issue is that paper proposes a method for intervention design, a well-known problem in causal discovery, which does not cite any of the work on intervention design or any work in which one co... | 3: The reviewer is 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
Iterative Causal Discovery with Active Interventions
### Paper Abstract
Multi-entity causal discovery is a fundamental problem in machine learning. Understanding the und... |
HJgeEh09KQ | ICLR.cc/2019/Conference | 2019 | Boosting Robustness Certification of Neural Networks | ["Gagandeep Singh", "Timon Gehr", "Markus P\u00fcschel", "Martin Vechev"] | We present a novel approach for the certification of neural networks against adversarial perturbations which combines scalable overapproximation methods with precise (mixed integer) linear programming. This results in significantly better precision than state-of-the-art verifiers on challenging feedforward and convolut... | ["Robustness certification", "Adversarial Attacks", "Abstract Interpretation", "MILP Solvers", "Verification of Neural Networks"] | ABSTRACTWe present a novel approach for the certification of neural networks against ad-versarial perturbations which combines scalable overapproximation methods withprecise (mixed integer) linear programming. This results in significantly betterprecision than state-of-the-art verifiers on challenging feedforward and c... | S1xULFL9n7 | a refined approach for robust verification, but experimental part could be stronger | 4: Ok but not good enough - rejection | This paper introduces a verifier that obtains improvement on both the precision of the incomplete verifiers and the scalability of the complete verifiers. The proposed approaches combines over-parameterization, mixed integer linear programming, and linear programming relaxation.
This paper is well written and well or... | 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
Boosting Robustness Certification of Neural Networks
### Paper Abstract
We present a novel approach for the certification of neural networks against adversarial perturba... |
Hksj2WWAW | ICLR.cc/2018/Conference | 2018 | Combining Symbolic Expressions and Black-box Function Evaluations in Neural Programs | ["Forough Arabshahi", "Sameer Singh", "Animashree Anandkumar"] | Neural programming involves training neural networks to learn programs, mathematics, or logic from data. Previous works have failed to achieve good generalization performance, especially on problems and programs with high complexity or on large domains. This is because they mostly rely either on black-box function eval... | ["symbolic reasoning", "mathematical equations", "recursive neural networks", "neural programing"] | ABSTRACTNeural programming involves training neural networks to learn programs, math-ematics, or logic from data. Previous works have failed to achieve good general-ization performance, especially on problems and programs with high complexityor on large domains. This is because they mostly rely either on black-box func... | BywjthjgG | A review of "Combining Symbolic Expressions..." | 5: Marginally below acceptance threshold | This paper proposes a model that predicts the validity of a mathematical expression (containing trigonometric or elementary algebraic expressions) using a recursive neural network (TreeLSTM). The idea is to take the parse tree of the expression, which is converted to the recursive neural network architecture, where we... | 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
Combining Symbolic Expressions and Black-box Function Evaluations in Neural Programs
### Paper Abstract
Neural programming involves training neural networks to learn pro... |
H1xLsjAqtX | ICLR.cc/2019/Conference | 2019 | Robust Text Classifier on Test-Time Budgets | ["Md Rizwan Parvez", "Tolga Bolukbasi", "Kai-Wei Chang", "Venkatesh Saligrama"] | In this paper, we design a generic framework for learning a robust text classification model that achieves accuracy comparable to standard full models under test-time
budget constraints. We take a different approach from existing methods and learn to dynamically delete a large fraction of unimportant words by a low-com... | ["Data Aggregation", "Budget Learning", "Speed Up", "Faster Inference", "Robust Classifier"] | ABSTRACTIn this paper, we design a generic framework for learning a robust text classi-fication model that achieves accuracy comparable to standard full models undertest-time budget constraints. We take a different approach from existing methodsand learn to dynamically delete a large fraction of unimportant words by a ... | B1xKKvdT2Q | The idea of jointly training a word/phrase/sentence mask with the classifier at training time (with some added complexity) to get faster predictions at test time is interesting; however, current results are largely preliminary and requires refinement of performance and understanding | 4: Ok but not good enough - rejection | The authors consider the setting of training a RNN-based text classification where there is a resource restriction on test-time prediction (e.g., time). Conceptually, the approach is essentially a masking mechanism to reduce the number of words/phrases/sentences used in the prediction at test time followed by a classif... | 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 Text Classifier on Test-Time Budgets
### Paper Abstract
In this paper, we design a generic framework for learning a robust text classification model that achieves... |
HkUfnZFt1Rw | ICLR.cc/2021/Conference | 2021 | Dissecting graph measures performance for node clustering in LFR parameter space | ["Vladimir Ivashkin", "Pavel Chebotarev"] | Graph measures can be used for graph node clustering using metric clustering algorithms. There are multiple measures applicable to this task, and which one performs better is an open question. We study the performance of 25 graph measures on generated graphs with different parameters. While usually measure comparisons ... | ["graph theory", "graph measures", "kernel k-means", "clustering"] | ABSTRACTGraph measures can be used for graph node clustering using metric clustering al-gorithms. There are multiple measures applicable to this task, and which one per-forms better is an open question. We study the performance of 25 graph measureson generated graphs with different parameters. While usually measure com... | 89WhVX3KfR4 | Interesting empirical analysis of different graph measures for graph clustering. The choice of LFR though limits the generalization of the observations. Also, the proposed approach lacks theoretical justification. | 3: Clear rejection | The paper deals with the problem of community detection on graphs, examining the impact of graph measures. To do so, the paper proposes an experimental framework where clustering is achieved using the kernel k-means algorithm, and the performance of graph measures is examined on various instances of artificially genera... | 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
Dissecting graph measures performance for node clustering in LFR parameter space
### Paper Abstract
Graph measures can be used for graph node clustering using metric clu... |
ByxmXnA9FQ | ICLR.cc/2019/Conference | 2019 | A Variational Dirichlet Framework for Out-of-Distribution Detection | ["Wenhu Chen", "Yilin Shen", "William Wang", "Hongxia Jin"] | With the recently rapid development in deep learning, deep neural networks have been widely adopted in many real-life applications. However, deep neural networks are also known to have very little control over its uncertainty for test examples, which potentially causes very harmful and annoying consequences in practica... | ["out-of-distribution detection", "variational inference", "Dirichlet distribution", "deep learning", "uncertainty measure"] | ABSTRACTWith the recently rapid development in deep learning, deep neural networks havebeen widely adopted in many real-life applications. However, deep neural net-works are also known to have very little control over its uncertainty for test ex-amples, which potentially causes very harmful and annoying consequences in... | HJghUGPw2Q | Not well motivated. | 5: Marginally below acceptance threshold | This paper proposes a new framework for out-of-distribution detection, based on variational inference and a prior Dirichlet distribution. The Dirichlet distribution is presented, and the way it is used withing the method is discussed (i.e. clipping, scaling, etc). Experiments on several datasets and comparison with the... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
A Variational Dirichlet Framework for Out-of-Distribution Detection
### Paper Abstract
With the recently rapid development in deep learning, deep neural networks have be... |
xdwZoGTiMJ- | graphicsinterface.org/Graphics_Interface/2022/Conference | 2022 | A Design Framework for Contextual and Embedded Information Visualizations in Spatial Augmented Reality | ["Nikhita Joshi", "Matthew Lakier", "Daniel Vogel", "Jian Zhao"] | Spatial augmented reality (SAR) displays digital content in a real environment in ways that are situated, peripheral, and viewable by multiple people. These capabilities change how embedded information visualizations are used, designed, and experienced. But a comprehensive design framework that captures the specific ch... | ["information visualization", "embedded visualizations", "spatial augmented reality", "design frameworks"] | ABSTRACTSpatial augmented reality (SAR) displays digital content in a realenvironment in ways that are situated, peripheral, and viewable bymultiple people. These capabilities change how embedded infor-mation visualizations are used, designed, and experienced. But acomprehensive design framework that captures the speci... | Cu9sckN-otp | This is a nice design study/model paper that is focused on an emerging aspect of AR, but it needs a bit of work. | 8: Top 50% of accepted papers, clear accept | This paper presents a design framework for spatial augmented reality. It includes an analysis of two existing works and a set of mock-up designs, each of which are described in the context of this framework.
This paper is a combination of a "design study" and "model" (following Munzner's classification of InfoVis pape... | 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 Design Framework for Contextual and Embedded Information Visualizations in Spatial Augmented Reality
### Paper Abstract
Spatial augmented reality (SAR) displays digita... |
Vbfr1jiMxYS | logconference.io/LOG/2022/Conference | 2022 | PatchGT: Transformer over Non-trainable Clusters for Learning Graph Representations | ["Han Gao", "Xu Han", "Jiaoyang Huang", "Jian-Xun Wang", "Liping Liu"] | Recently the Transformer structure has shown good performances in graph learning tasks. However, these Transformer models directly work on graph nodes and may have difficulties learning high-level information. Inspired by the vision transformer, which applies to image patches, we propose a new Transformer-based graph n... | ["Graph Neural Networks", "Unlearnable spectral clustering method", "Transformer on graphs"] | PatchGT: Transformer over Non-trainable Clusters forLearning Graph RepresentationsHan Gao∗University of Notre Damehgao1@nd.eduXu Han∗Tufts UniversityXu.Han@tufts.eduJiaoyang Huang∗University of Pennsylvaniahuangjy@wharton.upenn.eduJian-Xun WangUniversity of Notre Damejwang33@nd.eduLi-Ping LiuTufts UniversityLiping.Liu@... | z5_Nij6nNS | *Summary:* The paper proposes a graph transformer-based model for graph prediction tasks by taking inspiration from vision transformer models. In a first step, the presented approach clusters each graph using spectral clustering, resulting in patches of nodes. Patch representations are computed using a GCN both on 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
PatchGT: Transformer over Non-trainable Clusters for Learning Graph Representations
### Paper Abstract
Recently the Transformer structure has shown good performances in ... | |||
HyGDdsCcFQ | ICLR.cc/2019/Conference | 2019 | Better Generalization with On-the-fly Dataset Denoising | ["Jiaming Song", "Tengyu Ma", "Michael Auli", "Yann Dauphin"] | Memorization in over-parameterized neural networks can severely hurt generalization in the presence of mislabeled examples. However, mislabeled examples are to hard avoid in extremely large datasets. We address this problem using the implicit regularization effect of stochastic gradient descent with large learning rate... | ["dataset denoising", "supervised learning", "implicit regularization"] | ABSTRACTMemorization in over-parameterized neural networks can severely hurt general-ization in the presence of mislabeled examples. However, mislabeled examplesare to hard avoid in extremely large datasets. We address this problem usingthe implicit regularization effect of stochastic gradient descent with large learn-... | Hyx00AV5n7 | BETTER GENERALIZATION WITH ON-THE-FLY DATASET DENOISING | 6: Marginally above acceptance threshold | This paper presents ODD, a method that rejects incorrectly labeled / noisy examples from training on the fly. The motivation is sound, that with the capacity of modern neural networks, it's easy to memorize the mislabeled data and thus hurt generalization. If we could reject such mislabeled data, we may be able to get ... | 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
Better Generalization with On-the-fly Dataset Denoising
### Paper Abstract
Memorization in over-parameterized neural networks can severely hurt generalization in the pre... |
rkeu30EtvS | ICLR.cc/2020/Conference | 2020 | Network Deconvolution | ["Chengxi Ye", "Matthew Evanusa", "Hua He", "Anton Mitrokhin", "Tom Goldstein", "James A. Yorke", "Cornelia Fermuller", "Yiannis Aloimonos"] | Convolution is a central operation in Convolutional Neural Networks (CNNs), which applies a kernel to overlapping regions shifted across the image. However, because of the strong correlations in real-world image data, convolutional kernels are in effect re-learning redundant data. In this work, we show that this redun... | ["convolutional networks", "network deconvolution", "whitening"] | ABSTRACTConvolution is a central operation in Convolutional Neural Networks (CNNs),which applies a kernel to overlapping regions shifted across the image. However,because of the strong correlations in real-world image data, convolutional kernelsare in effect re-learning redundant data. In this work, we show that this r... | r1lY9K6aYS | Official Blind Review #1 | 8: Accept | This paper proposes an operation for removing the pixel-wise and channel-wise correlations of input features. This operation can be considered as a generalization form of the former proposed decorrelated batch normalization. The approach has a well-sounded neurological inspired motivation and a solid explanation 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
Network Deconvolution
### Paper Abstract
Convolution is a central operation in Convolutional Neural Networks (CNNs), which applies a kernel to overlapping regions shifte... | |
Dp5B1YhYlwY | AAAI.org/2022/Workshop/AdvML | 2022 | A Practical and Stealthy Adversarial Attack for Cyber-Physical Applications | ["Yifu Wu", "Jin Wei-Kocsis"] | Adversarial perturbations on misleading a well-trained machine learning (ML) model have been studied in computer vision (CV) and other related application areas. However, there is very limited focus on studying the impact of adversarial perturbations on ML models used in data-driven cyber-physical systems (CPSs) that n... | ["Adversarial Attack", "Cyber-Physical Systems", "Domain-Knowledge Constraints"] | A Practical and Stealthy Adversarial Attack for Cyber-Physical ApplicationsYifu Wu, Jin Wei-KocsisDepartment of Computer and Information Technology401 N Grant StreetWest Lafayette, Indiana 47907fwu1584, kocsis0g@purdue.eduAbstractAdversarial perturbations on misleading a well-trained ma-chine learning (ML) model have b... | XvoG6OtRhMd | Adversarial attacks for cyber-physical system. | 7: Good paper, accept | This paper proposes a novel adversarial attack method for cyber-physical systems (CSPs). An encoder-decoder model is first trained to learn and interpret the features of sensing data. The encoder is further used to train a discriminator. Finally, the encoder and discriminator constitute a surrogate model to generate ad... | 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 Practical and Stealthy Adversarial Attack for Cyber-Physical Applications
### Paper Abstract
Adversarial perturbations on misleading a well-trained machine learning (M... |
dGOeF3y_Weh | aclweb.org/ACL/2020/Workshop/NLP-COVID | 2020 | An Analysis of BERT FAQ Retrieval Models for COVID-19 Infobot | ["Shuo Sun", "Jo\u00e3o Sedoc"] | The outbreak of the COVID-19 pandemic has caused tremendous amounts of suffering and deaths around the world and greatly affected the lives of humanity. As the world sees more infected cases every day, the need and demand for reliable and up-to-date information on COVID-19 have never been higher. While recent pre-train... | ["COVID-19", "BERT", "QA retrieval", "infobot"] | An Analysis of BERT FAQ Retrieval Models for COVID-19 InfobotShuo SunJohns Hopkins Universityssun32@jhu.eduJo ̃ao SedocJohns Hopkins Universityjsedoc@jhu.eduAbstractThe outbreak of the COVID-19 pandemic hascaused tremendous amounts of suffering anddeaths around the world and greatly affectedthe lives of humanity. As th... | tA4dtD4iDn | Dataset can be of interest, but creation process is strange | 4: Ok but not good enough - rejection | This paper presents a pipeline for question-answering for covid-19. They create a novel dataset of covid-19 related questions and answers from multiple websites. Manual annotation of a subset of answers is collected. Various models, including BERT, Roberta, CovidBERT, and TwitterBERT, are evaluated on the dataset in un... | 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
An Analysis of BERT FAQ Retrieval Models for COVID-19 Infobot
### Paper Abstract
The outbreak of the COVID-19 pandemic has caused tremendous amounts of suffering and dea... |
Byx91R4twB | ICLR.cc/2020/Conference | 2020 | Adversarial Video Generation on Complex Datasets | ["Aidan Clark", "Jeff Donahue", "Karen Simonyan"] | Generative models of natural images have progressed towards high fidelity samples by the strong leveraging of scale. We attempt to carry this success to the field of video modeling by showing that large Generative Adversarial Networks trained on the complex Kinetics-600 dataset are able to produce video samples of subs... | ["GAN", "generative model", "generative adversarial network", "video prediction"] | ABSTRACTGenerative models of natural images have progressed towards high fidelity samplesby the strong leveraging of scale. We attempt to carry this success to the field ofvideo modeling by showing that large Generative Adversarial Networks trained onthe complex Kinetics-600 dataset are able to produce video samples of... | rklJIn8wYS | Official Blind Review #1 | 3: Weak Reject |
Summary:
This paper presents a method for training a generative adversarial network on high resolution videos and complex datasets. They propose decomposing the discriminator in Adversarial Networks into a spatial and temporal discriminator similar to previous works, however, the temporal discriminator downsamples 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
Adversarial Video Generation on Complex Datasets
### Paper Abstract
Generative models of natural images have progressed towards high fidelity samples by the strong lever... | |
p8agn6bmTbr | ICLR.cc/2021/Conference | 2021 | Usable Information and Evolution of Optimal Representations During Training | ["Michael Kleinman", "Alessandro Achille", "Daksh Idnani", "Jonathan Kao"] | We introduce a notion of usable information contained in the representation learned by a deep network, and use it to study how optimal representations for the task emerge during training. We show that the implicit regularization coming from training with Stochastic Gradient Descent with a high learning-rate and small b... | ["Usable Information", "Representation Learning", "Learning Dynamics", "Initialization", "SGD"] | ABSTRACTWe introduce a notion of usable information contained in the representationlearned by a deep network, and use it to study how optimal representations forthe task emerge during training. We show that the implicit regularization com-ing from training with Stochastic Gradient Descent with a high learning-rate ands... | iB4JfblGLD | Interesting findings, but not enough evidence of generality | 7: Good paper, accept | The paper studies how initialization and the implicit regularization of SGD affect the training dynamics of neural networks in terms of minimality and sufficiency of learned representations. The main findings are that 1) SGD with random initialization learns almost minimal and sufficient representations and 2) SGD with... | 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
Usable Information and Evolution of Optimal Representations During Training
### Paper Abstract
We introduce a notion of usable information contained in the representatio... |
HkljrTEKvr | ICLR.cc/2020/Conference | 2020 | Hierarchical Image-to-image Translation with Nested Distributions Modeling | ["Shishi Qiao", "Ruiping Wang", "Shiguang Shan", "Xilin Chen"] | Unpaired image-to-image translation among category domains has achieved remarkable success in past decades. Recent studies mainly focus on two challenges. For one thing, such translation is inherently multimodal due to variations of domain-specific information (e.g., the domain of house cat has multiple fine-grained su... | ["translation", "domains", "variations", "hierarchical", "nested distributions", "domain", "model", "multimodal", "unpaired", "category domains"] | ABSTRACTUnpaired image-to-image translation among category domains has achieved re-markable success in past decades. Recent studies mainly focus on two chal-lenges. For one thing, such translation is inherently multimodal due to variationsof domain-specific information (e.g., the domain of house cat has multiple fine-g... | r1e7p0CaKS | Official Blind Review #2 | 3: Weak Reject | This paper proposes a way to utilize the domain hierarchical structure information to help the general multimodal I2I (called joint multi-domain and multi-modal in the paper). The proposed method applies the hierarchy-regularized learning to learn the nested distributions which are reasonable for the data with a well-d... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Hierarchical Image-to-image Translation with Nested Distributions Modeling
### Paper Abstract
Unpaired image-to-image translation among category domains has achieved rem... | |
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 ... | rJxC9rZ2FH | Official Blind Review #2 | 3: Weak Reject | This paper introduces two novel techniques to help long term signal propagation in RNNs. One is an initialization strategy which uses inverse sigmoid function to avoid the decay of the contribution of the input earlier in time and another is a new design of a refine gate which pushes the value of the gate closer to 0 o... | <|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... | |
rkl-Nph79V | icaps-conference.org/ICAPS/2019/Workshop/XAIP | 2019 | Towards Explainable Planning as a Service | ["Anonymous"] | Explainable AI is an important area of research within which Explainable Planning is an emerging topic. In this paper, we argue that Explainable Planning can be designed as a service – that is, as a wrapper around an existing planning system that utilises the existing planner to assist in answering contrastive question... | ["Explainable Planning", "Explainable AI", "Contrastive Explanations"] | Towards Explainable AI Planning as a ServiceMichael Cashmore1, Anna Collins1, Benjamin Krarup1, Senka Krivic1,Daniele Magazzeni1, David Smith1King’s College London, United Kingdom, ffirstname.surnameg@kcl.ac.ukAbstractExplainable AI is an important area of research within whichExplainable Planning is an emerging topic.... | ryeMrlnP3N | Very good paper that perfect fits the workshop's aim | 5: Top 15% of accepted papers, strong accept |
The paper presents the Explainable Planning (XAIP) framework - introduced in the past IJCAI-XAI-17 workshop - to be provided "as a service". The key element of XAIP is the ability for the user to formulate contrastive questions (i.e., why did you do action A rather than B?). This would allow the user understanding th... | 3: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Towards Explainable Planning as a Service
### Paper Abstract
Explainable AI is an important area of research within which Explainable Planning is an emerging topic. In t... |
ryex8CEKPr | ICLR.cc/2020/Conference | 2020 | Knockoff-Inspired Feature Selection via Generative Models | ["Marco F. Duarte", "Siwei Feng"] | We propose a feature selection algorithm for supervised learning inspired by the recently introduced
knockoff framework for variable selection in statistical regression. While variable selection in statistics aims
to distinguish between true and false predictors, feature selection in machine learning aims to reduce t... | ["feature selection", "variable selection", "knockoff variables", "supervised learning"] | ABSTRACTWe propose a feature selection algorithm for supervised learning inspired bythe recently introduced knockoff framework for variable selection in statisticalregression. While variable selection in statistics aims to distinguish betweentrue and false predictors, feature selection in machine learning aims to reduc... | Hye4vzrzFS | Official Blind Review #3 | 3: Weak Reject | This paper proposes a new feature selection method by integrating the knockoff procedure and generative models.
The paper is clearly written and easy to read. However, I have the following concerns:
- The motivation is not clear. The advantage of the knockoff procedure is that it can find relevant features (variables)... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Knockoff-Inspired Feature Selection via Generative Models
### Paper Abstract
We propose a feature selection algorithm for supervised learning inspired by the recently in... | |
BJxh2j0qYm | ICLR.cc/2019/Conference | 2019 | Dynamic Channel Pruning: Feature Boosting and Suppression | ["Xitong Gao", "Yiren Zhao", "\u0141ukasz Dudziak", "Robert Mullins", "Cheng-zhong Xu"] | Making deep convolutional neural networks more accurate typically comes at the cost of increased computational and memory resources. In this paper, we reduce this cost by exploiting the fact that the importance of features computed by convolutional layers is highly input-dependent, and propose feature boosting and supp... | ["dynamic network", "faster CNNs", "channel pruning"] | Published as a conference paper at ICLR 2019Dynamic Channel Pruning:Feature Boosting and SuppressionXitong Gao1, Yiren Zhao2, Lukasz Dudziak3, Robert Mullins4, Cheng-zhong Xu51Shenzhen Institutes of Advanced Technology, Shenzhen, China2;3;4University of Cambridge, Cambridge, UK5University of Macau, Macau, China1xt.gao@... | rklHZy28TX | Review for "Dynamic Channel Pruning: Feature Boosting and Suppression" | 6: Marginally above acceptance threshold | The authors propose a dynamic inference technique for accelerating neural network prediction with minimal accuracy loss. The technique prunes channels in an input-dependent way through the addition of auxiliary channel saliency prediction+pruning connections.
Pros:
- The paper is well-written and clearly explains 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
Dynamic Channel Pruning: Feature Boosting and Suppression
### Paper Abstract
Making deep convolutional neural networks more accurate typically comes at the cost of incre... |
6Y_I9qc6JBD | eswc-conferences.org/ESWC/2021/Conference/Resources_Track | 2021 | A Knowledge Organization System for the United Nations Sustainable Development Goals | ["Amit Joshi", "Luis Gerardo Gonzalez Morales", "Szymon Klarman", "Armando Stellato", "Aaron Helton", "Christopher Sean Lovell", "Artur Haczek"] | This paper presents a formal knowledge organization system (KOS) to represent the United Nations Sustainable Development Goals (SDGs). The SDGs are a set of objectives adopted by all United Nations member states in 2015 to achieve a better and sustainable future. The developed KOS consists of an ontology that models th... | ["United Nations", "Sustainable Development Goals", "SDGs", "Ontology", "Linked Data", "Knowledge Organization Systems", "Metadata"] | A Knowledge Organization System for theUnited Nations Sustainable Development GoalsAmit Joshi1, Luis Gonzalez Morales1, Szymon Klarman2, ArmandoStellato3, Aaron Helton4, Sean Lovell1, and Artur Haczek21United Nations, Department of Economic and Social Aairs, New York, NY, USAfjoshi6, gonzalezmorales, lovells g@un.org2E... | rRnysIFcvvX | Ontology for SDG supported by UN | -1: Weak Reject | I've read the response and disagree with many points about the use of SKOS. Some claims (for example about the punning) also apply in the other direction. The same with the mappings, if their use don't harm why using the skos with no skos concepts and other mappings with SKOS concepts. I would undersating using the sam... | 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 Knowledge Organization System for the United Nations Sustainable Development Goals
### Paper Abstract
This paper presents a formal knowledge organization system (KOS) ... |
Ek7qrYhJMbn | ICLR.cc/2021/Conference | 2021 | Central Server Free Federated Learning over Single-sided Trust Social Networks | ["Chaoyang He", "Conghui Tan", "Hanlin Tang", "Shuang Qiu", "Ji Liu"] | Federated learning has become increasingly important for modern machine learning, especially for data privacy-sensitive scenarios. Existing federated learning mostly adopts the central server-based architecture or centralized architecture. However, in many social network scenarios, centralized federated learning is not... | ["central", "free", "federated learning", "users", "central server", "trust social networks", "learning", "important", "modern machine learning"] | ABSTRACTFederated learning has become increasingly important for modern machine learn-ing, especially for data privacy-sensitive scenarios. Existing federated learningmostly adopts the central server-based architecture or centralized architecture.However, in many social network scenarios, centralized federated learning... | 0P760t2vx84 | This work focuses on single-sided decentralized federated learning. The authors propose a push sum-based algorithm to relax the symmetric matrix assumption, which leads to a flexible decentralized training on a directed network graph. The authors analyze the regret bound, as well as run some numerical experiments to de... | 8: Top 50% of accepted papers, clear accept | This work focuses on single-sided decentralized federated learning. The authors propose a push
sum-based algorithm to relax the symmetric matrix assumption, which leads to a flexible
decentralized training on a directed network graph. The authors analyze the regret bound, as well as
run some numerical experiments to de... | 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
Central Server Free Federated Learning over Single-sided Trust Social Networks
### Paper Abstract
Federated learning has become increasingly important for modern machine... |
EAYd7OOsWpB | icaps-conference.org/ICAPS/2020/Workshop/HSDIP | 2020 | Online Saturated Cost Partitioning for Classical Planning | ["Anonymous"] | Saturated cost partitioning is a general method for admissibly adding heuristic estimates for optimal state-space search. The algorithm strongly depends on the order in which it considers the heuristics. The strongest previous approach precomputes a set of diverse orders and the corresponding saturated cost partitionin... | ["optimal classical planning", "admissible heuristics", "cost partitioning"] | Online Saturated Cost Partitioning for Classical PlanningJendrik SeippUniversity of BaselBasel, Switzerlandjendrik.seipp@unibas.chAbstractSaturated cost partitioning is a general method for admissiblyadding heuristic estimates for optimal state-space search. Thealgorithm strongly depends on the order in which it consid... | NHfKsSW_xoN | Official review | 7: Good paper, accept | Brief summary of the paper:
This short paper introduces and discusses a novel online approach for computing saturated cost partitioning (scp) heuristics. A key component of scp heuristics is the ordering on the heuristics that are used to modify (saturate) the cost function. Previous algorithms that compute saturated ... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Online Saturated Cost Partitioning for Classical Planning
### Paper Abstract
Saturated cost partitioning is a general method for admissibly adding heuristic estimates fo... |
SJev6JBtvH | ICLR.cc/2020/Conference | 2020 | Testing For Typicality with Respect to an Ensemble of Learned Distributions | ["Forrest Laine", "Claire Tomlin"] | Good methods of performing anomaly detection on high-dimensional data sets are
needed, since algorithms which are trained on data are only expected to perform
well on data that is similar to the training data. There are theoretical results on the
ability to detect if a population of data is likely to come from a known ... | ["anomaly detection", "density estimation", "generative models"] | ABSTRACTGood methods of performing anomaly detection on high-dimensional data sets areneeded, since algorithms which are trained on data are only expected to performwell on data that is similar to the training data. There are theoretical results on theability to detect if a population of data is likely to come from a k... | rygtsAw7YB | Official Blind Review #3 | 3: Weak Reject | This paper proposes to use ensembles of estimated probability distributions in hypothesis testing for anomaly detection.
While the problem of density estimation with its application to anomaly detection is relevant, I have a number of concerns listed below:
- Overall, this paper is not clearly written and it is diffic... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Testing For Typicality with Respect to an Ensemble of Learned Distributions
### Paper Abstract
Good methods of performing anomaly detection on high-dimensional data sets... | |
GNv-TyWu3PY | ICLR.cc/2021/Conference | 2021 | Robust Learning for Congestion-Aware Routing | ["Sreenivas Gollapudi", "Kostas Kollias", "Benjamin Plaut", "Ameya Velingker"] | We consider the problem of routing users through a network with unknown congestion functions over an infinite time horizon. On each time step $t$, the algorithm receives a routing request and must select a valid path. For each edge $e$ in the selected path, the algorithm incurs a cost $c_e^t = f_e(x_e^t) + \eta_e^t$, w... | ["routing algorithms", "adversarial learning", "congestion functions"] | ABSTRACTWe consider the problem of routing users through a network with unknown con-gestion functions over an infinite time horizon. On each time step t, the algorithmreceives a routing request and must select a valid path. For each edge ein the se-lected path, the algorithm incurs a cost cte=fe(xte) +te, wherexteis th... | AzSntIGEXNx | An interesting extension of combinatorial semi-bandits | 7: Good paper, accept | This work introduces an interesting generalization of stochastic combinatorial semi-bandits for routing in a static graph. The main differences are: (1) the expected loss of an edge e is f_e(x^t_e) where the flow x^t_e is revealed at the beginning of each round (for each edge) and f_e is an unknown Lipschitz function (... | 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
Robust Learning for Congestion-Aware Routing
### Paper Abstract
We consider the problem of routing users through a network with unknown congestion functions over an infi... |
XXtHQy0d8Y | MICCAI.org/2021/Challenge/KiTS | 2021 | Transfer learning for KiTS21 Challenge | ["Xi Yang", "Zhang Jianpeng", "Xia Yong"] | Transfer learning has witnessed a recent surge of interest after proving successful in multiple applications. However, it highly re- lies on the quantity of annotated data. Constrained by the labor cost and expertise, it is hard to annotate sufficient organs and tumors at the voxel level for medical image segmentation.... | ["Transfer learning", "Limited annotation", "Kidney tumor segmentation"] | Transfer learning for KiTS21 ChallengeNo Author GivenNo Institute GivenAbstract. Transfer learning has witnessed a recent surge of interestafter proving successful in multiple applications. However, it highly re-lies on the quantity of annotated data. Constrained by the labor costand expertise, it is hard to annotate s... | 8c5vn31TFmG | Official Review | 7: Good paper, accept | The authors present a interesting approach in which they pre-trained their models on a combination of other publicly availble medical imaging datasets and then fine tuned them on the KiTS21 data. However, they did not explain how they synthesized KiTS21's multiple annotations per case during training. Please be sure to... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Transfer learning for KiTS21 Challenge
### Paper Abstract
Transfer learning has witnessed a recent surge of interest after proving successful in multiple applications. H... | |
HPB8mTbpvJ3 | MICCAI.org/2021/Workshop/COMPAY | 2021 | Creating small but meaningful representations of digital pathology images | ["Corentin Guerendel", "Phil Arnold", "Benjamin Torben-Nielsen"] | Representation learning is a popular application of deep learning where an object (e.g.,
an image) is converted into a lower-dimensional representation that still encodes relevant
features of the original object. In digital pathology, however, this is more difficult because
whole slide images (WSIs) are tiled before pr... | ["Digital pathology", "representation learning", "classification", "similar image search"] | Meaningful representations of pathology slidesCreating small but meaningful representations of digitalpathology imagesEditor:AbstractRepresentation learning is a popular application of deep learning where an object (e.g.,an image) is converted into a lower-dimensional representation that still encodes relevantfeatures ... | 0a6dFINXj9 | The authors of this paper introduce a method to create whole slide image-level representations by calculating the proportion of tiles in clusters. | 5: Marginally below acceptance threshold | The authors of this paper introduce a method to create whole slide image-level representations by calculating the proportion of tiles in clusters. They showed their representations can successfully classify tissue types in TCGA data, predict the brain tumor stage (lower-grade glioma vs glioblastroma), and retrieve whol... | 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
Creating small but meaningful representations of digital pathology images
### Paper Abstract
Representation learning is a popular application of deep learning where an o... |
S1gE6TEYDB | ICLR.cc/2020/Conference | 2020 | Global reasoning network for image super-resolution | ["Jiahui Zhang", "Bin Zhou", "Qingchang Tao", "Deqiang Wang"] | Recent image super-resolution(SR) studies leverage very deep convolutional neural networks and the rich hierarchical features they offered, which leads to better reconstruction performance than conventional methods. However, the small receptive fields in the up-sampling and reconstruction process of those models stop t... | ["Global reasoning network", "upsampling module", "graph model", "image super-resolution"] | ABSTRACTRecent image super-resolution(SR) studies leverage very deep convolutional neu-ral networks and the rich hierarchical features they offered, which leads to betterreconstruction performance than conventional methods. However, the small re-ceptive fields in the up-sampling and reconstruction process of those mode... | rJlWt142KH | Official Blind Review #1 | 1: Reject | How to justify the receptive field is small? As the low-resolution image is small already, a network can definitely see holistic image as the receptive field is large enough to cover the whole image.
It seems contradictory to super-resolution process with the finding ``the brain generates the images we see based on 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
Global reasoning network for image super-resolution
### Paper Abstract
Recent image super-resolution(SR) studies leverage very deep convolutional neural networks and the... | |
HJxdTxHYvB | ICLR.cc/2020/Conference | 2020 | BREAKING CERTIFIED DEFENSES: SEMANTIC ADVERSARIAL EXAMPLES WITH SPOOFED ROBUSTNESS CERTIFICATES | ["Amin Ghiasi", "Ali Shafahi", "Tom Goldstein"] | Defenses against adversarial attacks can be classified into certified and non-certified. Certifiable defenses make networks robust within a certain $\ell_p$-bounded radius, so that it is impossible for the adversary to make adversarial examples in the certificate bound. We present an attack that maintains the impercept... | ["certified defenses", "semantic adversarial examples", "adversarial examples", "spoofed robustness certificates", "adversarial attacks", "certified", "certifiable defenses", "networks", "certain"] | ABSTRACTTo deflect adversarial attacks, a range of “certified” classifiers have been proposed.In addition to labeling an image, certified classifiers produce (when possible) acertificate guaranteeing that the input image is not an `p-bounded adversarial ex-ample. We present a new attack that exploits not only the label... | rJeb2tXkqB | Official Blind Review #3 | 8: Accept | The paper presents a new attack, called the shadow attack, that can maintain the imperceptibility of adversarial samples when out of the certified radius. This work not only aims to target the classifier label but also the certificate by adding large perturbations to the image. The attacks produce a 'spoofed' certifica... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
BREAKING CERTIFIED DEFENSES: SEMANTIC ADVERSARIAL EXAMPLES WITH SPOOFED ROBUSTNESS CERTIFICATES
### Paper Abstract
Defenses against adversarial attacks can be c... | |
wG5XIGi6nrt | ICLR.cc/2021/Conference | 2021 | Learning Private Representations with Focal Entropy | ["Tassilo Klein", "Moin Nabi"] | How can we learn a representation with good predictive power while preserving user privacy?
We present an adversarial representation learning method to sanitize sensitive content from the representation in an adversarial fashion.
Specifically, we propose focal entropy - a variant of entropy embedded in an adversarial r... | ["focal entropy", "private representations", "representation", "adversarial representation", "terms", "good predictive power", "user privacy", "sensitive content", "adversarial fashion", "variant"] | ABSTRACTHow can we learn a representation with good predictive power while preservinguser privacy? We present an adversarial representation learning method to sanitizesensitive content from the representation in an adversarial fashion. Specifically,we propose focal entropy - a variant of entropy embedded in an adversar... | 2fLzVfVuS7h | Review | 5: Marginally below acceptance threshold |
### Summary
This paper presented a method for learning private representations. This method is based on adversarial representation learning and the main technique contribution comes from focal entropy. The experimental results show that focal entropy can improve the accuracy of the target predictor without increasing ... | 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 Private Representations with Focal Entropy
### Paper Abstract
How can we learn a representation with good predictive power while preserving user privacy? We pre... |
rJe7CkrFvS | ICLR.cc/2020/Conference | 2020 | Improving Exploration of Deep Reinforcement Learning using Planning for Policy Search | ["Jakob J. Hollenstein", "Erwan Renaudo", "Justus Piater"] | Most Deep Reinforcement Learning methods perform local search and
therefore are prone to get stuck on non-optimal
solutions. Furthermore, in simulation based training, such as
domain-randomized simulation training, the availability of a simulation
model is not exploited, which potentially decreases
efficiency. To overc... | ["reinforcement learning", "kinodynamic planning", "policy search"] | ABSTRACTMost Deep Reinforcement Learning methods perform local search and thereforeare prone to get stuck on non-optimal solutions. Furthermore, in simulation basedtraining, such as domain-randomized simulation training, the availability of a sim-ulation model is not exploited, which potentially decreases efficiency. T... | ByxPXHFptS | Official Blind Review #2 | 1: Reject | The paper aims to improve exploration in DRL through the use of planning. This is claimed to increase state space coverage in exploration and yield better final policies than methods not augmented with planner derived data.
The current landscape of DRL research is very broad, but RRT can only directly be applied in ce... | <|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 Exploration of Deep Reinforcement Learning using Planning for Policy Search
### Paper Abstract
Most Deep Reinforcement Learning methods perform local search an... | |
rkgqCiRqKQ | ICLR.cc/2019/Conference | 2019 | Inferring Reward Functions from Demonstrators with Unknown Biases | ["Rohin Shah", "Noah Gundotra", "Pieter Abbeel", "Anca Dragan"] | Our goal is to infer reward functions from demonstrations. In order to infer the correct reward function, we must account for the systematic ways in which the demonstrator is suboptimal. Prior work in inverse reinforcement learning can account for specific, known biases, but cannot handle demonstrators with unknown bia... | ["Inverse reinforcement learning", "differentiable planning"] | ABSTRACTOur goal is to infer reward functions from demonstrations. In order to infer thecorrect reward function, we must account for the systematic ways in which thedemonstrator is suboptimal. Prior work in inverse reinforcement learning canaccount for specific, known biases, but cannot handle demonstrators with unknow... | BJxgiuKq2X | Excellent motivation of work but lacks technical merit; results not convincing | 5: Marginally below acceptance threshold | This paper has proposed algorithms for inferring reward functions from demonstrations with unknown biases. To achieve this, the authors have proposed to learn planners from demonstrations in multiple tasks via value iteration networks to learn the reward functions.
This paper has provided an excellent motivation of th... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Inferring Reward Functions from Demonstrators with Unknown Biases
### Paper Abstract
Our goal is to infer reward functions from demonstrations. In order to infer the cor... |
HygkpxStvr | ICLR.cc/2020/Conference | 2020 | Weakly-Supervised Trajectory Segmentation for Learning Reusable Skills | ["Parsa Mahmoudieh", "Trevor Darrell", "Deepak Pathak"] | Learning useful and reusable skill, or sub-task primitives, is a long-standing problem in sensorimotor control. This is challenging because it's hard to define what constitutes a useful skill. Instead of direct manual supervision which is tedious and prone to bias, in this work, our goal is to extract reusable skills f... | ["skills", "demonstration", "agent", "sub-task", "primitives", "robot learning", "manipulation"] | ABSTRACTLearning useful and reusable skill, or sub-task primitives, is a long-standing prob-lem in sensorimotor control. This is challenging because it’s hard to define whatconstitutes a useful skill. Instead of direct manual supervision which is tedious andprone to bias, in this work, our goal is to extract reusable s... | r1lG65CZKB | Official Blind Review #1 | 3: Weak Reject | This paper introduces a weakly supervised learning approach for trajectory segmentation, which relies on coarse labelling about the occurrence of a skill in a demonstration to segment trajectories. This is accomplished through a recurrent model predicting skill categories for each step in a trajectory, and a trajectory... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Weakly-Supervised Trajectory Segmentation for Learning Reusable Skills
### Paper Abstract
Learning useful and reusable skill, or sub-task primitives, is a long-standing ... | |
Hye1RJHKwB | ICLR.cc/2020/Conference | 2020 | Training Generative Adversarial Networks from Incomplete Observations using Factorised Discriminators | ["Daniel Stoller", "Sebastian Ewert", "Simon Dixon"] | Generative adversarial networks (GANs) have shown great success in applications such as image generation and inpainting.
However, they typically require large datasets, which are often not available, especially in the context of prediction tasks such as image segmentation that require labels. Therefore, methods such as... | ["Adversarial Learning", "Semi-supervised Learning", "Image generation", "Image segmentation", "Missing Data"] | ABSTRACTGenerative adversarial networks (GANs) have shown great success in applicationssuch as image generation and inpainting. However, they typically require largedatasets, which are often not available, especially in the context of predictiontasks such as image segmentation that require labels. Therefore, methods su... | SygA0eq1qr | Official Blind Review #3 | 8: Accept | I found this paper very easy and clear to follow - the authors present, what I believe to be an elegant, approach to training a GAN in the presence of missing data or where many marginal samples might be available but very few complete (e.g. paired) samples. The approach proceeds by identifying that the joint distribut... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Training Generative Adversarial Networks from Incomplete Observations using Factorised Discriminators
### Paper Abstract
Generative adversarial networks (GANs) have show... | |
BJeWUs05KQ | ICLR.cc/2019/Conference | 2019 | Directed-Info GAIL: Learning Hierarchical Policies from Unsegmented Demonstrations using Directed Information | ["Mohit Sharma", "Arjun Sharma", "Nicholas Rhinehart", "Kris M. Kitani"] | The use of imitation learning to learn a single policy for a complex task that has multiple modes or hierarchical structure can be challenging. In fact, previous work has shown that when the modes are known, learning separate policies for each mode or sub-task can greatly improve the performance of imitation learning. ... | ["Imitation Learning", "Reinforcement Learning", "Deep Learning"] | ABSTRACTThe use of imitation learning to learn a single policy for a complex task that hasmultiple modes or hierarchical structure can be challenging. In fact, previous workhas shown that when the modes are known, learning separate policies for eachmode or sub-task can greatly improve the performance of imitation learn... | H1l3BLbJpm | The InfoGail method extended to online latent code estimation at test time | 6: Marginally above acceptance threshold | The paper presents a learning-based method for learning the latent context codes from demonstrations along with a GAIL model.
This amounts to learning the option segments and the policies simultaneously.
The main contribution is the model the problem as a time-dependent context and then use a directed information flo... | 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
Directed-Info GAIL: Learning Hierarchical Policies from Unsegmented Demonstrations using Directed Information
### Paper Abstract
The use of imitation learning to learn a... |
B1x_K04YwS | ICLR.cc/2020/Conference | 2020 | Generalizing Natural Language Analysis through Span-relation Representations | ["Zhengbao Jiang", "Wei Xu", "Jun Araki", "Graham Neubig"] | A large number of natural language processing tasks exist to analyze syntax, semantics, and information content of human language. These seemingly very different tasks are usually solved by specially designed architectures. In this paper, we provide the simple insight that a great variety of tasks can be represented in... | ["Span-Relation Representation", "Task-Independent Framework", "Natural Language Analysis", "Benchmark"] | ABSTRACTA large number of natural language processing tasks exist to analyze syntax, se-mantics, and information content of human language. These seemingly very dif-ferent tasks are usually solved by specially designed architectures. In this paper,we provide the simple insight that a great variety of tasks can be repre... | BJgiPLnHFH | Official Blind Review #1 | 3: Weak Reject | This paper generalizes a wide range of natural language processing tasks as a single span-based framework and proposes a general architecture to solve all these problems. Specifically, each task can be modeled as either span identification/prediction or relation prediction based on the spans. The authors adopt an exist... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Generalizing Natural Language Analysis through Span-relation Representations
### Paper Abstract
A large number of natural language processing tasks exist to analyze synt... | |
oW4rUGjbMYg | ACM.org/ICMI/2023/Workshop/GENEA_Challenge | 2023 | The KCL-SAIR team's entry to the GENEA Challenge 2023 Exploring Role-based Gesture Generation in Dyadic Interactions: Listener vs. Speaker | ["Viktor Schmuck", "Nguyen Tan Viet Tuyen", "Oya Celiktutan"] | This paper presents the KCL-SAIR team's contribution to the GENEA Challenge 2023. As this year's challenge addressed gesture generation in a dyadic context instead of a monadic one, our aim was to investigate how the previous state-of-the-art approach can be improved to be more applicable for the generation of both spe... | ["datasets", "Tacotron2", "gesture generation", "dyadic interaction"] | ABSTRACTThis paper presents the KCL-SAIR team’s contribution to the GE-NEA Challenge 2023. As this year’s challenge addressed gesturegeneration in a dyadic context instead of a monadic one, our aimwas to investigate how the previous state-of-the-art approach can beimproved to be more applicable for the generation of bo... | yuUZRCWtP6 | Review of Role-based Gesture Generation in Dyadic Interactions | 6: Marginally above acceptance threshold | Paper Summary:
This paper presents a novel approach to gesture generation in dyadic interactions. The authors propose training two separate models based on the baseline model structure, one for when the agent is speaking (SM) and another for when the agent is listening (LM). The authors propose training two separate m... | 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
The KCL-SAIR team's entry to the GENEA Challenge 2023 Exploring Role-based Gesture Generation in Dyadic Interactions: Listener vs. Speaker
### Paper Abstract
This paper ... |
QkRbdiiEjM | ICLR.cc/2021/Conference | 2021 | AdaGCN: Adaboosting Graph Convolutional Networks into Deep Models | ["Ke Sun", "Zhanxing Zhu", "Zhouchen Lin"] | The design of deep graph models still remains to be investigated and the crucial part is how to explore and exploit the knowledge from different hops of neighbors in an efficient way. In this paper, we propose a novel RNN-like deep graph neural network architecture by incorporating AdaBoost into the computation of netw... | ["Graph Neural Networks", "AdaBoost"] | ABSTRACTThe design of deep graph models still remains to be investigated and the crucialpart is how to explore and exploit the knowledge from different hops of neighborsin an efficient way. In this paper, we propose a novel RNN-like deep graph neu-ral network architecture by incorporating AdaBoost into the computation ... | PmccgXMlmUr | Adaboosting GCNs | 7: Good paper, accept | Summary
In this paper, the authors study graph convolutional networks, where they propose to use AdaBoost for Deep GCNs. This method makes it possible to use information from multi-hop neighbours. Computationally, the proposed method is efficient, which is illustrated through various experiments. The paper is well writ... | 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
AdaGCN: Adaboosting Graph Convolutional Networks into Deep Models
### Paper Abstract
The design of deep graph models still remains to be investigated and the crucial par... |
ryj38zWRb | ICLR.cc/2018/Conference | 2018 | Optimizing the Latent Space of Generative Networks | ["Piotr Bojanowski", "Armand Joulin", "David Lopez-Paz", "Arthur Szlam"] | Generative Adversarial Networks (GANs) have achieved remarkable results in the task of generating realistic natural images. In most applications, GAN models share two aspects in common. On the one hand, GANs training involves solving a challenging saddle point optimization problem, interpreted as an adversarial game be... | ["generative models", "latent variable models", "image generation", "generative adversarial networks", "convolutional neural networks"] | ABSTRACTGenerative Adversarial Networks (GANs) have achieved remarkable results in thetask of generating realistic natural images. In most applications, GAN modelsshare two aspects in common. On the one hand, GANs training involves solvinga challenging saddle point optimization problem, interpreted as an adversarialgam... | BkILtntlz | OPTIMIZING THE LATENT SPACE OF GENERATIVE NETWORKS | 4: Ok but not good enough - rejection | Summary: The authors observe that the success of GANs can be attributed to two factors; leveraging the inductive bias of deep CNNs and the adversarial training protocol. In order to disentangle the factors of success, and they propose to eliminate the adversarial training protocol while maintaining the first factor. Th... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Optimizing the Latent Space of Generative Networks
### Paper Abstract
Generative Adversarial Networks (GANs) have achieved remarkable results in the task of generating r... |
_QdvdkxOii6 | ICLR.cc/2021/Conference | 2021 | Measuring Progress in Deep Reinforcement Learning Sample Efficiency | ["Florian E. Dorner"] | Sampled environment transitions are a critical input to deep reinforcement learning (DRL) algorithms. Current DRL benchmarks often allow for the cheap and easy generation of large amounts of samples such that perceived progress in DRL does not necessarily correspond to improved sample efficiency. As simulating real wor... | ["Deep Reinforcement Learning", "Sample Efficiency"] | ABSTRACTSampled environment transitions are a critical input to deep reinforcement learn-ing (DRL) algorithms. Current DRL benchmarks often allow for the cheap andeasy generation of large amounts of samples such that perceived progress in DRLdoes not necessarily correspond to improved sample efficiency. As simulating r... | KluXSnbjrUK | A meta-analysis of sample efficiency in deep RL | 4: Ok but not good enough - rejection | ##################################
Summary:
This paper conducts a meta-analysis of the trend in sample efficiency in deep RL. The authors argue that this is an informative measure of the progress in the field, in addition to the usual metrics of reward for given tasks, as it is an important consideration when applyin... | 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
Measuring Progress in Deep Reinforcement Learning Sample Efficiency
### Paper Abstract
Sampled environment transitions are a critical input to deep reinforcement learni... |
r1nTpv9eg | ICLR.cc/2017/conference | 2017 | Learning to Perform Physics Experiments via Deep Reinforcement Learning | ["Misha Denil", "Pulkit Agrawal", "Tejas D Kulkarni", "Tom Erez", "Peter Battaglia", "Nando de Freitas"] | When encountering novel objects, humans are able to infer a wide range of physical properties such as mass, friction and deformability by interacting with them in a goal driven way. This process of active interaction is in the same spirit as a scientist performing experiments to discover hidden facts. Recent advances i... | ["Deep learning", "Reinforcement Learning"] | ABSTRACTWhen encountering novel objects, humans are able to infer a wide range of phys-ical properties such as mass, friction and deformability by interacting with themin a goal driven way. This process of active interaction is in the same spirit asa scientist performing experiments to discover hidden facts. Recent adv... | B1YP4Db4l | mixed opinion | 7: Good paper, accept | This paper presents interesting experimental findings that state-of-the-art deep reinforcement learning methods enable agent learning of latent (physical) properties in its environment. The paper formulates the problem of an agent labeling environmental properties after interacting with the environment based on its act... | 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 to Perform Physics Experiments via Deep Reinforcement Learning
### Paper Abstract
When encountering novel objects, humans are able to infer a wide range of phys... |
r98z93GXhRt | ML_Reproducibility_Challenge/2021/Fall | 2021 | FakeFlow: Fake News Detection by Modeling the Flow of Affective Information | ["Anonymous"] | Unlike in short news texts, authors of longer articles stir the readers’ attention by means of emotional appeals that
arouse their feelings. To capture this, Bilal et al. (2021) propose in their paper to model the flow of affective information
in fake news articles using a neural architecture. The authors claim to intr... | ["flow", "affective information", "fakeflow", "fake news detection", "authors", "fake news articles", "affective information fakeflow", "short news texts", "longer articles", "readers"] | FakeFlow: Fake News Detection by Modeling the Flow ofAffective InformationAnonymous Author(s)AffiliationAddressemailReproducibility Summary1Scope of Reproducibility2Unlike in short news texts, authors of longer articles stir the readers’ attention by means of emotional appeals that3arouse their feelings. To capture thi... | rAxgn55tZGc | Review: Fake news detection by flow of information-FakeFlow | 8: Top 50% of accepted papers, clear accept | This work focused on detection of fake information in article of moderate length considering the information flow. The authors created a fake news dataset from several sources, and named MultiSourceFake. The work came up with deep networks involving CNN, bi-directional GRU layers for the task of identification. The wor... | 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
FakeFlow: Fake News Detection by Modeling the Flow of Affective Information
### Paper Abstract
Unlike in short news texts, authors of longer articles stir the readers’ a... |
r1exVhActQ | ICLR.cc/2019/Conference | 2019 | DEEP-TRIM: REVISITING L1 REGULARIZATION FOR CONNECTION PRUNING OF DEEP NETWORK | ["Chih-Kuan Yeh", "Ian E.H. Yen", "Hong-You Chen", "Chun-Pei Yang", "Shou-De Lin", "Pradeep Ravikumar"] | State-of-the-art deep neural networks (DNNs) typically have tens of millions of parameters, which might not fit into the upper levels of the memory hierarchy, thus increasing the inference time and energy consumption significantly, and prohibiting their use on edge devices such as mobile phones. The compression of DNN ... | ["L1 regularization", "deep neural network", "deep compression"] | ABSTRACTState-of-the-art deep neural networks (DNNs) typically have tens of millions ofparameters, which might not fit into the upper levels of the memory hierarchy,thus increasing the inference time and energy consumption significantly, andprohibiting their use on edge devices such as mobile phones. The compressionof ... | Byevdiz3nm | an interesting perspective on the L1 regularization of neural network | 6: Marginally above acceptance threshold | This paper discusses the effect of L1 penalization for deep neural network. In particular it shows the stationary point of an l1 regularized layer has bounded non-zero elements.
The perspective of the proof is interesting: By chain rule, the stationary point satisfies nnz(W^j) linear equations, but the subgradients o... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
DEEP-TRIM: REVISITING L1 REGULARIZATION FOR CONNECTION PRUNING OF DEEP NETWORK
### Paper Abstract
State-of-the-art deep neural networks (DNNs) typically have tens of mil... |
Syt0r4bRZ | ICLR.cc/2018/Conference | 2018 | Tree2Tree Learning with Memory Unit | ["Ning Miao", "Hengliang Wang", "Ran Le", "Chongyang Tao", "Mingyue Shang", "Rui Yan", "Dongyan Zhao"] | Traditional recurrent neural network (RNN) or convolutional neural net- work (CNN) based sequence-to-sequence model can not handle tree structural data well. To alleviate this problem, in this paper, we propose a tree-to-tree model with specially designed encoder unit and decoder unit, which recursively encodes tree in... | ["model", "embeddings", "tree", "rnn", "convolutional neural", "work", "cnn"] | ABSTRACTTraditional recurrent neural network (RNN) or convolutional neural net-work (CNN) based sequence-to-sequence model can not handle tree structuraldata well. To alleviate this problem, in this paper, we propose a tree-to-tree modelwith specially designed encoder unit and decoder unit, which recursively encodestre... | rkAf3gcgM | Please compare with other methods | 4: Ok but not good enough - rejection | This paper presents a model to encode and decode trees in distributed representations.
This is not the first attempt of doing these encoders and decoders. However, there is not a comparative evalution with these methods.
In fact, it has been demonstrated that it is possible to encode and decode trees in distributed st... | 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
Tree2Tree Learning with Memory Unit
### Paper Abstract
Traditional recurrent neural network (RNN) or convolutional neural net- work (CNN) based sequence-to-sequence mode... |
ByetGn0cYX | ICLR.cc/2019/Conference | 2019 | Probabilistic Planning with Sequential Monte Carlo methods | ["Alexandre Piche", "Valentin Thomas", "Cyril Ibrahim", "Yoshua Bengio", "Chris Pal"] | In this work, we propose a novel formulation of planning which views it as a probabilistic inference problem over future optimal trajectories. This enables us to use sampling methods, and thus, tackle planning in continuous domains using a fixed computational budget. We design a new algorithm, Sequential Monte Carlo... | ["control as inference", "probabilistic planning", "sequential monte carlo", "model based reinforcement learning"] | ABSTRACTIn this work, we propose a novel formulation of planning which views it as aprobabilistic inference problem over future optimal trajectories. This enables usto use sampling methods, and thus, tackle planning in continuous domains usinga fixed computational budget. We design a new algorithm, Sequential MonteCarl... | SJgI4p6O2m | Sequential Monte Carlo (SMC) has since its inception some 25 years ago proved to be a powerful and generally applicable tool. The authors of this paper continue this development in a very interesting and natural way by showing how SMC can be used to solve challenging planning problems. This is a enabled by reformulatin... | 8: Top 50% of accepted papers, clear accept | Sequential Monte Carlo (SMC) has since its inception some 25 years ago proved to be a powerful and generally applicable tool. The authors of this paper continue this development in a very interesting and natural way by showing how SMC can be used to solve challenging planning problems. This is a enabled by reformulatin... | 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
Probabilistic Planning with Sequential Monte Carlo methods
### Paper Abstract
In this work, we propose a novel formulation of planning which views it as a probabilistic ... |
eQe8DEWNN2W | ICLR.cc/2021/Conference | 2021 | Calibration of Neural Networks using Splines | ["Kartik Gupta", "Amir Rahimi", "Thalaiyasingam Ajanthan", "Thomas Mensink", "Cristian Sminchisescu", "Richard Hartley"] | Calibrating neural networks is of utmost importance when employing them in safety-critical applications where the downstream decision making depends on the predicted probabilities. Measuring calibration error amounts to comparing two empirical distributions. In this work, we introduce a binning-free calibration measure... | ["neural network calibration", "uncertainty", "calibration measure"] | ABSTRACTCalibrating neural networks is of utmost importance when employing them insafety-critical applications where the downstream decision making depends on thepredicted probabilities. Measuring calibration error amounts to comparing two em-pirical distributions. In this work, we introduce a binning-free calibration ... | ThYCYAtWQqC | Solid work on an important topic | 7: Good paper, accept | The paper presents a post-hoc calibration method for deep neural net classification. The method proposes to first reduces the well-known ECE score to a special case of the Kolmogorov-Smirnov (KS) test, and this way solves the dependency of ECE on the limiting binning assumption. The method proposes next to recalibrate ... | 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
Calibration of Neural Networks using Splines
### Paper Abstract
Calibrating neural networks is of utmost importance when employing them in safety-critical applications w... |
SJlYqRNKDS | ICLR.cc/2020/Conference | 2020 | Blockwise Adaptivity: Faster Training and Better Generalization in Deep Learning | ["Shuai Zheng", "James T. Kwok"] | Stochastic methods with coordinate-wise adaptive stepsize (such as RMSprop and Adam) have been widely used in training deep neural networks. Despite their fast convergence, they can generalize worse than stochastic gradient descent. In this paper, by revisiting the design of Adagrad, we propose to split the network pa... | ["optimization", "deep learning", "blockwise adaptivity"] | ABSTRACTStochastic methods with coordinate-wise adaptive stepsize (such as RMSprop andAdam) have been widely used in training deep neural networks. Despite theirfast convergence, they can generalize worse than stochastic gradient descent. Inthis paper, by revisiting the design of Adagrad, we propose to split the networ... | Hkgm9j2x5S | Official Blind Review #3 | 1: Reject | The paper proposes adaptive gradient approaches where the step-size is not determined on the per-coordinate basis but rather for blocks of coordinates. Theoretical results are presented in terms of regret in online convex optimization, regarding convergence in non-convex optimization, and with respect to uniform stabi... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Blockwise Adaptivity: Faster Training and Better Generalization in Deep Learning
### Paper Abstract
Stochastic methods with coordinate-wise adaptive stepsize (such as R... | |
E9W0QPxtZ_u | ICLR.cc/2021/Conference | 2021 | not-so-big-GAN: Generating High-Fidelity Images on Small Compute with Wavelet-based Super-Resolution | ["Seungwook Han", "Akash Srivastava", "Cole Lincoln Hurwitz", "Prasanna Sattigeri", "David Daniel Cox"] | State-of-the-art models for high-resolution image generation, such as BigGAN and VQVAE-2, require an incredible amount of compute resources and/or time (512 TPU-v3 cores) to train, putting them out of reach for the larger research community. On the other hand, GAN-based image super-resolution models, such as ESRGAN, ca... | ["deep generative modeling", "GAN", "super-resolution", "wavelet transformation", "energy efficient"] | ABSTRACTState-of-the-art models for high-resolution image generation, such as BigGANand VQV AE-2, require an incredible amount of compute resources and/or time(512 TPU-v3 cores) to train, putting them out of reach for the larger research com-munity. On the other hand, GAN-based image super-resolution models, such asESR... | rVG0ALVHmb1 | A cost-effective two-step GAN framework | 5: Marginally below acceptance threshold | **Summary**
This paper proposes a cost-effective two-step training GAN framework (NSB-GAN). NSB-GAN contains a learned sampler in the wavelet domain, and a decoder to super-resolve images from the wavelet domain to the RGB space. Compared with the baseline BigGAN model their method reduces the cost of training and gen... | 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
not-so-big-GAN: Generating High-Fidelity Images on Small Compute with Wavelet-based Super-Resolution
### Paper Abstract
State-of-the-art models for high-resolution image... |
S1eBzhRqK7 | ICLR.cc/2019/Conference | 2019 | Evolutionary-Neural Hybrid Agents for Architecture Search | ["Krzysztof Maziarz", "Andrey Khorlin", "Quentin de Laroussilhe", "Andrea Gesmundo"] | Neural Architecture Search has recently shown potential to automate the design of Neural Networks. The use of Neural Network agents trained with Reinforcement Learning can offer the possibility to learn complex patterns, as well as the ability to explore a vast and compositional search space. On the other hand, evoluti... | ["Evolutionary", "Architecture Search", "NAS"] | ABSTRACTNeural Architecture Search has recently shown potential to automate the design ofNeural Networks. The use of Neural Network agents trained with ReinforcementLearning can offer the possibility to learn complex patterns, as well as the ability toexplore a vast and compositional search space. On the other hand, ev... | Syxf0zQch7 | Interesting method. However, empirical results are not convincing enough. | 4: Ok but not good enough - rejection | Summary:
The paper proposes a hybrid approach which combines evolution and RL. The key idea is to conduct tournament selection over a population of architectures with learned mutations. The mutations are defined as the output of an RNN controller which either reuses or alters the sequence descriptor of the parent at e... | 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
Evolutionary-Neural Hybrid Agents for Architecture Search
### Paper Abstract
Neural Architecture Search has recently shown potential to automate the design of Neural Net... |
r1rqJyHKg | ICLR.cc/2017/workshop | 2017 | Efficient Sparse-Winograd Convolutional Neural Networks | ["Xingyu Liu", "Song Han", "Huizi Mao", "William J. Dally"] | Convolutional Neural Networks (CNNs) are compute intensive which limits their application on mobile devices. Their energy is dominated by the number of multiplies needed to perform the convolutions. Winograd’s minimal filtering algorithm (Lavin (2015)) and network pruning (Han et al. (2015)) reduce the operation count.... | ["Deep learning"] | ABSTRACTConvolutional Neural Networks (CNNs) are compute intensive which limits theirapplication on mobile devices. Their energy is dominated by the number of mul-tiplies needed to perform the convolutions. Winograd’s minimal filtering algo-rithm (Lavin (2015)) and network pruning (Han et al. (2015)) reduce the operati... | BkMwvOejg | Nice trick to reduce the number of multiplications required to evaluate a CNN | 7: Good paper, accept | The paper presents two tricks to reduce the number of multiplications needed to compute convolutions with small kernels. First, the trainable filters are stored directly in the Winograd domain, rather than stored in feature domain and then transformed. This directly impacts the number of multiplications that need to be... | 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
Efficient Sparse-Winograd Convolutional Neural Networks
### Paper Abstract
Convolutional Neural Networks (CNNs) are compute intensive which limits their application on m... |
tuIt1aIb6Co | ICLR.cc/2021/Conference | 2021 | Counterfactual Self-Training | ["Ruijiang Gao", "Max Biggs", "Wei Sun", "Ligong Han"] | Unlike traditional supervised learning, in many settings only partial feedback is available. We may only observe outcomes for the chosen actions, but not the counterfactual outcomes associated with other alternatives. Such settings encompass a wide variety of applications including pricing, online marketing and precis... | ["self-training", "counterfactual inference"] | ABSTRACTUnlike traditional supervised learning, in many settings only partial feedback isavailable. We may only observe outcomes for the chosen actions, but not the coun-terfactual outcomes associated with other alternatives. Such settings encompassa wide variety of applications including pricing, online marketing and ... | 9aKz3JGkI6z | Review | 5: Marginally below acceptance threshold | The paper proposes to use self-training to tackle the fundamental problem of causal inference where only one potential outcome is seen. The proposed self-training method is iterative: after training a model on the observational dataset, they run points with different actions (treatments) through the trained model and c... | 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
Counterfactual Self-Training
### Paper Abstract
Unlike traditional supervised learning, in many settings only partial feedback is available. We may only observe outcomes... |
H1Nyf7W0Z | ICLR.cc/2018/Conference | 2018 | Alpha-divergence bridges maximum likelihood and reinforcement learning in neural sequence generation | ["Sotetsu Koyamada", "Yuta Kikuchi", "Atsunori Kanemura", "Shin-ichi Maeda", "Shin Ishii"] | Neural sequence generation is commonly approached by using maximum- likelihood (ML) estimation or reinforcement learning (RL). However, it is known that they have their own shortcomings; ML presents training/testing discrepancy, whereas RL suffers from sample inefficiency. We point out that it is difficult to resolve a... | ["neural network", "reinforcement learning", "natural language processing", "machine translation", "alpha-divergence"] | ABSTRACTNeural sequence generation is commonly approached by using maximum-likelihood (ML) estimation or reinforcement learning (RL). However, it is knownthat they have their own shortcomings; ML presents training/testing discrepancy,whereas RL suffers from sample inefficiency. We point out that it is difficult toresol... | BkwRmW1Wz | Review of the paper: unfortunately I do not understand main points of this paper and cannot give accurate reviews | 4: Ok but not good enough - rejection | This paper considers a dichitomy between ML and RL based methods for sequence generation. It is argued that the ML approach has some "discrepancy" between the optimization objective and the learning objective, and the RL approach suffers from bad sample complexity. An alpha-divergence formulation is considered to combi... | 1: The reviewer's evaluation is an educated guess | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Alpha-divergence bridges maximum likelihood and reinforcement learning in neural sequence generation
### Paper Abstract
Neural sequence generation is commonly approached... |
HklPzxHFwB | ICLR.cc/2020/Conference | 2020 | Zero-Shot Policy Transfer with Disentangled Attention | ["Josh Roy", "George Konidaris"] | Domain adaptation is an open problem in deep reinforcement learning (RL). Often, agents are asked to perform in environments where data is difficult to obtain. In such settings, agents are trained in similar environments, such as simulators, and are then transferred to the original environment. The gap between visual o... | ["Transfer Learning", "Reinforcement Learning", "Attention", "Domain Adaptation", "Representation Learning", "Feature Extraction"] | ABSTRACTDomain adaptation is an open problem in deep reinforcement learning (RL). Of-ten, agents are asked to perform in environments where data is difficult to ob-tain. In such settings, agents are trained in similar environments, such as sim-ulators, and are then transferred to the original environment. The gap betwe... | HyxtTsDjKH | Official Blind Review #2 | 1: Reject | Pros:
This paper proposed a new method for zero-shot transfer learning under the reinforcement learning setting. The use of attention weights to regularize the latent states was fairly interesting.
Cons:
Limited applicability of the proposed methods
- The paper was restricted in a setting where rewards, actions, and 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
Zero-Shot Policy Transfer with Disentangled Attention
### Paper Abstract
Domain adaptation is an open problem in deep reinforcement learning (RL). Often, agents are aske... | |
BkesJ3R9YX | ICLR.cc/2019/Conference | 2019 | Where and when to look? Spatial-temporal attention for action recognition in videos | ["Lili Meng", "Bo Zhao", "Bo Chang", "Gao Huang", "Frederick Tung", "Leonid Sigal"] | Inspired by the observation that humans are able to process videos efficiently by only paying attention when and where it is needed, we propose a novel spatial-temporal attention mechanism for video-based action recognition. For spatial attention, we learn a saliency mask to allow the model to focus on the most salient... | ["visual attention", "video action recognition", "network interpretability"] | ABSTRACTInspired by the observation that humans are able to process videos efficiently byonly paying attention where andwhen it is needed, we propose a novel spatial-temporal attention mechanism for video-based action recognition. For spatial at-tention, we learn a saliency mask only using convolutional layers to allow... | S1gJRUfs3X | Minor novelty | 3: Clear rejection | A method for activity recognition in videos is presented, which uses spatial soft attention combined with temporal soft attention. In a nutshell, a pixelwise mask is output and elementwise combined with feature maps for spatial attention, and temporal attention is a distribution over frames. The method is tested on sev... | 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
Where and when to look? Spatial-temporal attention for action recognition in videos
### Paper Abstract
Inspired by the observation that humans are able to process videos... |
S1ljpTEFPB | ICLR.cc/2020/Conference | 2020 | Sparsity Learning in Deep Neural Networks | ["Amirsina Torfi", "Rouzbeh A. Shirvani", "Sobhan Soleymani", "Nasser M. Nasrabadi"] | The main goal of network pruning is imposing sparsity on the neural network by increasing the number of parameters with zero value in order to reduce the architecture size and the computational speedup. | ["Neural Networks", "Deep Learning", "Sparsity", "Guided Attention"] | ABSTRACTThe main goal of network pruning is imposing sparsity on the neural networkby increasing the number of parameters with zero value in order to reduce thearchitecture size and the computational speedup. In most of the previous researchworks, sparsity is imposed stochastically without considering any prior knowled... | HylAMrHoFS | Official Blind Review #1 | 1: Reject | # Summary
In this paper, the authors propose an approach to induce (structured) sparsity in deep neural networks during training.
The proposed approach builds heavily on Structured Sparsity Learning (SSL) [1], which in a nutshell pioneered the use of a group-Lasso regularizer [2] to induce sparsity at the level of f... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Sparsity Learning in Deep Neural Networks
### Paper Abstract
The main goal of network pruning is imposing sparsity on the neural network by increasing the number of par... | |
HozL9BGbnKr | AKBC.ws/2021/Conference | 2021 | A Joint Training Framework for Open-World Knowledge Graph Embeddings | ["Karthik V", "Beethika Tripathi", "Mitesh M Khapra", "Balaraman Ravindran"] | Knowledge Graphs(KGs) represent factual information as graphs of entities connected by relations. Knowledge graph embeddings have emerged as a popular approach to encode this information for various downstream tasks like natural language inference, question answering and dialogue generation. As knowledge bases expand, ... | ["Knowledge Graph", "Knowledge Graph completion", "open-world embeddings"] | Automated Knowledge Base Construction (2021) Conference paperA Joint Training Framework for Open-World KnowledgeGraph EmbeddingsKarthik Venkat Ramanan vkarthik2019@gmail.comBeethika Tripathi beethika.tripathi@gmail.comMitesh M. Khapra miteshk@cse.iitm.ac.inBalaraman Ravindran ravi@cse.iitm.ac.inRobert Bosch Centre for ... | nZKKD9oHYMo | Interesting fusion of text and conventional KG Embeddings, consistent improvements over baselines for all datasets/metrics | 6: Marginally above acceptance threshold | The authors propose a new method for computing Knowledge Graph Embeddings on KGs where textual descriptions of entities are available. They evaluate the effectiveness of their method on a number of downstream tasks, including open-world KG completion, closed-world KG completion and entity classification. They show that... | 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 Joint Training Framework for Open-World Knowledge Graph Embeddings
### Paper Abstract
Knowledge Graphs(KGs) represent factual information as graphs of entities connect... |
HJWGdbbCW | ICLR.cc/2018/Conference | 2018 | Reinforcement and Imitation Learning for Diverse Visuomotor Skills | ["Yuke Zhu", "Ziyu Wang", "Josh Merel", "Andrei Rusu", "Tom Erez", "Serkan Cabi", "Saran Tunyasuvunakool", "J\u00e1nos Kram\u00e1r", "Raia Hadsell", "Nando de Freitas", "Nicolas Heess"] | We propose a general deep reinforcement learning method and apply it to robot manipulation tasks. Our approach leverages demonstration data to assist a reinforcement learning agent in learning to solve a wide range of tasks, mainly previously unsolved. We train visuomotor policies end-to-end to learn a direct mapping f... | ["reinforcement learning", "imitation learning", "robotics", "visuomotor skills"] | ABSTRACTWe propose a general deep reinforcement learning method and apply it to robotmanipulation tasks. Our approach leverages demonstration data to assist a re-inforcement learning agent in learning to solve a wide range of tasks, mainlypreviously unsolved. We train visuomotor policies end-to-end to learn a directmap... | HJge1dvgz | Good experimental work, but some pretty severe over-statements and claims | 6: Marginally above acceptance threshold | Paper summary: The authors propose a number of tricks to enable training policies for pick and place style tasks using a combination of GAIL-based imitation learning and hand-specified rewards, as well as use of unobserved state information during training and hand-designed curricula. The results demonstrate manipulati... | 5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Reinforcement and Imitation Learning for Diverse Visuomotor Skills
### Paper Abstract
We propose a general deep reinforcement learning method and apply it to robot manip... |
8C_CfbhFjQG | ICLR.cc/2022/Track/Blog | 2022 | Large Scale Image Completion via Co-Modulated Generative Adversarial Networks | ["Anonymous"] | Co-modulated GANs link image-conditional GANs and unconditional modulated models to address large-scale image completion tasks. Co-modulation brings stochastic and conditional style representations together. To improve existing metrics for image completion, the proposed Paired/Unpaired Inception Discriminative Score (P... | ["GAN", "co-modulation", "image completion", "image-to-image translation"] | BcxQKfs0Rt | A condensed version of the paper, with additional insights to capabilities and limitations | 5: Marginally below acceptance threshold | ## Summary of the contribution
The blog post (BP) addresses the Co-Mod-GAN paper my Zhao et al, ICLR 2021.
The paper highlights 2 contributions: a new information flow to improve GAN-based in-painting tasks for large missing areas, and a new SVM-based metric.
The BP provides an overview of the paper's contribution th... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Large Scale Image Completion via Co-Modulated Generative Adversarial Networks
### Paper Abstract
Co-modulated GANs link image-conditional GANs and unconditional modulate... | |
SJQNqLFgl | ICLR.cc/2017/conference | 2017 | Deep Convolutional Neural Network Design Patterns | ["Leslie N. Smith", "Nicholay Topin"] | Recent research in the deep learning field has produced a plethora of new architectures. At the same time, a growing number of groups are applying deep learning to new applications. Some of these groups are likely to be composed of inexperienced deep learning practitioners who are baffled by the dizzying array of arc... | ["groups", "recent research", "deep learning field", "plethora", "new architectures", "time", "number", "deep learning", "new applications"] | ABSTRACTRecent research in the deep learning field has produced a plethora of new ar-chitectures. At the same time, a growing number of groups are applying deeplearning to new applications. Some of these groups are likely to be composedof inexperienced deep learning practitioners who are baffled by the dizzying ar-ray ... | SJEc2TZVl | 3: Clear rejection | The paper formulates a number of rules for designing convolutional neural network architectures for image processing and computer vision problems. Essentially, it reads like a review paper about modern CNN architectures. It also proposes a few new architectural ideas inspired by these rules. These are experimentally ev... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Deep Convolutional Neural Network Design Patterns
### Paper Abstract
Recent research in the deep learning field has produced a plethora of new architectures. At the sam... | |
HJxR7R4FvS | ICLR.cc/2020/Conference | 2020 | RaCT: Toward Amortized Ranking-Critical Training For Collaborative Filtering | ["Sam Lobel*", "Chunyuan Li*", "Jianfeng Gao", "Lawrence Carin"] | We investigate new methods for training collaborative filtering models based on actor-critic reinforcement learning, to more directly maximize ranking-based objective functions. Specifically, we train a critic network to approximate ranking-based metrics, and then update the actor network to directly optimize against t... | ["Collaborative Filtering", "Recommender Systems", "Actor-Critic", "Learned Metrics"] | ABSTRACTWe investigate new methods for training collaborative filtering models based onactor-critic reinforcement learning, to more directly maximize ranking-based objec-tive functions. Specifically, we train a critic network to approximate ranking-basedmetrics, and then update the actor network to directly optimize ag... | r1edoAD6tS | Official Blind Review #3 | 6: Weak Accept | The work presents a two-level architecture for building an efficient ranking solution for standard collaborative filtering task in recommender systems. The core of this solution is the actor-critic approach, where the critic (motivated by ideas from RL) tries to directly approximate a ranking-based metric using as an i... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
RaCT: Toward Amortized Ranking-Critical Training For Collaborative Filtering
### Paper Abstract
We investigate new methods for training collaborative filtering models b... | |
BkVsWbbAW | ICLR.cc/2018/Conference | 2018 | Deep Generative Dual Memory Network for Continual Learning | ["Nitin Kamra", "Umang Gupta", "Yan Liu"] | Despite advances in deep learning, artificial neural networks do not learn the same way as humans do. Today, neural networks can learn multiple tasks when trained on them jointly, but cannot maintain performance on learnt tasks when tasks are presented one at a time -- this phenomenon called catastrophic forgetting is ... | ["Continual Learning", "Catastrophic Forgetting", "Sequential Multitask Learning", "Deep Generative Models", "Dual Memory Networks", "Deep Learning"] | ABSTRACTDespite advances in deep learning, artificial neural networks do not learn the sameway as humans do. Today, neural networks can learn multiple tasks when trainedon them jointly, but cannot maintain performance on learnt tasks when tasks arepresented one at a time – this phenomenon called catastrophic forgetting... | S1t718Tef | A moderately surprising but useful study on replay based learning | 7: Good paper, accept | This paper introduces a neural network architecture for continual learning. The model is inspired by current knowledge about long term memory consolidation mechanisms in humans. As a consequence, it uses:
- One temporary memory storage (inspired by hippocampus) and a long term memory
- A notion of memory replay, implem... | 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
Deep Generative Dual Memory Network for Continual Learning
### Paper Abstract
Despite advances in deep learning, artificial neural networks do not learn the same way as ... |
hLElJeJKxzY | ICLR.cc/2021/Conference | 2021 | Deep Q Learning from Dynamic Demonstration with Behavioral Cloning | ["Xiaoshuang Li", "Junchen Jin", "Xiao Wang", "Fei-Yue Wang"] | Although Deep Reinforcement Learning (DRL) has proven its capability to learn optimal policies by directly interacting with simulation environments, how to combine DRL with supervised learning and leverage additional knowledge to assist the DRL agent effectively still remains difficult. This study proposes a novel ... | ["deep q", "dynamic demonstration", "behavioral", "drl", "drl model", "model", "deep reinforcement learning", "capability", "optimal policies", "simulation environments"] | ABSTRACTAlthough Deep Reinforcement Learning (DRL) has proven its capability to learnoptimal policies by directly interacting with simulation environments, how tocombine DRL with supervised learning and leverage additional knowledge to as-sist the DRL agent effectively still remains difficult. This study proposes a nov... | tZdEeok0gxC | Tend to accept | 6: Marginally above acceptance threshold | Summary:
This paper proposes integrating deep Q-learning from dynamic demonstrations with a behavioral cloning model (DQfDD-BC). Compared with DQfD, the proposed approach introduces a behavior cloning model, which was first pre-trained by leveraging historical demonstrations and then updated using generated dynamic 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
Deep Q Learning from Dynamic Demonstration with Behavioral Cloning
### Paper Abstract
Although Deep Reinforcement Learning (DRL) has proven its capability to learn o... |
-kigPjfTIGd | ICLR.cc/2021/Conference | 2021 | SSW-GAN: Scalable Stage-wise Training of Video GANs | ["Lluis Castrejon", "Nicolas Ballas", "Aaron Courville"] | Current state-of-the-art generative models for videos have high computational requirements that impede high resolution generations beyond a few frames. In this work we propose a stage-wise strategy to train Generative Adversarial Networks (GANs) for videos. We decompose the generative process to first produce a downsa... | ["video generation", "GANs", "scalable methods"] | ABSTRACTCurrent state-of-the-art generative models for videos have high computationalrequirements that impede high resolution generations beyond a few frames. In thiswork we propose a stage-wise strategy to train Generative Adversarial Networks(GANs) for videos. We decompose the generative process to first produce adow... | S_P6fy7E5qA | This paper proposes a stage-wise strategy to train Generative Adversarial Networks for videos. The contribution of this paper is very limited and the experiments are not convincing. | 3: Clear rejection | Pros:
1. A stage-wise approach to train GANs for video is defined to reduce the computational costs needed to generate long high resolution videos.
2. The authors provide some quality results of the proposed approach.
Cons:
1. The contribution of this paper is very limited. The authors just do some incremental improve... | 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
SSW-GAN: Scalable Stage-wise Training of Video GANs
### Paper Abstract
Current state-of-the-art generative models for videos have high computational requirements that im... |
Sklsm20ctX | ICLR.cc/2019/Conference | 2019 | Competitive experience replay | ["Hao Liu", "Alexander Trott", "Richard Socher", "Caiming Xiong"] | Deep learning has achieved remarkable successes in solving challenging reinforcement learning (RL) problems when dense reward function is provided. However, in sparse reward environment it still often suffers from the need to carefully shape reward function to guide policy optimization. This limits the applicability of... | ["reinforcement learning", "sparse reward", "goal-based learning"] | ABSTRACTDeep learning has achieved remarkable successes in solving challenging reinforcementlearning (RL) problems when dense reward function is provided. However, in sparsereward environment it still often suffers from the need to carefully shape reward functionto guide policy optimization. This limits the applicabili... | BygCiBx9nm | To address the sparse reward problems, the authors propose a relabeling strategy called Competitive Experience Reply (CER). This strategy relabels states, and places learning in the context of an exploration competition between a pair of agents. The experiments support some parts of authors’ claim well. However, the... | 6: Marginally above acceptance threshold | The authors propose a states relabeling strategy (CER) to encourage exploration in RL algorithms by organizing a competitive game between a pair of agents.
To verify their strategy, they extend MADDPG as their framework. Then, they compare the performance of agents trained with HER, and both variants of CER, and both ... | 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
Competitive experience replay
### Paper Abstract
Deep learning has achieved remarkable successes in solving challenging reinforcement learning (RL) problems when dense r... |
H8mMlaG73CY | ML_Reproducibility_Challenge/2021/Fall | 2021 | [RE] Bad Seeds: Evaluating Lexical Methods for Bias Measurement | ["Anonymous"] |
In this work we verify the results of Bad Seeds: Evaluating Lexical Methods for Bias Measurement.
We replicate the experiments conducted and verify the main claims made in the original paper: (1) Bias measurements depend on seeds and models. (2) Shuffled seed pairs can result in a significant different bias subspace ... | ["Seeds", "NLP", "Bias", "Embedding", "PCA", "WEAT"] | [RE] Bad Seeds: Evaluating Lexical Methods for BiasMeasurementAnonymous Author(s)AffiliationAddressemailReproducibility Summary 1Scope of Reproducibility 2In this work we verify the results of Bad Seeds: Evaluating Lexical Methods for Bias Measurement (Antoniak and 3Mimno, 2021). We replicate the experiments conducted ... | rhNgWzkNqf5 | Review of "[RE] Bad Seeds: Evaluating Lexical Methods for Bias Measurement" | 4: Ok but not good enough - rejection | **Reproducibility Summary:**
This report tries to reproduce the results presented in the paper titled Bad Seeds: Evaluating Lexical Methods for Bias Measurement (Antoniak and Mimno, 2021). It has a proper first page for the summary of their work. Scope of reproducibility part in this page, the authors pick to use "veri... | 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
[RE] Bad Seeds: Evaluating Lexical Methods for Bias Measurement
### Paper Abstract
In this work we verify the results of Bad Seeds: Evaluating Lexical Methods for Bias ... |
VdWaMgaTKtX | JSYS/2021/Mar_Papers | 2021 | SoK: Function-as-a-Service: From An Application Developer's Perspective | ["Ali Raza", "Ibrahim Matta", "Nabeel Akhtar", "Vasiliki Kalavri", "Vatche Isahagian"] | In the past few years, FaaS has gained significant popularity and became a go-to choice for deploying cloud applications and micro-services. FaaS with its unique 'pay as you go' pricing model and key performance benefits over other cloud services, offers an easy and intuitive programming model to build cloud applicatio... | ["serverless", "faas", "lambda", "gcf", "cloud"] | Journal of Systems Research (JSys) 2021Function -as-a-Service : From an Application Developer ’sPerspectiveAli RazaBoston Universityaraza@bu.eduIbrahim MattaBoston Universitymatta@bu.eduNabeel AkhtarAkamai Technologies Inc.nakhtar@akamai.comVasiliki KalavriBoston Universityvkalavri@bu.eduVatche IsahagianIBM Researchvat... | bFjZY7icVxy | Review for "Serverless Computing: From An Application Developer’s Perspective" | The paper presents a SOK on the subject of Serverless Computing, with the focus on the information that can be useful for application developers (as opposed to infrastructure managers). I must confess that I was unexcited about reading yet another Serverless SOK paper, but I found the perspective refreshing and enjoyed... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
SoK: Function-as-a-Service: From An Application Developer's Perspective
### Paper Abstract
In the past few years, FaaS has gained significant popularity and became a go-... | ||
HygpthEtvr | ICLR.cc/2020/Conference | 2020 | ProxSGD: Training Structured Neural Networks under Regularization and Constraints | ["Yang Yang", "Yaxiong Yuan", "Avraam Chatzimichailidis", "Ruud JG van Sloun", "Lei Lei", "Symeon Chatzinotas"] | In this paper, we consider the problem of training neural networks (NN). To promote a NN with specific structures, we explicitly take into consideration the nonsmooth regularization (such as L1-norm) and constraints (such as interval constraint). This is formulated as a constrained nonsmooth nonconvex optimization prob... | ["stochastic gradient descent", "regularization", "constrained optimization", "nonsmooth optimization"] | ABSTRACTIn this paper, we consider the problem of training structured neural networks (NN)with nonsmooth regularization (e.g. `1-norm) and constraints (e.g. interval con-straints). We formulate training as a constrained nonsmooth nonconvex optimiza-tion problem, and propose a convergent proximal-type stochastic gradien... | HJxQie4LcS | Official Blind Review #1 | 6: Weak Accept | The paper proposes a new gradient-based stochastic optimization algorithm (with gradient averaging) by adapting theory for proximal algorithms (originally developed for convex problems) to the non-convex setting. The main idea is to first use an averaged gradient plus a quadratic term to locally approximate the non-con... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
ProxSGD: Training Structured Neural Networks under Regularization and Constraints
### Paper Abstract
In this paper, we consider the problem of training neural networks (... | |
SJe8DsR9tm | ICLR.cc/2019/Conference | 2019 | Dynamic Early Terminating of Multiply Accumulate Operations for Saving Computation Cost in Convolutional Neural Networks | ["Yu-Yi Su", "Yung-Chih Chen", "Xiang-Xiu Wu", "Shih-Chieh Chang"] | Deep learning has been attracting enormous attention from academia as well as industry due to its great success in many artificial intelligence applications. As more applications are developed, the need for implementing a complex neural network model on an energy-limited edge device becomes more critical. To this end, ... | ["Convolutional neural network", "Early terminating", "Dynamic model optimization"] | ABSTRACTDeep learning has been attracting enormous attention from academia as well as in-dustry due to its great success in many artificial intelligence applications. As moreapplications are developed, the need for implementing a complex neural networkmodel on an energy-limited edge device becomes more critical. To thi... | ryxtgdYQnQ | Simple and effective method which might still have room to improve | 6: Marginally above acceptance threshold | In convolutional neural networks, a convolutional filter conducts a series of multiply-accumulate (MAC) operations, which is computationally heavy. To save computational cost, this manuscript proposes an algorithm to set a checkpoint in the MAC process to determine whether a filter could terminate early based on the in... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Dynamic Early Terminating of Multiply Accumulate Operations for Saving Computation Cost in Convolutional Neural Networks
### Paper Abstract
Deep learning has been attrac... |
HygtiTEYvS | ICLR.cc/2020/Conference | 2020 | Self-Supervised Policy Adaptation | ["Christopher Mutschler", "Sebastian Pokutta"] | We consider the problem of adapting an existing policy when the environment representation changes. Upon a change of the encoding of the observations the agent can no longer make use of its policy as it cannot correctly interpret the new observations. This paper proposes Greedy State Representation Learning (GSRL) to t... | ["reinforcement learning", "environment representation", "representation learning", "model mismatch"] | ABSTRACTWe consider the problem of adapting an existing policy when the environmentrepresentation changes. Upon a change of the encoding of the observations theagent can no longer make use of its policy as it cannot correctly interpret thenew observations. This paper proposes Greedy State Representation Learning(GSRL) ... | BylWnJiXtH | Official Blind Review #3 | 1: Reject | The paper investigates a setting in which the observation function changes while the underlying environmental dynamic stays the same.
In order to re-use the policy which was trained on the old observation function, they propose to learn a mapping function to map the new observations to the old ones.
I believe the wor... | <|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-Supervised Policy Adaptation
### Paper Abstract
We consider the problem of adapting an existing policy when the environment representation changes. Upon a change of... | |
lukVnNC-Hc | EMNLP/2020/Workshop/NLP-COVID | 2020 | COVID-19: A Semantic-Based Pipeline for Recommending Biomedical Entities | ["Marcia Afonso Barros", "Andre Lamurias", "Diana Sousa", "Pedro Ruas", "Francisco M. Couto"] | With the increasing number of publications about COVID-19, it is a challenge to extract personalized knowledge suitable for each researcher. This work aims to build a new semantic-based pipeline for recommending biomedical entities to scientific researchers. To this end, we developed a pipeline that creates an implicit... | ["COVID-19", "NER", "Relation Extraction", "Recommender System", "Ontology"] | COVID-19: A Semantic-Based Pipeline for Recommending BiomedicalEntitiesMarcia Barros1,2, Andre Lamurias1, Diana Sousa1, Pedro Ruas1andFrancisco M. Couto11LASIGE, Faculdade de Ci ˆencias, Universidade de Lisboa2CENTRA, Faculdade de Ci ˆencias, Universidade de Lisboamcbarros@fc.ul.pt, alamurias@lasige.di.fc.ul.pt,dfsousa... | 3uf1IeDYoKc | Need to improve the experiment design | 4: Ok but not good enough - rejection | This paper proposes a system for biomedical entity recommendation. The system performs entity recommendation in a pipeline manner with three steps: named entity recognition, relation extraction, and entity recognition. The paper shows that the recommendation accuracy can be enhanced if multiple ontologies are used for ... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
COVID-19: A Semantic-Based Pipeline for Recommending Biomedical Entities
### Paper Abstract
With the increasing number of publications about COVID-19, it is a challenge ... |
V6BjBgku7Ro | ICLR.cc/2021/Conference | 2021 | Planning from Pixels using Inverse Dynamics Models | ["Keiran Paster", "Sheila A. McIlraith", "Jimmy Ba"] | Learning dynamics models in high-dimensional observation spaces can be challenging for model-based RL agents. We propose a novel way to learn models in a latent space by learning to predict sequences of future actions conditioned on task completion. These models track task-relevant environment dynamics over a distribut... | ["model based reinforcement learning", "deep reinforcement learning", "multi-task learning", "deep learning", "goal-conditioned reinforcement learning"] | ABSTRACTLearning task-agnostic dynamics models in high-dimensional observation spacescan be challenging for model-based RL agents. We propose a novel way to learnlatent world models by learning to predict sequences of future actions conditionedon task completion. These task-conditioned models adaptively focus modelingc... | SaAICBJEUS4 | Review for Planning from Pixels using Inverse Dynamics Models | 6: Marginally above acceptance threshold | Summary: The author proposes Goal-Conditioned Latent Action Models for RL (GLAMOR) a novel approach to learn latent world models by modeling inverse dynamics. The proposed approach learns to track task-relevant dynamics for a diverse distribution of tasks and provide a strong heuristic that enables efficient planning. ... | 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
Planning from Pixels using Inverse Dynamics Models
### Paper Abstract
Learning dynamics models in high-dimensional observation spaces can be challenging for model-based ... |
nyhZXiaotm | MIDL.io/2020/Conference | 2020 | Robust Image Segmentation Quality Assessment | ["Leixin Zhou", "Wenxiang Deng", "Xiaodong Wu"] | Deep learning based image segmentation methods have achieved great success, even having human-level accuracy in some applications. However, due to the black box nature of deep learning, the best method may fail in some situations. Thus predicting segmentation quality without ground truth would be very crucial especiall... | ["Robust", "Segmentation Quality", "Deep Learning"] | Medical Imaging with Deep Learning 2020 1{5 MIDL 2020 { Short PaperRobust Image Segmentation Quality AssessmentLeixin Zhou leixin-zhou@uiowa.eduWenxiang Deng wenxiang-deng@uiowa.eduXiaodong Wu xiaodong-wu@uiowa.eduDepartment of Electrical and Computer Engineering, University of Iowa, USAAbstractDeep learning based imag... | POeg0b8EPs | Interesting idea and evaluation against adversarial examples | 3: Weak accept | The authors proposed a method for segmentation quality assessment. The method consists of learning a reconstruction network for the masked input images and a regression network for quality assessment. The reconstruction network aims to only faithfully reconstruct input images masked correctly by the segmentation, while... | 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 Image Segmentation Quality Assessment
### Paper Abstract
Deep learning based image segmentation methods have achieved great success, even having human-level accu... |
Hyg-JC4FDr | ICLR.cc/2020/Conference | 2020 | Imitation Learning via Off-Policy Distribution Matching | ["Ilya Kostrikov", "Ofir Nachum", "Jonathan Tompson"] | When performing imitation learning from expert demonstrations, distribution matching is a popular approach, in which one alternates between estimating distribution ratios and then using these ratios as rewards in a standard reinforcement learning (RL) algorithm. Traditionally, estimation of the distribution ratio requi... | ["reinforcement learning", "deep learning", "imitation learning", "adversarial learning"] | ABSTRACTWhen performing imitation learning from expert demonstrations, distributionmatching is a popular approach, in which one alternates between estimating dis-tribution ratios and then using these ratios as rewards in a standard reinforcementlearning (RL) algorithm. Traditionally, estimation of the distribution rati... | Hkg8cdP6YB | Official Blind Review #3 | 6: Weak Accept | This paper provides a novel off policy objective to solve imitation learning. It resolves the limitation of the famous GAIL algorithm that we need on-policy samples to interact with the environment. The new algorithm is simple but efficient, and can handle off-policy settings. The derivation of equation (12) is nice 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
Imitation Learning via Off-Policy Distribution Matching
### Paper Abstract
When performing imitation learning from expert demonstrations, distribution matching is a popu... | |
aaI4jKANEH4 | ICML.cc/2020/Workshop/SAS | 2020 | Revisiting Representation Learning for Singing Voice Separation with Sinkhorn Distances | ["Stylianos Ioannis Mimilakis", "Konstantinos Drossos", "Gerald Schuller"] | In this work we present a method for unsupervised learning of audio representations, focused on the task of singing voice separation. We
build upon a previously proposed method for learning representations of time-domain music signals with a re-parameterized denoising autoencoder, extending it by using the family of Si... | ["Representation learning", "unsupervised learning", "music source separation"] | Revisiting Representation Learning for Singing Voice Separation withSinkhorn DistancesStylianos I. Mimilakis* 1Konstantinos Drossos* 2Gerald Schuller3AbstractIn this work we present a method for unsupervisedlearning of audio representations, focused on thetask of singing voice separation. We build upona previously prop... | VcDsxBgN5z | Unsupervised approach for singing voice separation with Sinkhorn distance as regularizer | 6: Marginally above acceptance threshold | The paper builds on prior work by the authors in recent conferences, notably in Eusipco 2020 where they proposed the core algorithm which the paper extends. The proposed approach builds upon the denoising auto encoder model and uses sinusoids to reconstruct the original signal in the decoder. In prior work the autho... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Revisiting Representation Learning for Singing Voice Separation with Sinkhorn Distances
### Paper Abstract
In this work we present a method for unsupervised learning of ... |
HHSEKOnPvaO | ICLR.cc/2021/Conference | 2021 | Graph-Based Continual Learning | ["Binh Tang", "David S. Matteson"] | Despite significant advances, continual learning models still suffer from catastrophic forgetting when exposed to incrementally available data from non-stationary distributions. Rehearsal approaches alleviate the problem by maintaining and replaying a small episodic memory of previous samples, often implemented as an a... | ["continual learning", "array", "continual", "significant advances", "continual learning models", "catastrophic", "available data", "distributions", "rehearsal", "problem"] | ABSTRACTDespite significant advances, continual learning models still suffer from catas-trophic forgetting when exposed to incrementally available data from non-stationarydistributions. Rehearsal approaches alleviate the problem by maintaining and re-playing a small episodic memory of previous samples, often implemente... | 2p_Y6d8W0qn | an interesting integration of continual learning, random graph, and attention network | 7: Good paper, accept | This paper presented a memory-based continual learning model where relationships between training samples are represented with a random graph that is defined from the non-linear embedding of the input data. Catastrophic forgetting between tasks is partially (1) alleviated with a graph regularization that penalizes chan... | 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
Graph-Based Continual Learning
### Paper Abstract
Despite significant advances, continual learning models still suffer from catastrophic forgetting when exposed to incre... |
ryQu7f-RZ | ICLR.cc/2018/Conference | 2018 | On the Convergence of Adam and Beyond | ["Sashank J. Reddi", "Satyen Kale", "Sanjiv Kumar"] | Several recently proposed stochastic optimization methods that have been successfully used in training deep networks such as RMSProp, Adam, Adadelta, Nadam are based on using gradient updates scaled by square roots of exponential moving averages of squared past gradients. In many applications, e.g. learning with large... | ["optimization", "deep learning", "adam", "rmsprop"] | ABSTRACTSeveral recently proposed stochastic optimization methods that have been suc-cessfully used in training deep networks such as RMS PROP , ADAM , ADADELTA ,NADAM are based on using gradient updates scaled by square roots of exponentialmoving averages of squared past gradients. In many applications, e.g. learningw... | Hyl2iJgGG | Theoretical (maybe practical) issues with ADAM, and an way to solve them | 8: Top 50% of accepted papers, clear accept | This paper examines the very popular and useful ADAM optimization algorithm, and locates a mistake in its proof of convergence (for convex problems). Not only that, the authors also show a specific toy convex problem on which ADAM fails to converge. Once the problem was identified to be the decrease in v_t (and increas... | 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 Convergence of Adam and Beyond
### Paper Abstract
Several recently proposed stochastic optimization methods that have been successfully used in training deep net... |
B1lC62EKwr | ICLR.cc/2020/Conference | 2020 | Evidence-Aware Entropy Decomposition For Active Deep Learning | ["Weishi Shi", "Xujiang Zhao", "Feng Chen", "Qi Yu"] | We present a novel multi-source uncertainty prediction approach that enables deep learning (DL) models to be actively trained with much less labeled data. By leveraging the second-order uncertainty representation provided by subjective logic (SL), we conduct evidence-based theoretical analysis and formally decompose th... | ["active learning", "entropy decomposition", "uncertainty"] | ABSTRACTWe present a novel multi-source uncertainty prediction approach that enables deeplearning (DL) models to be actively trained with much less labeled data. By lever-aging the second-order uncertainty representation provided by subjective logic(SL), we conduct evidence-based theoretical analysis and formally decom... | SylVvHLkcH | Official Blind Review #1 | 6: Weak Accept | This paper propose an active deep learning model. By leveraging subjective Logic, they propose to decompose the entropy of a predicted class distribution into vacuity (lack of evidence) and dissonance (conflict of strong evidence). Instead of using the predicted class distribution, they estimate the supporting evidence... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Evidence-Aware Entropy Decomposition For Active Deep Learning
### Paper Abstract
We present a novel multi-source uncertainty prediction approach that enables deep learn... | |
rJfHoM-C- | ICLR.cc/2018/Conference | 2018 | Few-Shot Learning with Simplex | ["Bowen Zhang", "Xifan Zhang", "Fan Cheng", "Deli Zhao"] | Deep learning has made remarkable achievement in many fields. However, learning
the parameters of neural networks usually demands a large amount of labeled
data. The algorithms of deep learning, therefore, encounter difficulties when applied
to supervised learning where only little data are available. This specific tas... | ["One-shot learning", "few-shot learning", "deep learning", "simplex"] | ABSTRACTDeep learning has made remarkable achievement in many fields. However, learn-ing the parameters of neural networks usually demands a large amount of labeleddata. The algorithms of deep learning, therefore, encounter difficulties when ap-plied to supervised learning where only little data are available. This spe... | ryht5AYlM | Interesting approach but missing baselines and state of the art claims wrong | 4: Ok but not good enough - rejection | This paper proposes a geometric based approach to solving the problem of one-shot and few-shot learning. The basic idea is to use the feature vectors of a particular class to construct a simplex. (I am assuming the dimensions of the vectors are selected so as to exactly construct a simplex? It is not clearly written in... | 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
Few-Shot Learning with Simplex
### Paper Abstract
Deep learning has made remarkable achievement in many fields. However, learning the parameters of neural networks usual... |
Syzn9i05Ym | ICLR.cc/2019/Conference | 2019 | Learning Neural Random Fields with Inclusive Auxiliary Generators | ["Yunfu Song", "Zhijian Ou"] | Neural random fields (NRFs), which are defined by using neural networks to implement potential functions in undirected models, provide an interesting family of model spaces for machine learning. In this paper we develop a new approach to learning NRFs with inclusive-divergence minimized auxiliary generator - the inclus... | ["Neural random fields", "Deep generative models", "Unsupervised learning", "Semi-supervised learning"] | ABSTRACTNeural random fields (NRFs), which are defined by using neural networks toimplement potential functions in undirected models, provide an interesting familyof model spaces for machine learning. In this paper we develop a new approachto learning NRFs with inclusive-divergence minimized auxiliary generator - thein... | Hkxe-51c3Q | Interesting but incremental | 6: Marginally above acceptance threshold | This paper addresses an important problem of learning the random field using neural networks by using a inclusive auxiliary generator. Comparing to existing state-of-the-art methods for learning neural random fields, this paper used a the inclusive-divergence (KL divergence of the density approximate and the auxiliary ... | 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 Neural Random Fields with Inclusive Auxiliary Generators
### Paper Abstract
Neural random fields (NRFs), which are defined by using neural networks to implement... |
ryAe2WBee | ICLR.cc/2017/conference | 2017 | Multi-label learning with semantic embeddings | ["Liping Jing", "MiaoMiao Cheng", "Liu Yang", "Alex Gittens", "Michael W. Mahoney"] | Multi-label learning aims to automatically assign to an instance (e.g., an image or a document) the most relevant subset of labels from a large set of possible labels. The main challenge is to maintain accurate predictions while scaling efficiently on data sets with extremely large label sets and many training data poi... | ["Supervised Learning"] | ABSTRACTMulti-label learning aims to automatically assign to an instance (e.g., an image ora document) the most relevant subset of labels from a large set of possible labels.The main challenge is to maintain accurate predictions while scaling efficiently ondata sets with extremely large label sets and many training dat... | SyqLAIg4x | review | 4: Ok but not good enough - rejection | The paper presents the semantic embedding model for multi-label prediction.
In my questions, I pointed that the proposed approach assumes the number of labels to predict is known, and the authors said this was an orthogonal question, although I don't think it is!
I was trying to understand how different is SEM from a b... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Multi-label learning with semantic embeddings
### Paper Abstract
Multi-label learning aims to automatically assign to an instance (e.g., an image or a document) the most... |
AUa_CiMnZ9 | graphicsinterface.org/Graphics_Interface/2023/Conference_SD | 2023 | FaceZip: Automatic Texture Compression for Facial Blendshapes | ["Yiyi Cai", "Shinichi Kinuwaki", "Nobuyuki Umetani"] | Recently, numerous cinematic and interactive entertainment production companies have adopted advanced capture systems to acquire faithful facial geometries and corresponding textures. However, it is difficult to animate these models in a controllable manner for real-time applications. Although blendshape is typically u... | ["digital human", "facial animation", "blendshape", "texture compression"] | ABSTRACTRecently, numerous cinematic and interactive entertainment produc-tion companies have adopted advanced capture systems to acquirefaithful facial geometries and corresponding textures. However, itis difficult to animate these models in a controllable manner forreal-time applications. Although blendshape is typic... | ebdBlWFmlP | need more limitation/discussion | 7: Good paper, accept | The paper discusses the importance for realistic digital human models in various applications and the use of photogrammetry to capture 3D geometries and materials of actors. However, skilled artists are still needed to bring life to the captured model due to difficulties in parameterizing the model's surface material a... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
FaceZip: Automatic Texture Compression for Facial Blendshapes
### Paper Abstract
Recently, numerous cinematic and interactive entertainment production companies have ado... |
SJe8DsR9tm | ICLR.cc/2019/Conference | 2019 | Dynamic Early Terminating of Multiply Accumulate Operations for Saving Computation Cost in Convolutional Neural Networks | ["Yu-Yi Su", "Yung-Chih Chen", "Xiang-Xiu Wu", "Shih-Chieh Chang"] | Deep learning has been attracting enormous attention from academia as well as industry due to its great success in many artificial intelligence applications. As more applications are developed, the need for implementing a complex neural network model on an energy-limited edge device becomes more critical. To this end, ... | ["Convolutional neural network", "Early terminating", "Dynamic model optimization"] | ABSTRACTDeep learning has been attracting enormous attention from academia as well as in-dustry due to its great success in many artificial intelligence applications. As moreapplications are developed, the need for implementing a complex neural networkmodel on an energy-limited edge device becomes more critical. To thi... | rylgW2I53m | Unclear if it will save wall clock time. | 6: Marginally above acceptance threshold | This paper motivates itself by observing that not all convolutional weights are required to make an accurate prediction. In the introduction the authors envision a system similar to a cascaded classifier [Viola and Jones 2001] (I draw this conclusion not the paper). However the wording of the introduction is not clear ... | 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
Dynamic Early Terminating of Multiply Accumulate Operations for Saving Computation Cost in Convolutional Neural Networks
### Paper Abstract
Deep learning has been attrac... |
50N86aYZ3n | eswc-conferences.org/ESWC/2021/Conference/Resources_Track | 2021 | P2P-O: A Purchase-To-Pay Ontology for Enabling Semantic Invoices | ["Michael Schulze", "Markus Schr\u00f6der", "Christian Jilek", "Torsten Albers", "Heiko Maus", "Andreas Dengel"] | Small and medium-sized enterprises increasingly adopt electronic invoices and digitized purchase-to-pay processes. A purchase-to-pay process begins with making a purchase order and ends with completing the payment process. Even when organizations adopt electronic invoices, knowledge work in such processes is characteri... | ["Semantic invoice", "Purchase-To-Pay", "Enterprise knowledge graph", "Corporate memory", "RML"] | P2P-O: A Purchase-To-Pay Ontology forEnabling Semantic InvoicesMichael Schulze1;2;3, Markus Schr oder1;2, Christian Jilek1;2, Torsten Albers3,Heiko Maus2, and Andreas Dengel1;21Computer Science Department, Technische Universit at Kaiserslautern, Germany2Smart Data & Knowledge Services Department, Deutsches Forschungsze... | rtAPhJaYd7z | Semantic invoices ontology. Worth to be presented during the conference. | 1: Weak Accept |
Thank you very much for the response. I've have read it and decided not to change my previous review.
----
The resource presented in this represent a modular ontology for representing invoices semantically. The paper is well written and easy to follow. A brief description of each module is provided and the result ha... | 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
P2P-O: A Purchase-To-Pay Ontology for Enabling Semantic Invoices
### Paper Abstract
Small and medium-sized enterprises increasingly adopt electronic invoices and digitiz... |
a4-fFL7aCi0 | AKBC.ws/2021/Conference | 2021 | Commonsense Reasoning with Implicit Knowledge in Natural Language | ["Pratyay Banerjee", "Swaroop Mishra", "Kuntal Kumar Pal", "Arindam Mitra", "Chitta Baral"] | Commonsense Reasoning is a research challenge studied from the early days of AI. In recent years, several natural language QA task have been proposed where commonsense reasoning is important. Two common approaches to this are (i) Use of well-structured commonsense present in knowledge graphs, and (ii) Use of prog... | ["Commonsense Reasoning", "Question Answering"] | Automated Knowledge Base Construction (2021) Conference paperCommonsense Reasoning with Implicit Knowledge inNatural LanguagePratyay Banerjeeypbanerj6@asu.eduSwaroop Mishraysrmishr1@asu.eduKuntal Kumar Palykkpal@asu.eduArindam Mitrazarindam.mitra@microsoft.comChitta Baralychitta@asu.eduySchool of Computing and AI, Ariz... | 15bF0ZCu_DW | A solid work with detailed empirical analysis. | 7: Good paper, accept | The paper presents a detailed study on the commonsense reasoning ability of pre-trained Transformer-based LMs such as BERT and RoBERTa. The authors build an experimental framework for evaluating different models with different knowledge sources under different infusion strategies, e.g., continually pre-training (revisi... | 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
Commonsense Reasoning with Implicit Knowledge in Natural Language
### Paper Abstract
Commonsense Reasoning is a research challenge studied from the early days of AI. ... |
HJlHzJBFwB | ICLR.cc/2020/Conference | 2020 | Accelerating Monte Carlo Bayesian Inference via Approximating Predictive Uncertainty over the Simplex | ["Yufei Cui", "Wuguannan Yao", "Qiao Li", "Antoni Chan", "Chun Jason Xue"] | Estimating the predictive uncertainty of a Bayesian learning model is critical in various decision-making problems, e.g., reinforcement learning, detecting adversarial attack, self-driving car. As the model posterior is almost always intractable, most efforts were made on finding an accurate approximation the true post... | ["predictive uncertainty", "model posterior", "simplex", "bayesian learning model", "critical", "various", "problems", "reinforcement learning"] | ABSTRACTEstimating the predictive uncertainty of a Bayesian learning model is critical in var-ious decision-making problems, e.g., reinforcement learning, detecting adversarialattack, self-driving car. As the model posterior is almost always intractable, mostefforts were made on finding an accurate approximation the tr... | Skg2gUHcFr | Official Blind Review #4 | 3: Weak Reject | Overall I liked several results presented in this paper. The findings in Figure 2 gives clear illustration on how Bayesian classification models distinguish between in-distribution difficult-to-classify data and out-of-distribution data, namely uncertain predicted mean and large predicted variance. Though I believe thi... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Accelerating Monte Carlo Bayesian Inference via Approximating Predictive Uncertainty over the Simplex
### Paper Abstract
Estimating the predictive uncertainty of a Bayes... | |
LVotkZmYyDi | ICLR.cc/2021/Conference | 2021 | Proximal Gradient Descent-Ascent: Variable Convergence under KŁ Geometry | ["Ziyi Chen", "Yi Zhou", "Tengyu Xu", "Yingbin Liang"] | The gradient descent-ascent (GDA) algorithm has been widely applied to solve minimax optimization problems. In order to achieve convergent policy parameters for minimax optimization, it is important that GDA generates convergent variable sequences rather than convergent sequences of function value or gradient norm. How... | ["Kurdyka-\u0141ojasiewicz geometry", "minimax", "nonconvex", "proximal gradient descent-ascent", "variable convergence"] | ABSTRACTThe gradient descent-ascent (GDA) algorithm has been widely applied to solveminimax optimization problems. In order to achieve convergent policy parametersfor minimax optimization, it is important that GDA generates convergent variablesequences rather than convergent sequences of function values or gradient nor... | IB3g5HFmEuE | The results are nice and well-explained; the paper could be improved by adding experiments and better explanation of its relevance to ML | 7: Good paper, accept | ## Summary
This paper discusses the convergence of the proximal gradient algorithm in the case of KL geometry. It does not propose a new algorithm but rather focuses on understanding the simplest (although not the best) algorithm that can be applied to the saddle point objective.
I enjoyed reading this paper, I find t... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Proximal Gradient Descent-Ascent: Variable Convergence under KŁ Geometry
### Paper Abstract
The gradient descent-ascent (GDA) algorithm has been widely applied to solve ... |
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... | HJx3a8J0FB | Official Blind Review #3 | 1: Reject | The paper conjectures that the so called Jeffreys prior over the parameters of a neural network is the prior leading to the best generalization performance. The authors test this conjecture on an artificial task using a small neural network, and investigate the sensitivity of the results to noise.
I like the general i... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
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... | |
rkx6MJSFPH | ICLR.cc/2020/Conference | 2020 | Unrestricted Adversarial Attacks For Semantic Segmentation | ["Guangyu Shen", "Chengzhi Mao", "Junfeng Yang", "Baishakhi Ray"] | Despite the rapid development of adversarial attacks for machine learning models, many types of new adversarial examples still remain unknown. Uncovered types of adversarial attacks pose serious concern for the safety of the models, which raise the question about the effectiveness of current adversarial robustness ev... | ["Adversarial Attacks", "Semantic Segmentation", "Computer Vision"] | ABSTRACTDespite the rapid development of adversarial attacks for machine learning models,many types of new adversarial examples still remain unknown. Uncovered typesof adversarial attacks pose serious concern for the safety of the models, whichraise the question about the effectiveness of current adversarial robustness... | ByeCQCuCKH | Official Blind Review #1 | 1: Reject | This paper studies the problem of unrestricted adversarial attack in the image domain. Building upon the idea of previous work (Song et al. 2018), this paper leverages the state-of-the-art image-to-image translation model SPADE (Park et al. 2019) to synthesize high-resolution adversarial images from semantic segmentati... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Unrestricted Adversarial Attacks For Semantic Segmentation
### Paper Abstract
Despite the rapid development of adversarial attacks for machine learning models, many type... | |
S6AtYQLzXOY | ICLR.cc/2021/Conference | 2021 | Learning Self-Similarity in Space and Time as a Generalized Motion for Action Recognition | ["Heeseung Kwon", "Manjin Kim", "Suha Kwak", "Minsu Cho"] | Spatio-temporal convolution often fails to learn motion dynamics in videos and thus an effective motion representation is required for video understanding in the wild. In this paper, we propose a rich and robust motion representation method based on spatio-temporal self-similarity (STSS). Given a sequence of frames, ST... | ["Action recognition", "Video understanding", "Motion analysis"] | ABSTRACTSpatio-temporal convolution often fails to learn motion dynamics in videos andthus an effective motion representation is required for video understanding in thewild. In this paper, we propose a rich and robust motion representation methodbased on spatio-temporal self-similarity (STSS). Given a sequence of frame... | WmuKjlP7SKL | This is an interesting work utilized STSS to help learn representation end-to-end and achieved good empirical results on several action recognition benchmarks. | 6: Marginally above acceptance threshold | ### Summary:
This submission proposed a motion representation method based on spatio-temporal self-similarity (STSS), which represents each local region as similarities to its neighbors in both spatial and temporal dimension. There are previous works (e.g., Ref[1] , [2], [5] listed here) which utilize STSS for featur... | 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 Self-Similarity in Space and Time as a Generalized Motion for Action Recognition
### Paper Abstract
Spatio-temporal convolution often fails to learn motion dyna... |
nDe2D8DDXKR | AKBC.ws/2021/Conference | 2021 | A Case Study in Bootstrapping Ontology Graphs from Textbooks | ["Vinay K. Chaudhri", "Matthew Boggess", "Han Lin Aung", "Debshila Basu Mallick", "Andrew C Waters", "Richard Baraniuk"] | Ontology graphs are graphs in which the nodes are generic classes and edges have labels that specify the relationships between the classes. In this paper, we address the question:to what extent can automated extraction and crowdsourcing techniques be combined to boostrap the creation of comprehensive and accurat... | ["Ontology", "Education", "Lanugage Models", "BERT"] | Automated Knowledge Base Construction (2021) Conference paperA Case Study in Bootstrapping Ontology Graphs fromTextbooksVinay K. Chaudhri vinayc@stanford.eduMatthew Boggess mattbogg@stanford.eduHan L. Aung hanlaung@cs.stanford.eduDepartment of Computer Science, Stanford University, Stanford, CA, 94305, USADebshila B. M... | vbta3BEO7yL | An interesting experiment on an interesting research topic | 6: Marginally above acceptance threshold | Overall, I like this work. It is well written and easy to follow. It feels like there is still a fair amount of work to get to well-rounded conclusions, but this seems like a good stage to already make initial results available for the community. Here and there, the text could be more clear by only discussing what has ... | 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 Case Study in Bootstrapping Ontology Graphs from Textbooks
### Paper Abstract
Ontology graphs are graphs in which the nodes are generic classes and edges have labels t... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.