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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
RgeMS1Tf1zs | JSYS/2022/Aug_Papers | 2022 | Solution: Inspecting Traffic in Residential Networks with Opportunistically Outsourced Middleboxes | ["Anonymous"] | Since they lack the powerful tools and personnel available in enterprise-grade security solution, home networks have particularly difficult network security challenges. While prior efforts outsource network traffic to cloud or cloudlet services, such measures redirect network traffic out of the home network, which gran... | ["residential networks", "security", "middleboxes"] | SOLUTION : INSPECTING TRAFFIC IN RESIDENTIAL NETWORKS WITHOPPORTUNISTICALLY OUTSOURCED MIDDLEBOXESAnonymous authorsPaper under double-blind reviewAbstractSince they lack the powerful tools and personnel availablein enterprise-grade security solution, home networks haveparticularly difficult network security challenges.... | WHR0nUBnA5F | Review for paper on Inspecting Traffic in Residential Networks with Opportunistically Outsourced Middleboxes | ## Paper summary
The paper evaluates the use of smartphone devices as a proxy to capture and analyse traffic between the client and the Internet. The smartphone is connected as a proxy to the router through NAT, traffic is redirected from the router to the smartphone, before being delivered to the server. It then compa... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Solution: Inspecting Traffic in Residential Networks with Opportunistically Outsourced Middleboxes
### Paper Abstract
Since they lack the powerful tools and personnel av... | ||
3v-mxwmpw2P | IEEE.org/2022/Workshop/altVIS | 2022 | I Learn to Diffuse, or Data Alchemy 101: a Mnemonic Manifesto | ["Anonymous"] | In this manifesto, we put forward the idea of data alchemy as a narrative device to discuss storytelling and transdisciplinarity in visualization. We use text-to-image generative art mainly based on Disco Diffusion (DD) to develop the concept, and structure our argument in ten propositions, as if they were ten issues o... | ["Visualization", "Transdisciplinarity", "AI Art", "Extended Creativity", "Data Alchemy"] | ABSTRACTIn this manifesto, we put forward the idea of data alchemy as anarrative device to discuss storytelling and transdisciplinarity in vi-sualization. If data is the prima materia of modern science, how doesone perform the Great Work? We use text-to-image diffusion-basedgenerative art to develop the concept, and st... | 5YUwKrb1DB | This is beautiful work—a well-executed paper with a strong idea. I agree with the concept of generating images as a form of visualization of the latent space. | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
I Learn to Diffuse, or Data Alchemy 101: a Mnemonic Manifesto
### Paper Abstract
In this manifesto, we put forward the idea of data alchemy as a narrative device to disc... | |||
D04TGKz5rfF | ICLR.cc/2021/Conference | 2021 | A frequency domain analysis of gradient-based adversarial examples | ["Bochen Lv", "Pu Yang", "Zehao Wang", "Zhanxing Zhu"] | It is well known that deep neural networks are vulnerable to adversarial examples. We attempt to understand adversarial examples from the perspective of frequency analysis. Several works have empirically shown that the gradient-based adversarial attacks perform differently in the low-frequency and high-frequency part o... | ["adversarial examples", "part", "frequency domain analysis", "model", "adversarial", "deep neural networks", "vulnerable", "perspective", "frequency analysis", "several works"] | ABSTRACTIt is well known that deep neural networks are vulnerable to adversarial examples.We attempt to understand adversarial examples from the perspective of frequencyanalysis . Several works have empirically shown that the gradient-based adver-sarial attacks perform differently in the low-frequency and high-frequenc... | HxBeYiYABTh | Interesting results but the paper needs improvement | 4: Ok but not good enough - rejection | This paper analyzes the frequency spectrum of adversarial perturbations during normal training. The authors show that the low frequency component (LFC) of adversarial perturbation is increasing during training, but it is not increasing fast enough, so the LFC of adversarial perturbation is not as dense as the input nat... | 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 frequency domain analysis of gradient-based adversarial examples
### Paper Abstract
It is well known that deep neural networks are vulnerable to adversarial examples. ... |
rkxWpCNKvS | ICLR.cc/2020/Conference | 2020 | Improved Image Augmentation for Convolutional Neural Networks by Copyout and CopyPairing | ["Philip May"] | Image augmentation is a widely used technique to improve the performance of convolutional neural networks (CNNs). In common image shifting, cropping, flipping, shearing and rotating are used for augmentation. But there are more advanced techniques like Cutout and SamplePairing.
In this work we present two improvements... | ["image augmentation", "cnn", "images", "augmentation"] | ABSTRACTImage augmentation is a widely used technique to improve the performance ofconvolutional neural networks (CNNs). In common image shifting, cropping,flipping, shearing and rotating are used for augmentation. But there are moreadvanced techniques like Cutout andSamplePairing .In this work we present two improveme... | SylmDA_-FS | Official Blind Review #3 | 1: Reject | I think this paper is not enough to accept in ICLR because
- Lack of novelty.
- CutMix [1] is very similar to Copyout.
- To verify the novelty, a more sophisticated description and experimental supports should be required.
- Insufficient experiments for supporting the effectiveness of the proposed method.
- 6-lay... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Improved Image Augmentation for Convolutional Neural Networks by Copyout and CopyPairing
### Paper Abstract
Image augmentation is a widely used technique to improve the ... | |
rJA5Pz7lHKb | ICLR.cc/2021/Conference | 2021 | Improved Autoregressive Modeling with Distribution Smoothing | ["Chenlin Meng", "Jiaming Song", "Yang Song", "Shengjia Zhao", "Stefano Ermon"] | While autoregressive models excel at image compression, their sample quality is often lacking. Although not realistic, generated images often have high likelihood according to the model, resembling the case of adversarial examples. Inspired by a successful adversarial defense method, we incorporate randomized smoothing... | ["generative models", "autoregressive models"] | ABSTRACTWhile autoregressive models excel at image compression, their sample quality isoften lacking. Although not realistic, generated images often have high likelihoodaccording to the model, resembling the case of adversarial examples. Inspired by asuccessful adversarial defense method, we incorporate randomized smoo... | vWh4UgeXY6u | Interesting empirical results but might benefit from more careful discussion. | 7: Good paper, accept | #### SHORT DESCRIPTION
This paper proposes a two-stage generative modeling approach, first learning a distribution over noised data, then learning the original data distribution conditioned on this noised data. The paper demonstrates that this leads to improved sample quality compared to fitting the data distribution ... | 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
Improved Autoregressive Modeling with Distribution Smoothing
### Paper Abstract
While autoregressive models excel at image compression, their sample quality is often lac... |
rJeeKTNKDB | ICLR.cc/2020/Conference | 2020 | Hierarchical Graph-to-Graph Translation for Molecules | ["Wengong Jin", "Regina Barzilay", "Tommi Jaakkola"] | The problem of accelerating drug discovery relies heavily on automatic tools to optimize precursor molecules to afford them with better biochemical properties. Our work in this paper substantially extends prior state-of-the-art on graph-to-graph translation methods for molecular optimization. In particular, we realize ... | ["graph generation", "deep learning"] | ABSTRACTThe problem of accelerating drug discovery relies heavily on automatic tools tooptimize precursor molecules to afford them with better biochemical properties.Our work in this paper substantially extends prior state-of-the-art on graph-to-graphtranslation methods for molecular optimization. In particular, we rea... | B1lPC22v5H | Official Blind Review #3 | 3: Weak Reject | The authors present a heirarchical graph-to-graph translation method for generating novel organic molecules.
Working from the model of Jin et al. (2019), the authors introduce a three step heirarchy - the model first determines where a substructure should be generated, what is the substructure, then the attachments 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
Hierarchical Graph-to-Graph Translation for Molecules
### Paper Abstract
The problem of accelerating drug discovery relies heavily on automatic tools to optimize precurs... | |
BkxHEW5cF4 | icaps-conference.org/ICAPS/2019/Workshop/SPARK | 2019 | Advantages and Challenges of Using AI Planning in Cloud Migration | ["Hongtan Sun", "Maja Vukovic", "John Rofrano", "Chen Lin"] | Cloud Migration transforms customer’s data, application and services from original IT platform to one or more cloud en- vironment, with the goal of improving the performance of the IT system while reducing the IT management cost. The enterprise level Cloud Migration projects are generally com- plex, involves dynamicall... | ["AI planning", "Cloud Migration"] | Advantages and Challenges of Using AI Planning in Cloud MigrationHongtan Sun, Maja Vukovic, John Rofrano, Chen LinIBM T.J. Watson Research Center1101 Kitchawan Rd,Yorktown Heights, New York 10598hongtan.sun@ibm.com, fmaja, rofranog@us.ibm.com, liana.lin@ibm.comAbstractCloud Migration transforms customer’s data, applica... | rkliadzUjE | potentially interesting topic; submission is light on details | 3: Marginally above acceptance threshold | This short submission (2 pages) seeks to motivate the use of AI planning technology to support the migration of computing resources to a Cloud environment. It outlines the problem at a high level, briefly describes an initial prototype built on Metric-FF for a simple version of the problem, and summarizes open challeng... | 2: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Advantages and Challenges of Using AI Planning in Cloud Migration
### Paper Abstract
Cloud Migration transforms customer’s data, application and services from original I... |
BygXY34FDr | ICLR.cc/2020/Conference | 2020 | VUSFA:Variational Universal Successor Features Approximator | ["Shamane Siriwardhana", "Rivindu Weerasakera", "Denys J.C. Matthies", "Suranga Nanayakkara"] | In this paper, we show how novel transfer reinforcement learning techniques can be applied to the complex task of target-driven navigation using the photorealisticAI2THOR simulator. Specifically, we build on the concept of Universal SuccessorFeatures with an A3C agent. We introduce the novel architectural1contribution ... | ["Universal Successor Features", "Successor Features", "Model Free Deep Reinforcement Learning"] | ABSTRACTIn this paper, we show how novel transfer reinforcement learning techniques canbe applied to the complex task of target driven navigation using the photorealisticAI2THOR simulator. Specifically, we build on the concept of Universal SuccessorFeatures with an A3C agent. We introduce the novel architectural1contri... | SygtgL1FYS | Official Blind Review #2 | 1: Reject | Comments and Questions :
- The core contribution of the paper is to introduce a new architecture for a policy that is useful for successor features. This new architecture uses a variational information bottleneck approach and can be implemented readily on top of existing successor feature based approaches like USFA.... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
VUSFA:Variational Universal Successor Features Approximator
### Paper Abstract
In this paper, we show how novel transfer reinforcement learning techniques can be applie... | |
S1gV6AVKwB | ICLR.cc/2020/Conference | 2020 | Cross Domain Imitation Learning | ["Kun Ho Kim", "Yihong Gu", "Jiaming Song", "Shengjia Zhao", "Stefano Ermon"] | We study the question of how to imitate tasks across domains with discrepancies such as embodiment and viewpoint mismatch. Many prior works require paired, aligned demonstrations and an additional RL procedure for the task. However, paired, aligned demonstrations are seldom obtainable and RL procedures are expensive. I... | ["Imitation Learning", "Domain Adaptation", "Reinforcement Learning", "Zeroshot Learning", "Machine Learning", "Artificial Intelligence"] | ABSTRACTWe study the question of how to imitate tasks across domains with discrepancies suchas embodiment and viewpoint mismatch. Many prior works require paired, aligneddemonstrations and an additional RL procedure for the task. However, paired, aligneddemonstrations are seldom obtainable and RL procedures are expensi... | H1g_go22tH | Official Blind Review #2 | 3: Weak Reject | Summary
This paper proposes a method to perform alignment between demonstrations in different domains in order to adapt a policy from expert domain to the test domain. They also include a formalism that studies alignments of different MDPs.
Strengths
1) The paper presents a formal framework to study many recent method... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Cross Domain Imitation Learning
### Paper Abstract
We study the question of how to imitate tasks across domains with discrepancies such as embodiment and viewpoint misma... | |
RFRpPPiV6wa | eswc-conferences.org/ESWC/2021/Conference/Poster_and_Demo_Track | 2021 | Coverage-Based Summaries for RDF KBs | ["Giannis Vassiliou", "Georgia Troullinou", "Nikos Papadakis", "Kostas Stefanidis", "Evangelia Pitoura", "Haridimos Kondylakis"] | As more and more data become available as linked data, the need for efficient and effective methods for their exploration becomes apparent. Semantic summaries try to extract meaning from data, while reducing its size. State of the art structural semantic summaries, focus primarily on the graph structure of the data, tr... | ["RDF Summaries", "Coverage", "Summaries", "RDF"] | Coverage -Based Summaries for RDF KB s Giannis Vassiliou1, Georgia Troullinou2, Nikos Papadakis1, Kostas Stefanidis3, Evangelia Pitoura4, Haridimos Kondylakis1,2 1 Department of Electrical and Computer Engineering , Hellenic Mediterranean University, He-raklion, Greece 2 FORTH -ICS, Heraklion, Greece 3Tampere Universit... | Q2FFgZP-oTo | A well written paper proposing an algorithm which outperforms the state of the art. A survey was performed previously and next steps are outlined. A promising contribution worthy to be presented and to receive feedback. | 8: Top 50% of accepted papers, clear accept | This poster paper proposes an algorithm to maximize query coverage by using result diversification, thus not only taking center nodes but also perimeter nodes into account.
## Summary
The problem is well defined, a relevant survey from the authors is linked which covers related work and other relevant papers are refe... | 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
Coverage-Based Summaries for RDF KBs
### Paper Abstract
As more and more data become available as linked data, the need for efficient and effective methods for their exp... |
rJgjGxrFPS | ICLR.cc/2020/Conference | 2020 | A Simple and Scalable Shape Representation for 3D Reconstruction | ["Mateusz Michalkiewicz", "Eugene Belilovsky", "Mahsa Baktashmotagh", "Anders Eriksson"] | Deep learning applied to the reconstruction of 3D shapes has seen growing interest. A popular approach to 3D reconstruction and generation in recent years has been the CNN decoder-encoder model often applied in voxel space. However this often scales very poorly with the resolution limiting the effectiveness of these mo... | ["Computer Vision", "3D Reconstruction"] | ABSTRACTDeep learning applied to the reconstruction of 3D shapes has seen growing in-terest. A popular approach to 3D reconstruction and generation in recent yearshas been the CNN decoder-encoder model often applied in voxel space. Howeverthis often scales very poorly with the resolution limiting the effectiveness of t... | rJe5U-qh9r | Official Blind Review #2 | 3: Weak Reject | This paper studies the problem of learning the feature representation for predicting the 3D shape of objects, from a single image or a point cloud. The proposed approach performs PCA on the SDF field. And then the transformed feature map is learned and used as input to task-specific decoders for 3D shape prediction. Th... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
A Simple and Scalable Shape Representation for 3D Reconstruction
### Paper Abstract
Deep learning applied to the reconstruction of 3D shapes has seen growing interest. A... | |
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... | jGckYg0zj_1 | A resource using the NeOn Methodology to descrive an represent machine-readable electronic invoices | 1: Weak Accept | Summary
========
This resource paper proposes an ontology following the NeOn methodology to describe and represent electronic invoices in a machine-readable format. The authors use mainly the European Standard EN 16931-1 for their proposal. The ontology is published on a stable PURL with access to the modules of the on... | 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
P2P-O: A Purchase-To-Pay Ontology for Enabling Semantic Invoices
### Paper Abstract
Small and medium-sized enterprises increasingly adopt electronic invoices and digitiz... |
B1gqipNYwH | ICLR.cc/2020/Conference | 2020 | Option Discovery using Deep Skill Chaining | ["Akhil Bagaria", "George Konidaris"] | Autonomously discovering temporally extended actions, or skills, is a longstanding goal of hierarchical reinforcement learning. We propose a new algorithm that combines skill chaining with deep neural networks to autonomously discover skills in high-dimensional, continuous domains. The resulting algorithm, deep skill c... | ["Hierarchical Reinforcement Learning", "Reinforcement Learning", "Skill Discovery", "Deep Learning", "Deep Reinforcement Learning"] | ABSTRACTAutonomously discovering temporally extended actions, or skills , is a longstand-ing goal of hierarchical reinforcement learning. We propose a new algorithmthat combines skill chaining with deep neural networks to autonomously discoverskills in high-dimensional, continuous domains. The resulting algorithm, deep... | Bkgp_NaFYB | Official Blind Review #1 | 6: Weak Accept | This paper studies the problem of learning suitable action abstractions (i.e., options or skills) that can be composed hierarchically to solve control tasks. The starting point for the paper is the (classic) observation that one skill should end where another can start. The paper then proposes a recursive algorithm for... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Option Discovery using Deep Skill Chaining
### Paper Abstract
Autonomously discovering temporally extended actions, or skills, is a longstanding goal of hierarchical rei... | |
Utc4Yd1RD_s | ICLR.cc/2021/Conference | 2021 | Towards Defending Multiple Adversarial Perturbations via Gated Batch Normalization | ["Aishan Liu", "Shiyu Tang", "Xianglong Liu", "Xinyun Chen", "Lei Huang", "Zhuozhuo Tu", "Dawn Song", "Dacheng Tao"] | There is now extensive evidence demonstrating that deep neural networks are vulnerable to adversarial examples, motivating the development of defenses against adversarial attacks. However, existing adversarial defenses typically improve model robustness against individual specific perturbation types. Some recent method... | ["adversarial examples", "multiple adversarial peturbation types", "adversarial robustness"] | ABSTRACTThere is now extensive evidence demonstrating that deep neural networks are vul-nerable to adversarial examples, motivating the development of defenses againstadversarial attacks. However, existing adversarial defenses typically improvemodel robustness against individual specific perturbation types. Some recent... | tT3Vj-Ps4Tt | Method is interesting. The results show the effectiveness. But the reviewer worries whether the evaluation is valid. | 5: Marginally below acceptance threshold | Existing defenses are usually robust against certain types of attacks. This paper proposes to build a defense against different types of norm-bounded attacks via gated batch normalization. This work treats different types of adversarial attacks as data from different domains and proposes to use gated batch normalizatio... | 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 Defending Multiple Adversarial Perturbations via Gated Batch Normalization
### Paper Abstract
There is now extensive evidence demonstrating that deep neural netw... |
5k8F6UU39V | ICLR.cc/2021/Conference | 2021 | Autoregressive Entity Retrieval | ["Nicola De Cao", "Gautier Izacard", "Sebastian Riedel", "Fabio Petroni"] | Entities are at the center of how we represent and aggregate knowledge. For instance, Encyclopedias such as Wikipedia are structured by entities (e.g., one per Wikipedia article). The ability to retrieve such entities given a query is fundamental for knowledge-intensive tasks such as entity linking and open-domain ques... | ["entity retrieval", "document retrieval", "autoregressive language model", "entity linking", "end-to-end entity linking", "entity disambiguation", "constrained beam search"] | ABSTRACTEntities are at the center of how we represent and aggregate knowledge. For in-stance, Encyclopedias such as Wikipedia are structured by entities (e.g., one perWikipedia article). The ability to retrieve such entities given a query is fundamen-tal for knowledge-intensive tasks such as entity linking and open-do... | gmngTl8-zQe | The paper proposes a brand new approach for entity retrieval, which leverages an encoder-decoder architecture to generate the target entity directly. Intensive experiments on entity disambiguation, entity linking, and document retrieval tasks prove the effectiveness of the approach. | 7: Good paper, accept | The paper proposes a brand new approach for entity retrieval, which leverages an encoder-decoder architecture to generate the target entity directly. To tackle the problem of invalid generation, a trie-constrained beam search is used for decoding. The author performs intensive experiments on entity disambiguation, enti... | 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
Autoregressive Entity Retrieval
### Paper Abstract
Entities are at the center of how we represent and aggregate knowledge. For instance, Encyclopedias such as Wikipedia ... |
r1gEqiC9FX | ICLR.cc/2019/Conference | 2019 | Equi-normalization of Neural Networks | ["Pierre Stock", "Benjamin Graham", "R\u00e9mi Gribonval", "Herv\u00e9 J\u00e9gou"] | Modern neural networks are over-parametrized. In particular, each rectified linear hidden unit can be modified by a multiplicative factor by adjusting input and out- put weights, without changing the rest of the network. Inspired by the Sinkhorn-Knopp algorithm, we introduce a fast iterative method for minimizing the l... | ["convolutional neural networks", "Normalization", "Sinkhorn", "Regularization"] | ABSTRACTModern neural networks are over-parametrized. In particular, each rectified linearhidden unit can be modified by a multiplicative factor by adjusting input and out-put weights, without changing the rest of the network. Inspired by the Sinkhorn-Knopp algorithm, we introduce a fast iterative method for minimizing... | SkeZ8Jeq2Q | A bit too incremental and needs comparison with Weight Normalization | 7: Good paper, accept | Summary:
This paper introduces equi-normalization (ENorm): a normalization technique that relies on the scaling invariance properties of the ReLU, similarly to Path-SGD. Their method explicitly use this property to balance the weights of the network, without changing the function computed by the network. The main diffe... | 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
Equi-normalization of Neural Networks
### Paper Abstract
Modern neural networks are over-parametrized. In particular, each rectified linear hidden unit can be modified b... |
DEa4JdMWRHp | ICLR.cc/2021/Conference | 2021 | Interpretable Models for Granger Causality Using Self-explaining Neural Networks | ["Ri\u010dards Marcinkevi\u010ds", "Julia E Vogt"] | Exploratory analysis of time series data can yield a better understanding of complex dynamical systems. Granger causality is a practical framework for analysing interactions in sequential data, applied in a wide range of domains. In this paper, we propose a novel framework for inferring multivariate Granger causality u... | ["time series", "Granger causality", "interpretability", "inference", "neural networks"] | ABSTRACTExploratory analysis of time series data can yield a better understanding of com-plex dynamical systems. Granger causality is a practical framework for analysinginteractions in sequential data, applied in a wide range of domains. In this paper,we propose a novel framework for inferring multivariate Granger caus... | GL25BZf1vqH | Paper is well presented but does not offer a fundamentally unique point of view as it combines existing ideas. | 4: Ok but not good enough - rejection | This paper primarily deals with learning Granger-causal relationships in multivariate time series in the nonlinear dynamics setting. The core method uses vector autoregressive modeling with sparsity inducing regularizers (elastic net and smoothness based fused lasso) along with the recently proposed with self-explainin... | 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
Interpretable Models for Granger Causality Using Self-explaining Neural Networks
### Paper Abstract
Exploratory analysis of time series data can yield a better understan... |
BJlAzTEKwS | ICLR.cc/2020/Conference | 2020 | Attraction-Repulsion Actor-Critic for Continuous Control Reinforcement Learning | ["Thang Doan", "Bogdan Mazoure", "Audrey Durand", "Joelle Pineau", "R Devon Hjelm"] | Continuous control tasks in reinforcement learning are important because they provide an important framework for learning in high-dimensional state spaces with deceptive rewards, where the agent can easily become trapped into suboptimal solutions.
One way to avoid local optima is to use a population of agents to ensure... | ["reinforcement learning", "continuous control", "multi-agent", "mujoco"] | ABSTRACTIn reinforcement learning, robotic control tasks are often useful for understandinghow agents perform in environments with deceptive rewards where the agent caneasily become trapped into suboptimal solutions. One way to avoid these localoptima is to use a population of agents to ensure coverage of the policy sp... | BJx3HfFoYB | Official Blind Review #3 | 3: Weak Reject | The paper proposes an ensemble method for reinforcement learning in which the policy updates are modulated with a loss which encourages diversity among all experienced policies. It is a combination of SAC, normalizing flow policies, and an approach to diversity considered by Hong et al. (2018). The work seems rather in... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Attraction-Repulsion Actor-Critic for Continuous Control Reinforcement Learning
### Paper Abstract
Continuous control tasks in reinforcement learning are important becau... | |
SyNbRj09Y7 | ICLR.cc/2019/Conference | 2019 | Visual Imitation Learning with Recurrent Siamese Networks | ["Glen Berseth", "Christopher J. Pal"] | People are incredibly skilled at imitating others by simply observing them. They achieve this even in the presence of significant morphological differences and capabilities. Further, people are able to do this from raw perceptions of the actions of others, without direct access to the abstracted demonstration actions a... | ["Reinforcement Learning", "Imitation Learning", "Deep Learning"] | ABSTRACTPeople are incredibly skilled at imitating others by simply observing them. Theyachieve this even in the presence of significant morphological differences and ca-pabilities. Further, people are able to do this from raw perceptions of the actions ofothers, without direct access to the abstracted demonstration ac... | r1erAUxeaQ | Issues with significance of results | 4: Ok but not good enough - rejection | This paper proposes an imitation learning method solely from video demonstrations by learning recurrent image-based distance model and in conjunction using RL to track that distance.
Clarity: The paper writing is mostly clear. The motivation for using videos as a demonstration source could be more clearly stated. One ... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Visual Imitation Learning with Recurrent Siamese Networks
### Paper Abstract
People are incredibly skilled at imitating others by simply observing them. They achieve thi... |
ryG8UsR5t7 | ICLR.cc/2019/Conference | 2019 | MERCI: A NEW METRIC TO EVALUATE THE CORRELATION BETWEEN PREDICTIVE UNCERTAINTY AND TRUE ERROR | ["michel moukari", "lo\u00efc simon", "sylvaine picard", "fr\u00e9d\u00e9ric jurie"] | As deep learning applications are becoming more and more pervasive, the question of evaluating the reliability of a prediction becomes a central question in the machine learning community. This domain, known as predictive uncertainty, has come under the scrutiny of research groups developing Bayesian approaches to deep... | ["evaluation metric", "predictive uncertainty", "deep learning"] | ABSTRACTAs deep learning applications are becoming more and more pervasive, the ques-tion of evaluating the reliability of a prediction becomes a central question inthe machine learning community. This domain, known as predictive uncertainty,has come under the scrutiny of research groups developing Bayesian approachest... | BkgP0S052m | Ok, but not good enough | 4: Ok but not good enough - rejection | The authors propose mean rescaled confidence interval (MERCI) as a way to measure the quality of predictive uncertainty for regression problems. The main idea is to rescale confidence intervals, and use average width of the confidence intervals as a measure for calibration. Due to the rescaling, the MERCI score is inse... | 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
MERCI: A NEW METRIC TO EVALUATE THE CORRELATION BETWEEN PREDICTIVE UNCERTAINTY AND TRUE ERROR
### Paper Abstract
As deep learning applications are becoming more and more... |
SkghN205KQ | ICLR.cc/2019/Conference | 2019 | Search-Guided, Lightly-supervised Training of Structured Prediction Energy Networks | ["Amirmohammad Rooshenas", "Dongxu Zhang", "Gopal Sharma", "Andrew McCallum"] | In structured output prediction tasks, labeling ground-truth training output is often expensive. However, for many tasks, even when the true output is unknown, we can evaluate predictions using a scalar reward function, which may be easily assembled from human knowledge or non-differentiable pipelines. But searching ... | ["structured prediction energy networks", "indirect supervision", "search-guided training", "reward functions"] | ABSTRACTIn structured output prediction tasks, labeling ground-truth training output is of-ten expensive. However, for many tasks, even when the true output is unknown,we can evaluate predictions using a scalar reward function, which may be easilyassembled from human knowledge or non-differentiable pipelines. But searc... | rkxbqGkLh7 | Review | 4: Ok but not good enough - rejection | The paper proposes to use a reward function to guide the learning of energy-based models for structured prediction. The idea is to update the energy function based on a random search algorithm guided by a reward function. At each iteration, the SPEN proposes a solution, then a better one is found by the search algorith... | 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
Search-Guided, Lightly-supervised Training of Structured Prediction Energy Networks
### Paper Abstract
In structured output prediction tasks, labeling ground-truth tra... |
dYlO0elyczY | vivoconference.org/VIVO/2020/Conference | 2020 | VIVO-DataConnect: Towards an Architectural Model for Interconnecting Heterogeneous Data Sources to Populate the VIVO Triplestore | ["Michel Heon", "Nicolas Dickner", "Rachid Belkouch", "Alexander J. Jerabek"] | In a large organization, corporate data is rarely stored in a single data source. Data is most often stored sparsely in distributed systems that communicate more or less well with each other. In this context, the integration of a new data source such as VIVO is sometimes perceived as a complexification of the infrastru... | ["Data Loading", "Data Synchronization", "Interoperability", "Enterprise Architecture", "Data Transformation", "Kafka", "Data Graph", "Knowledge Graph"] | IlaAX8vKx_m | Presentation about a common use-case | 7: Good paper, accept | Distributed data sources are the rule rather than the exception. The thoughts outlined here sound very promising, especially if the authors can fulfull their plans "to build a generalizable and adaptive solution to different organizational contexts". | 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
VIVO-DataConnect: Towards an Architectural Model for Interconnecting Heterogeneous Data Sources to Populate the VIVO Triplestore
### Paper Abstract
In a large organizati... | |
ryg2wlSFwS | ICLR.cc/2020/Conference | 2020 | Uncertainty-aware Variational-Recurrent Imputation Network for Clinical Time Series | ["Ahmad Wisnu Mulyadi", "Eunji Jun", "Heung-Il Suk"] | Electronic Health Records (EHR) comprise of longitudinal clinical observations portrayed with sparsity, irregularity, and high-dimensionality which become the major obstacles in drawing reliable downstream outcome. Despite greatly numbers of imputation methods are being proposed to tackle these issues, most of the exis... | ["Missing data imputation", "electronic health Records", "deep generative models", "deep learning"] | ABSTRACTElectronic Health Records (EHR) comprise of longitudinal clinical observationsportrayed with sparsity, irregularity, and high-dimensionality which become themajor obstacles in drawing reliable downstream outcome. Despite greatly num-bers of imputation methods are being proposed to tackle these issues, most of t... | SJx-qKJ6tS | Official Blind Review #1 | 1: Reject | The paper presented a method that combines VAE and uncertainty aware GRU for sequential missing data imputation and outcome prediction.
Pros:
>modeling heterogeneous sequential data with missing value is an important task, especially in the healthcare domain.
>The presentation of the method is easy to follow.
>Th... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Uncertainty-aware Variational-Recurrent Imputation Network for Clinical Time Series
### Paper Abstract
Electronic Health Records (EHR) comprise of longitudinal clinical ... | |
BylctiCctX | ICLR.cc/2019/Conference | 2019 | Guiding Physical Intuition with Neural Stethoscopes | ["Fabian Fuchs", "Oliver Groth", "Adam Kosiorek", "Alex Bewley", "Markus Wulfmeier", "Andrea Vedaldi", "Ingmar Posner"] | Model interpretability and systematic, targeted model adaptation present central challenges in deep learning. In the domain of intuitive physics, we study the task of visually predicting stability of block towers with the goal of understanding and influencing the model's reasoning. Our contributions are two-fold. First... | ["Deep Learning", "Intuitive Physics", "Stability Prediction", "Adversarial Training", "Auxiliary Training", "Multi-Task Learning"] | ABSTRACTModel interpretability and systematic, targeted model adaptation present centralchallenges in deep learning. In the domain of intuitive physics, we study the taskof visually predicting stability of block towers with the goal of understandingand influencing the model’s reasoning. Our contributions are two-fold. ... | SklxtSc93Q | Preliminary report, the idea could be shaped into a better one | 4: Ok but not good enough - rejection | This paper combines the global and local stability prediction and tries to get interpretable results using the stethoscope design, which is actually a weighted subbranch for the main branch. There are several concerns regarding the proposed framework.
1) How to choose \lambda? A better design could be a learnable \la... | 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
Guiding Physical Intuition with Neural Stethoscopes
### Paper Abstract
Model interpretability and systematic, targeted model adaptation present central challenges in dee... |
KRKGJrbPcKE | ICLR.cc/2021/Conference | 2021 | Distribution Based MIL Pooling Filters are Superior to Point Estimate Based Counterparts | ["Mustafa Umit Oner", "Jared Marc Song", "Hwee Kuan Lee", "Wing-Kin Sung"] | Multiple instance learning (MIL) is a machine learning paradigm which learns the mapping between bags of instances and bag labels. There are different MIL tasks which can be solved by different MIL methods. One common component of all MIL methods is the MIL pooling filter, which obtains bag level representations from e... | ["multiple instance learning", "mil", "mil pooling filters", "distribution pooling", "point estimate based pooling"] | ABSTRACTMultiple instance learning (MIL) is a machine learning paradigm which learns themapping between bags of instances and bag labels. There are different MIL taskswhich can be solved by different MIL methods. One common component of allMIL methods is the MIL pooling filter, which obtains bag level representationsfr... | _LNCABv-eC | Gaussian distribution pooling for MIL | 4: Ok but not good enough - rejection | **Summary:**
This paper analyzes pooling functions for Multiple Instance Learning (MIL). The authors presented and compared several point estimate and distribution pooling functions. They also showed some theoretical properties of some point estimate and Gaussian distribution based poolings. They extended the Gaussian... | 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
Distribution Based MIL Pooling Filters are Superior to Point Estimate Based Counterparts
### Paper Abstract
Multiple instance learning (MIL) is a machine learning paradi... |
gVBvtRqU1_ | robot-learning.org/CoRL/2023/Conference | 2023 | OVIR-3D: Open-Vocabulary 3D Instance Retrieval Without Training on 3D Data | ["Shiyang Lu", "Haonan Chang", "Eric Pu Jing", "Abdeslam Boularias", "Kostas Bekris"] | This work presents OVIR-3D, a straightforward yet effective method for open-vocabulary 3D object instance retrieval without using any 3D data for training. Given a language query, the proposed method is able to return a ranked set of 3D object instance segments based on the feature similarity of the instance and the te... | ["Open-Vocabulary", "3D Instance Retrieval"] | OVIR-3D: Open-Vocabulary 3D Instance RetrievalWithout Training on 3D DataShiyang Lu Haonan Chang Eric Pu Jing Abdeslam Boularias Kostas BekrisRutgers Universityhttps://github.com/shiyoung77/OVIR-3DAbstract: This work presents OVIR-3D, a straightforward yet effective methodfor open-vocabulary 3D object instance retrieva... | 8ReVAu-XEv | ## Strengths
The paper is well-written and easy to understand. The approach is simple, and explained in a concise manner, with sufficient detail to enable a skilled practitioner to implement it. From a sequence of input images, an open-vocabulary object detector (here, Detic) is used to obtain 2D bounding boxes. Each ... | 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
OVIR-3D: Open-Vocabulary 3D Instance Retrieval Without Training on 3D Data
### Paper Abstract
This work presents OVIR-3D, a straightforward yet effective method for open... | ||
rklraTNFwB | ICLR.cc/2020/Conference | 2020 | Robust Instruction-Following in a Situated Agent via Transfer-Learning from Text | ["Felix Hill", "Sona Mokra", "Nathaniel Wong", "Tim Harley"] | Recent work has described neural-network-based agents that are trained to execute language-like commands in simulated worlds, as a step towards an intelligent agent or robot that can be instructed by human users. However, the instructions that such agents are trained to follow are typically generated from templates (by... | ["agent", "language", "3D", "simulation", "policy", "instruction", "transfer"] | ABSTRACTRecent work has described neural-network-based agents that are trained to exe-cute language-like commands in simulated worlds, as a step towards an intelligentagent or robot that can be instructed by human users. However, the instructionsthat such agents are trained to follow are typically generated from templa... | r1epwfv15r | Official Blind Review #2 | 3: Weak Reject | This paper considers the task of instruction following where an agent navigates/interacts with a 3D environment conditioned on goals provided in natural language. While several existing approaches use synthetic language for instructions, the authors tackle this problem under the setting of noisy instructions provided b... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Robust Instruction-Following in a Situated Agent via Transfer-Learning from Text
### Paper Abstract
Recent work has described neural-network-based agents that are traine... | |
Sks9_ajex | ICLR.cc/2017/conference | 2017 | Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer | ["Sergey Zagoruyko", "Nikos Komodakis"] | Attention plays a critical role in human visual experience. Furthermore, it has recently been demonstrated that attention can also play an important role in the context of applying artificial neural networks to a variety of tasks from fields such as computer vision and NLP. In this work we show that, by properly defini... | ["Computer vision", "Deep learning", "Supervised Learning"] | ABSTRACTAttention plays a critical role in human visual experience. Furthermore, it hasrecently been demonstrated that attention can also play an important role in thecontext of applying artificial neural networks to a variety of tasks from fields suchas computer vision and NLP. In this work we show that, by properly d... | H1NPJoPNg | official review | 6: Marginally above acceptance threshold | The paper proposes a new way of transferring knowledge.
I like the idea of transferring attention maps instead of activations.
However, the experiments don’t show a big improvement compared with knowledge distillation alone and I think more experiments are required in IMAGENET section.
I would consider updating the sco... | 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
Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer
### Paper Abstract
Attention plays a critical role ... |
SkgEaj05t7 | ICLR.cc/2019/Conference | 2019 | On the Relation Between the Sharpest Directions of DNN Loss and the SGD Step Length | ["Stanis\u0142aw Jastrz\u0119bski", "Zachary Kenton", "Nicolas Ballas", "Asja Fischer", "Yoshua Bengio", "Amos Storkey"] | The training of deep neural networks with Stochastic Gradient Descent (SGD) with a large learning rate or a small batch-size typically ends in flat regions of the weight space, as indicated by small eigenvalues of the Hessian of the training loss. This was found to correlate with a good final generalization performance... | ["optimization", "generalization", "theory of deep learning", "SGD", "hessian"] | Published as a conference paper at ICLR 2019On the relation between the sharpest direc-tions of DNN loss and the SGD step lengthStanisław Jastrzębski1,2, Zachary Kenton2, Nicolas Ballas3, Asja Fischer4, Yoshua Bengio2,5,and Amos Storkey61Jagiellonian University2Mila / University of Montreal3Facebook AI Research4Ruhr U... | BkenqdEc27 | see review | 6: Marginally above acceptance threshold | The paper discusses connections between the properties of DNN loss surfaces and the step length SGD algorithms take, a timely topic. On the whole, reasonably well done, with some interesting observations.
It makes several claims, most notably that there is an initial regime where SGD visits increasingly sharp regions... | 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 Relation Between the Sharpest Directions of DNN Loss and the SGD Step Length
### Paper Abstract
The training of deep neural networks with Stochastic Gradient Desc... |
HJy_5Mcll | ICLR.cc/2017/conference | 2017 | ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation | ["Adam Paszke", "Abhishek Chaurasia", "Sangpil Kim", "Eugenio Culurciello"] | The ability to perform pixel-wise semantic segmentation in real-time is of paramount importance in practical mobile applications. Recent deep neural networks aimed at this task have the disadvantage of requiring a large number of floating point operations and have long run-times that hinder their usability. In this pap... | ["Deep learning"] | ABSTRACTThe ability to perform pixel-wise semantic segmentation in real-time is ofparamount importance in practical mobile applications. Recent deep neural net-works aimed at this task have the disadvantage of requiring a large number offloating point operations and have long run-times that hinder their usability. In t... | ByyhOMQVx | fast system, but needs more thorough account | 5: Marginally below acceptance threshold | This paper describes a fast image semantic segmentation network. Many different techniques are combined to create a system much faster than the baseline SegNet approach, with accuracy comparable or somewhat worse in most of three datasets evaluated.
The choices and techniques used to achieve these speed optimizations... | 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
ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation
### Paper Abstract
The ability to perform pixel-wise semantic segmentation in real-time is o... |
3uhxsmMACaC | graphicsinterface.org/Graphics_Interface/2020/Conference | 2020 | Evaluating Temporal Delays and Spatial Gaps in Overshoot-avoiding Mouse-pointing Operations | ["Shota Yamanaka"] | For hover-based UIs (e.g., pop-up windows) and scrollable UIs, we investigated mouse-pointing performance for users trying to avoid overshooting a target while aiming for it. Three experiments were conducted with a 1D pointing task in which overshooting was accepted (a) within a temporal delay, (b) via a spatial gap be... | ["delay", "temporal delays", "spatial gaps", "operations", "target", "gap", "movement times", "uis", "windows", "scrollable uis"] | ABSTRACTFor hover-based UIs (e.g., pop-up windows) and scrollable UIs,we investigated mouse-pointing performance for users trying toavoid overshooting a target while aiming for it. Three experimentswere conducted with a 1D pointing task in which overshooting wasaccepted (a) within a temporal delay, (b) via a spatial ga... | e990sE8BvqZ | Good paper on gaps and delays on mouse-pointing targets, three studies and with Fitt's law evaluation | 7: Good paper, accept | This paper studies the effects of gaps around and time delays on UI elements that are sensitive to overshooting. These elements were studied as the targets of mouse-pointing tasks and evaluated using Fitt's law.
The works is very well motivated by two common usage scenarios: popup elements and scrolling type targetin... | 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
Evaluating Temporal Delays and Spatial Gaps in Overshoot-avoiding Mouse-pointing Operations
### Paper Abstract
For hover-based UIs (e.g., pop-up windows) and scrollable ... |
GlytqeT39Nc | vivoconference.org/VIVO/2020/Conference | 2020 | euroCRIS Track at the 2020 VIVO annual conference: "VIVO in the euroCRIS Directory of Research Information Systems (DRIS): a growing presence, multiple use cases" | ["Pablo De Castro"] | Taking the under-construction euroCRIS DRIS as a starting point, the session will feature an introductory presentation by a euroCRIS representative exploring the geographic distribution and the various configurations for the VIVO implementations listed in the directory. A number of case study presentations will follow ... | ["VIVO", "research information management", "current research information systems", "system interoperability", "case studies"] | WuKsMKvCMIb | Exciting opportunity to explore collaboration | 9: Top 15% of accepted papers, strong accept | The presence of the euroCRIS track at the VIVO conference represents an important opportunity to explore collaboration between the projects. The four presentations in this session, and the roundtable, should develop questions regarding the relationship between CRIS systems, the DRIS, and VIVO systems. The perspective... | 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
euroCRIS Track at the 2020 VIVO annual conference: "VIVO in the euroCRIS Directory of Research Information Systems (DRIS): a growing presence, multiple use cases"
### Pa... | |
uCY5MuAxcxU | ICLR.cc/2021/Conference | 2021 | Why Are Convolutional Nets More Sample-Efficient than Fully-Connected Nets? | ["Zhiyuan Li", "Yi Zhang", "Sanjeev Arora"] | Convolutional neural networks often dominate fully-connected counterparts in generalization performance, especially on image classification tasks. This is often explained in terms of \textquotedblleft better inductive bias.\textquotedblright\ However, this has not been made mathematically rigorous, and the hurdle is t... | ["sample complexity separation", "equivariance", "convolutional neural networks", "fully-connected"] | ABSTRACTConvolutional neural networks often dominate fully-connected counterparts ingeneralization performance, especially on image classification tasks. This is oftenexplained in terms of “better inductive bias.” However, this has not been mademathematically rigorous, and the hurdle is that the sufficiently wide fully... | uFCqE-aXBOy | Interesting theoretical investigation, some points are unclear | 7: Good paper, accept | The paper studies simple distributional settings in which convolutional neural networks give a provable sample complexity advantage over fully connected networks. This perspective is a valuable complement to prior work in statistical learning theory that often focuses on distribution-free results, which make it harder ... | 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
Why Are Convolutional Nets More Sample-Efficient than Fully-Connected Nets?
### Paper Abstract
Convolutional neural networks often dominate fully-connected counterparts ... |
rkgQL6VFwr | ICLR.cc/2020/Conference | 2020 | Learning Generative Image Object Manipulations from Language Instructions | ["Martin L\u00e4ngkvist", "Andreas Persson", "Amy Loutfi"] | The use of adequate feature representations is essential for achieving high performance in high-level human cognitive tasks in computational modeling. Recent developments in deep convolutional and recurrent neural networks architectures enable learning powerful feature representations from both images and natural langu... | ["language instructions", "objects", "use", "adequate feature representations", "essential", "high performance", "human cognitive tasks", "computational modeling", "recent developments"] | ABSTRACTThe use of adequate feature representations is essential for achieving high perfor-mance in high-level human cognitive tasks in computational modeling. Recentdevelopments in deep convolutional and recurrent neural networks architecturesenable learning powerful feature representations from both images and natura... | ByxOfQOJcr | Official Blind Review #2 | 3: Weak Reject | This paper proposes an architecture for generating images with objects manipulated according to user-specified or conditional instructions. I find the domain very interesting and do believe that tasks like these are critical for learning human-like cognitive capabilities.
This paper is also very clear and easy to foll... | <|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 Generative Image Object Manipulations from Language Instructions
### Paper Abstract
The use of adequate feature representations is essential for achieving high ... | |
SyfIfnC5Ym | ICLR.cc/2019/Conference | 2019 | Improving the Generalization of Adversarial Training with Domain Adaptation | ["Chuanbiao Song", "Kun He", "Liwei Wang", "John E. Hopcroft"] | By injecting adversarial examples into training data, adversarial training is promising for improving the robustness of deep learning models. However, most existing adversarial training approaches are based on a specific type of adversarial attack. It may not provide sufficiently representative samples from the adversa... | ["adversarial training", "domain adaptation", "adversarial example", "deep learning"] | ABSTRACTBy injecting adversarial examples into training data, adversarial training ispromising for improving the robustness of deep learning models. However, mostexisting adversarial training approaches are based on a specific type of adversarialattack. It may not provide sufficiently representative samples from the ad... | BJx8VSaIhQ | Good paper | 6: Marginally above acceptance threshold | This paper addresses the generalization of adversarial training by proposing a new domain adaptation method. In order to have robust defense for adversarial examples, they combine supervised and unsupervised learning for domain adaptation. The idea of domain adaptation is to increase the similarity between clear and ad... | 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
Improving the Generalization of Adversarial Training with Domain Adaptation
### Paper Abstract
By injecting adversarial examples into training data, adversarial training... |
HyezBa4tPB | ICLR.cc/2020/Conference | 2020 | Dirichlet Wrapper to Quantify Classification Uncertainty in Black-Box Systems | ["Jos\u00e9 Mena Rold\u00e1n", "Oriol Pujol Vila", "Jordi Vitri\u00e0 Marca"] | Nowadays, machine learning models are becoming a utility in many sectors. AI companies deliver pre-trained encapsulated models as application programming interfaces (APIs) that developers can combine with third party components, their models, and proprietary data, to create complex data products. This complexity and th... | ["uncertainty", "black-box classifiers", "rejection", "deep learning", "NLP", "CV"] | ABSTRACTNowadays, machine learning models are becoming a utility in many sectors. AIcompanies deliver pre-trained encapsulated models as application programminginterfaces (APIs) that developers can combine with third party components, theirmodels, and proprietary data, to create complex data products. This complexityan... | Byev-9Zwcr | Official Blind Review #4 | 1: Reject | Motivated by real-world challenges in applying pre-trained models, the authors propose a model for selective prediction (prediction with an option for abstention) that wraps an existing black-box classification model. The resulting model output is a Dirichlet distribution with mean equal to the categorical distribution... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Dirichlet Wrapper to Quantify Classification Uncertainty in Black-Box Systems
### Paper Abstract
Nowadays, machine learning models are becoming a utility in many sectors... | |
2xJCo7lgB8 | MIDL.io/2020/Conference | 2020 | Exploring Bayesian Deep Learning Uncertainty Measures for Segmentation of New Lesions in Longitudinal MRIs | ["Nazanin Mohammadi Sepahvand", "Raghav Mehta", "Douglas Lorne Arnold", "Doina Precup", "Tal Arbel"] | In this paper, we develop a modified U-Net architecture to accurately segment new and
enlarging lesions in longitudinal MRI, based on multi-modal MRI inputs, as well as subtrac-
tion images between timepoints, in the context of large-scale clinical trial data for patients
with Multiple Sclerosis (MS). We explore whethe... | ["Multiple Sclerosis", "New and enlarging lesions", "longitudinal MRI", "Bayesian Deep Learning."] | Proceedings of Machine Learning Research { Under Review:1{4, 2020 Short Paper { MIDL 2020 submissionExploring Bayesian Deep Learning Uncertainty Measures forSegmentation of New Lesions in Longitudinal MRIsAbstractIn this paper, we develop a modied U-Net architecture to accurately segment new andenlarging lesions in lon... | P0ujWLgdIQ | MCDO for the segmentation of 'New/Enlarging' lesions, well-written, promising results, but proprietary data | 3: Weak accept | The paper uses MCDO for uncertainty estimation in the segmentation of 'New/Enlarging' brain lesions. it is well written and the problem statement is clear. With the use of uncertainty information, they leap from deterministic segmentation, which can be perilous in the given medical context, to a probabilistic approach.... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Exploring Bayesian Deep Learning Uncertainty Measures for Segmentation of New Lesions in Longitudinal MRIs
### Paper Abstract
In this paper, we develop a modified U-Net ... |
P5UQFFoQ4PJ | eswc-conferences.org/ESWC/2021/Conference/Resources_Track | 2021 | RuBQ 2.0: An Innovated Russian Question Answering Dataset | ["Ivan Rybin", "Vladislav Korablinov", "Pavel Efimov", "Pavel Braslavski"] | We describe the second version of RuBQ, a Russian dataset for knowledge base question answering (KBQA) over Wikidata. Whereas the first version builds on Q&A pairs harvested online, the extension is based on questions obtained through search engine query suggestion services. The questions underwent crowdsourced and in-... | ["knowledge base question answering", "multilingual question answering", "machine reading comprehension", "evaluation resources", "Russian language resources"] | RuBQ 2.0: An Innovated Russian QuestionAnswering DatasetIvan Rybin1?, Vladislav Korablinov1;2, Pavel Emov1, and PavelBraslavski3;4[000000026964458X]1ITMO University, Saint Petersburg, Russiafisrybin,vladislav.korablinov,pavel.vl.efimov g@gmail.com2Yandex, Saint Petersburg, Russia3Ural Federal University, Yekaterinburg,... | olu-AI3eSNq | RuBQ review | 2: Accept | The authors introduce an extension of their data set for answering Russian questions over knowledge graphs. They nearly double the size of the first version (1,410 new questions) by collecting questions from search engine suggestions and also add means for machine-reading comprehension tasks. The data set is of high qu... | 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
RuBQ 2.0: An Innovated Russian Question Answering Dataset
### Paper Abstract
We describe the second version of RuBQ, a Russian dataset for knowledge base question answer... |
r1lZgyBYwS | ICLR.cc/2020/Conference | 2020 | HiLLoC: lossless image compression with hierarchical latent variable models | ["James Townsend", "Thomas Bird", "Julius Kunze", "David Barber"] | We make the following striking observation: fully convolutional VAE models trained on 32x32 ImageNet can generalize well, not just to 64x64 but also to far larger photographs, with no changes to the model. We use this property, applying fully convolutional models to lossless compression, demonstrating a method to scale... | ["compression", "variational inference", "lossless compression", "deep latent variable models"] | ABSTRACTWe make the following striking observation: fully convolutional V AE modelstrained on 3232 ImageNet can generalize well, not just to 64 64 but also to farlarger photographs, with no changes to the model. We use this property, applyingfully convolutional models to lossless compression, demonstrating a method tos... | BygIsY5aKr | Official Blind Review #3 | 8: Accept | Summary:
This paper focuses on lossless source compression with bits back coding for hierarchical fully convolutional VAEs. The focus/contribution is three-fold: 1. Improve the compression rate performance by adapting the discretization of latent space required for the entropy coder ANS. The newly proposed discretizati... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
HiLLoC: lossless image compression with hierarchical latent variable models
### Paper Abstract
We make the following striking observation: fully convolutional VAE models... | |
ZcKPWuhG6wy | ICLR.cc/2021/Conference | 2021 | Tradeoffs in Data Augmentation: An Empirical Study | ["Raphael Gontijo-Lopes", "Sylvia Smullin", "Ekin Dogus Cubuk", "Ethan Dyer"] | Though data augmentation has become a standard component of deep neural network training, the underlying mechanism behind the effectiveness of these techniques remains poorly understood. In practice, augmentation policies are often chosen using heuristics of distribution shift or augmentation diversity. Inspired by the... | ["Generalization", "Interpretability", "Understanding Data Augmentation"] | ABSTRACTThough data augmentation has become a standard component of deep neuralnetwork training, the underlying mechanism behind the effectiveness of thesetechniques remains poorly understood. In practice, augmentation policies are oftenchosen using heuristics of distribution shift or augmentation diversity. Inspired b... | tCS3MXCOhMq | This paper is interesting and can be improved. | 6: Marginally above acceptance threshold | # Summary
This paper analyzes data augmentation for image classification using two measures: Affinity and Diversity. These measures depend on both training data and model and are easy to compute. The authors show that the performance of image classifiers depends on both of these measures through extensive experiments.... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Tradeoffs in Data Augmentation: An Empirical Study
### Paper Abstract
Though data augmentation has become a standard component of deep neural network training, the under... |
ZaYZfu8pT_N | ICLR.cc/2021/Conference | 2021 | Non-iterative Parallel Text Generation via Glancing Transformer | ["Lihua Qian", "Hao Zhou", "Yu Bao", "Mingxuan Wang", "Lin Qiu", "Weinan Zhang", "Yong Yu", "Lei Li"] | Although non-autoregressive models with one-iteration generation achieve remarkable inference speed-up, they still fall behind their autoregressive counterparts in prediction accuracy. The non-autoregressive models with the best accuracy currently rely on multiple decoding iterations, which largely sacrifice the infere... | ["models", "transformer", "parallel text generation", "multiple decoding iterations", "generation", "remarkable inference", "autoregressive counterparts", "prediction accuracy", "best accuracy", "inference speed"] | ABSTRACTAlthough non-autoregressive models with one-iteration generation achieve remark-able inference speed-up, they still fall behind their autoregressive counterpartsin prediction accuracy. The non-autoregressive models with the best accuracycurrently rely on multiple decoding iterations, which largely sacrifice the... | feTdiFsKjcF | Inspiring idea for training "one-iteration" NAT | 7: Good paper, accept | - Overall Comment
This submission improves non-autoregressive translation (NAT) by proposing a non-iterative parallel text generation model called Glancing Transformer (GLAT), which includes the explicit word dependency modeling in NAT via a proposed Glancing Language Model (GLM). Compared to previous work, biggest con... | 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
Non-iterative Parallel Text Generation via Glancing Transformer
### Paper Abstract
Although non-autoregressive models with one-iteration generation achieve remarkable in... |
0QA2qomtW3- | iscaconf.org/ISCA/2023/Workshop/ASSYST | 2023 | Towards Efficient Multi-Agent Learning Systems | ["Kailash Gogineni", "Peng Wei", "Tian Lan", "Guru Prasadh Venkataramani"] | Multi-Agent Reinforcement Learning (MARL) is an increasingly popular domain for modeling and controlling multiple large-scale autonomous systems. Existing multi-agent learning implementations typically involve intensive computations regarding training time and power requirements arising from large observation-action sp... | ["Multi-Agent Systems", "Performance Analysis", "Reinforcement Learning", "Performance Optimization"] | ML for Computer Architecture and Systems (MLArchSys), ISCA, 2023Towards Efficient Multi-Agent Learning SystemsKailash Gogineni, Peng Wei, Tian Lan and Guru VenkataramaniThe George Washington University, Washington, DC, USAE-mail: {kailashg26, pwei, tlan, guruv }@gwu.eduAbstract —Multi-Agent Reinforcement Learning (MARL... | bRGVl8edDwM | Profiling result is interesting, but the optimization is less so | 5: Marginally below acceptance threshold | Multi-Agent Reinforcement Learning (MARL) is an interesting emerging workload to optimize for. The nature of quadratically increased computation and memory requirement in this multi-agent scenario is a challenging and requires more study.
This paper provides an detailed profiling and workload analysis for two popular ... | 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
Towards Efficient Multi-Agent Learning Systems
### Paper Abstract
Multi-Agent Reinforcement Learning (MARL) is an increasingly popular domain for modeling and controllin... |
PRr_3HPakQ | ICLR.cc/2021/Conference | 2021 | Learning to Generate Questions by Recovering Answer-containing Sentences | ["Seohyun Back", "Akhil Kedia", "Sai Chetan Chinthakindi", "Haejun Lee", "Jaegul Choo"] | To train a question answering model based on machine reading comprehension (MRC), significant effort is required to prepare annotated training data composed of questions and their answers from contexts. To mitigate this issue, recent research has focused on synthetically generating a question from a given context and a... | ["Question Generation", "Question Answering", "Data Augmentation", "Machine Reading Comprehension"] | ABSTRACTTo train a question answering model based on machine reading comprehension(MRC), significant effort is required to prepare annotated training data composedof questions and their answers from contexts. To mitigate this issue, recent re-search has focused on synthetically generating a question from a given contex... | up--IA1trW | A new method to pre-train question generation systems for MRC data augmentation which leads to consistent improvements. The initial selection step needs to be exposed better but the findings are otherwise supported by thorough experiments. | 7: Good paper, accept | ### Summary
This work presents a new multi-step method to pre-train a question generation system, which can then be used to create synthetic data to improve a Machine Reading Comprehension system.
First, the authors train a system to identify spans in a text paragraphs which would constitute likely answers for questi... | 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 Generate Questions by Recovering Answer-containing Sentences
### Paper Abstract
To train a question answering model based on machine reading comprehension (M... |
Tp7kI90Htd | ICLR.cc/2021/Conference | 2021 | Generalization in data-driven models of primary visual cortex | ["Konstantin-Klemens Lurz", "Mohammad Bashiri", "Konstantin Willeke", "Akshay Jagadish", "Eric Wang", "Edgar Y. Walker", "Santiago A Cadena", "Taliah Muhammad", "Erick Cobos", "Andreas S. Tolias", "Alexander S Ecker", "Fabian H. Sinz"] | Deep neural networks (DNN) have set new standards at predicting responses of neural populations to visual input. Most such DNNs consist of a convolutional network (core) shared across all neurons which learns a representation of neural computation in visual cortex and a neuron-specific readout that linearly combines t... | ["neuroscience", "cognitive science", "multitask learning", "transfer learning", "representation learning", "network architecture", "computational biology", "visual perception"] | ABSTRACTDeep neural networks (DNN) have set new standards at predicting responses ofneural populations to visual input. Most such DNNs consist of a convolutionalnetwork (core) shared across all neurons which learns a representation of neuralcomputation in visual cortex and a neuron-specific readout that linearly combin... | v0OBKdMoHsl | Innovative state-of-the-art predictor of mouse visual responses | 8: Top 50% of accepted papers, clear accept | Summarize what the paper claims to contribute.
The paper introduces a deep-network-based approach to regression of responses to natural stimuli in mouse primary visual cortex. There is closely related work in the literature, but this paper achieves very good performance, partly through a new way of accounting for neur... | 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
Generalization in data-driven models of primary visual cortex
### Paper Abstract
Deep neural networks (DNN) have set new standards at predicting responses of neural popu... |
SklSQgHFDS | ICLR.cc/2020/Conference | 2020 | Scheduled Intrinsic Drive: A Hierarchical Take on Intrinsically Motivated Exploration | ["Jingwei Zhang", "Niklas Wetzel", "Nicolai Dorka", "Joschka Boedecker", "Wolfram Burgard"] | Exploration in sparse reward reinforcement learning remains an open challenge. Many state-of-the-art methods use intrinsic motivation to complement the sparse extrinsic reward signal, giving the agent more opportunities to receive feedback during exploration. Commonly these signals are added as bonus rewards, which res... | ["Reinforcement Learning", "Exploration", "Intrinsic Motivation", "Sparse Rewards"] | ABSTRACTExploration in sparse reward reinforcement learning remains an open challenge.Many state-of-the-art methods use intrinsic motivation to complement the sparseextrinsic reward signal, giving the agent more opportunities to receive feedbackduring exploration. Commonly these signals are added as bonus rewards, whic... | BJev5lEbcS | Official Blind Review #1 | 8: Accept | This paper tackles the problem of how to integrate intrinsic rewards most effectively, in the episodic sparse reward setting. It has two main technical contributions. The first is Scheduled Intrinsic Drive (SID), which trains two separate policies -- one for maximizing the extrinsic (i.e., task) reward and another for ... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Scheduled Intrinsic Drive: A Hierarchical Take on Intrinsically Motivated Exploration
### Paper Abstract
Exploration in sparse reward reinforcement learning remains an o... | |
SyxS0T4tvS | ICLR.cc/2020/Conference | 2020 | RoBERTa: A Robustly Optimized BERT Pretraining Approach | ["Yinhan Liu", "Myle Ott", "Naman Goyal", "Jingfei Du", "Mandar Joshi", "Danqi Chen", "Omer Levy", "Mike Lewis", "Luke Zettlemoyer", "Veselin Stoyanov"] | Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We ... | ["Deep learning", "language representation learning", "natural language understanding"] | ABSTRACTLanguage model pretraining has led to significant performance gains but carefulcomparison between different approaches is challenging. Training is computation-ally expensive, often done on private datasets of different sizes, and, as we show,hyperparameter choices have significant impact on the final results. W... | ByeB-Xqatr | Official Blind Review #3 | 6: Weak Accept | This paper is a replication study of BERT for training large language models. Its main modification is simple: training longer with more data. Significantly improvements have been reported, and the work achieves on-par or higher accuracy over a large set of downstream tasks compared to XLNet, which is a state-of-the-ar... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
RoBERTa: A Robustly Optimized BERT Pretraining Approach
### Paper Abstract
Language model pretraining has led to significant performance gains but careful comparison bet... | |
awOO1NGKZIx | NeurIPS.cc/2022/Workshop/Federated_Learning | 2022 | On the Vulnerability of Backdoor Defenses for Federated Learning | ["Pei Fang", "Jinghui Chen"] | Federated learning (FL) is a popular distributed machine learning paradigm which enables jointly training a global model without sharing clients' data. However, its repetitive server-client communication gives room for possible backdoor attacks which aims to mislead the global model into a targeted misprediction when a... | ["Federared Learning", "Backdoor Attack"] | On the Vulnerability of Backdoor Defenses forFederated LearningAnonymous Author(s)AffiliationAddressemailAbstractFederated learning (FL) is a distributed machine learning paradigm which enables 1jointly training a global model without sharing clients’ data. However, its repetitive 2server-client communication gives roo... | mbE9nbBjuh | Official Review | Summary
The authors propose a more persistent and stealthy backdoor attack. Specifically, the attack selectively flips the signs of a small proportion of network weights and jointly optimizes the trigger pattern with the model.
Experiments show the effectiveness of the attack against the defenses from three major cat... | <|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 Vulnerability of Backdoor Defenses for Federated Learning
### Paper Abstract
Federated learning (FL) is a popular distributed machine learning paradigm which enab... | ||
rImGHBOL9e5 | ICLR.cc/2022/Workshop/OSC | 2022 | Learning Fourier-Sparse Functions on DAGs | ["Bastian Seifert", "Chris Wendler", "Markus P\u00fcschel"] | We show that the classical Moebius transform from combinatorics can be interpreted as a causal form of Fourier transform on directed acyclic graphs (DAGs). The associated Fourier basis, which is spanned by the columns of the zeta transform, enables us to make use of Fourier-sparse learning methods to learn functions on... | ["causality", "Moebius inversion", "partial order", "directed acyclic graph", "Fourier transform"] | ABSTRACTWe show that the classical Moebius transform from combinatorics can be inter-preted as a causal form of Fourier transform on directed acyclic graphs (DAGs).The associated Fourier basis, which is spanned by the columns of the zeta trans-form, enables us to make use of Fourier-sparse learning methods to learn fun... | BtWeWPv-_Z5 | Interesting idea, elegant paper | 3: Strong accept | The paper discusses Fourier transformations of signals on partial orders and uses these to define convolutions on such signals. For partial orders, a signal $c$ causes a signal $s$ when one sums the value of the predecessors. This is a bijection, whose inverse is the Möbius transform. For any node $q$, it defines a lin... | 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
Learning Fourier-Sparse Functions on DAGs
### Paper Abstract
We show that the classical Moebius transform from combinatorics can be interpreted as a causal form of Fouri... |
pQq3oLH9UmL | ICLR.cc/2021/Conference | 2021 | Achieving Explainability in a Visual Hard Attention Model through Content Prediction | ["Samrudhdhi B. Rangrej", "James J. Clark"] | A visual hard attention model actively selects and observes a sequence of subregions in an image to make a prediction. Unlike in the deep convolution network, in hard attention it is explainable which regions of the image contributed to the prediction. However, the attention policy used by the model to select these reg... | ["visual hard attention", "glimpses", "explainability", "bayesian optimal experiment design", "variational autoencoder"] | ABSTRACTA visual hard attention model actively selects and observes a sequence of subre-gions in an image to make a prediction. Unlike in the deep convolution network,in hard attention it is explainable which regions of the image contributed to theprediction. However, the attention policy used by the model to select th... | p9ETLFbbs76 | Review of "Achieving Explainability in a Visual Hard Attention Model through Content Prediction" | 4: Ok but not good enough - rejection | This paper presents a visual hard-attention image classification model. The difference to standard classification methods such as CNN is that the model provides an explainable inner structure by default, that can be inspected to see what the model focused on. The difference to other state-or-the-art hard-attention mode... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Achieving Explainability in a Visual Hard Attention Model through Content Prediction
### Paper Abstract
A visual hard attention model actively selects and observes a seq... |
r1Chut9xl | ICLR.cc/2017/conference | 2017 | Inference and Introspection in Deep Generative Models of Sparse Data | ["Rahul G. Krishnan", "Matthew Hoffman"] | Deep generative models such as deep latent Gaussian models (DLGMs) are powerful and popular density estimators. However, they have been applied almost exclusively to dense data such as images; DLGMs are rarely applied to sparse, high-dimensional integer data such as word counts or product ratings. One reason is that th... | ["Unsupervised Learning", "Deep learning"] | ABSTRACTDeep generative models such as deep latent Gaussian models (DLGMs) are pow-erful and popular density estimators. However, they have been applied almostexclusively to dense data such as images; DLGMs are rarely applied to sparse,high-dimensional integer data such as word counts or product ratings. One reasonis t... | HyQznHgVe | Weak reject | 5: Marginally below acceptance threshold | This paper presents a small trick to improve the model quality of variational autoencoders (further optimizing the ELBO while initializing it from the predictions of the q network, instead of just using those directly) and the idea of using Jacobian vectors to replace simple embeddings when interpreting variational aut... | 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
Inference and Introspection in Deep Generative Models of Sparse Data
### Paper Abstract
Deep generative models such as deep latent Gaussian models (DLGMs) are powerful a... |
H1lVIxHtPS | ICLR.cc/2020/Conference | 2020 | Learning Multi-Agent Communication Through Structured Attentive Reasoning | ["Murtaza Rangwala", "Ryan Williams"] | Learning communication via deep reinforcement learning has recently been shown to be an effective way to solve cooperative multi-agent tasks. However, learning which communicated information is beneficial for each agent's decision-making remains a challenging task. In order to address this problem, we introduce a fully... | ["Multi-Agent", "Deep Reinforcement Learning", "Communication"] | ABSTRACTLearning communication via deep reinforcement learning has recently beenshown to be an effective way to solve cooperative multi-agent tasks. However,learning which communicated information is beneficial for each agent’s decision-making remains a challenging task. In order to address this problem, we intro-duce ... | HygUdt80FB | Official Blind Review #3 | 3: Weak Reject | The paper presents a new method SARNet for multiagent communication which uses ideas from transformers as well as MAC to perform better than baselines. The model first extracts out important stuff from current information (broadcasted communication from other agents as well own private observation) and then extract imp... | <|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 Multi-Agent Communication Through Structured Attentive Reasoning
### Paper Abstract
Learning communication via deep reinforcement learning has recently been sho... | |
HJB8ntJPG | ICLR.cc/2018/Workshop | 2018 | Multi-Modal Few-Shot Learning: A Benchmark | ["Frederik Pahde", "Moin Nabi", "Tassilo Klein"] | The state-of-the-art deep learning algorithms generally require large amounts of data for model training. Lack thereof can severely deteriorate the performance. To this end, we propose a multi-modal approach that facilitates bridging the information gap by means of meaningful joint embeddings. Specifically, we present ... | ["Few-Shot Learning", "Multi-Modal", "Fine-grained Recognition", "Meta-Learning"] | ABSTRACTThe state-of-the-art deep learning algorithms generally require large amounts ofdata for model training. Lack thereof can severely deteriorate the performance.To this end, we propose a multi-modal approach that facilitates bridging the infor-mation gap by means of meaningful joint embeddings. Specifically, we p... | ByZO0IJFz | interesting idea | 6: Marginally above acceptance threshold | Though the idea of this paper has been explored in some degree in previous works, I still think the idea of this paper is interesting and novel enough | 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
Multi-Modal Few-Shot Learning: A Benchmark
### Paper Abstract
The state-of-the-art deep learning algorithms generally require large amounts of data for model training. L... |
r1xwA34KDB | ICLR.cc/2020/Conference | 2020 | Learning Invariants through Soft Unification | ["Nuri Cingillioglu", "Alessandra Russo"] | Human reasoning involves recognising common underlying principles across many examples by utilising variables. The by-products of such reasoning are invariants that capture patterns across examples such as "if someone went somewhere then they are there" without mentioning specific people or places. Humans learn what va... | ["representation learning", "neural networks", "unification"] | ABSTRACTHuman reasoning involves recognising common underlying principles acrossmany examples by utilising variables. The by-products of such reasoning areinvariants that capture patterns across examples such as “if someone went some-where then they are there” without mentioning specific people or places. Humanslearn w... | BJxGVXNpKB | Official Blind Review #3 | 1: Reject | This paper explores a very interesting idea: can a model learn what variables are and how to use them? Unfortunately, the paper doesn't seem quite ready: the model description was very hard to follow and it's not clear the approach has found a compelling use case.
I read the paper carefully three times, and try as I m... | <|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 Invariants through Soft Unification
### Paper Abstract
Human reasoning involves recognising common underlying principles across many examples by utilising varia... | |
ucEXZQncukK | ICLR.cc/2021/Conference | 2021 | Bayesian Online Meta-Learning | ["Pauching Yap", "Hippolyt Ritter", "David Barber"] | Neural networks are known to suffer from catastrophic forgetting when trained on sequential datasets. While there have been numerous attempts to solve this problem for large-scale supervised classification, little has been done to overcome catastrophic forgetting for few-shot classification problems. Few-shot meta-lear... | ["Bayesian online learning", "few-shot learning", "meta-learning"] | ABSTRACTNeural networks are known to suffer from catastrophic forgetting when trainedon sequential datasets. While there have been numerous attempts to solve thisproblem for large-scale supervised classification, little has been done to over-come catastrophic forgetting for few-shot classification problems. Few-shot me... | k12Rk1IC6qe | Review of Bayesian Online Meta-Learning | 5: Marginally below acceptance threshold | This work proposes a Bayesian approach to meta-learning from sequential data. Two algorithms are proposed. The first is based on the Laplace approximation to the model posterior which is made tractable by using K-FAC approximation of the Hessian. The second approaches uses a variational approximation for the posterior,... | 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
Bayesian Online Meta-Learning
### Paper Abstract
Neural networks are known to suffer from catastrophic forgetting when trained on sequential datasets. While there have b... |
r1eiqi09K7 | ICLR.cc/2019/Conference | 2019 | Riemannian Adaptive Optimization Methods | ["Gary Becigneul", "Octavian-Eugen Ganea"] | Several first order stochastic optimization methods commonly used in the Euclidean domain such as stochastic gradient descent (SGD), accelerated gradient descent or variance reduced methods have already been adapted to certain Riemannian settings. However, some of the most popular of these optimization tools - namely A... | ["Riemannian optimization", "adaptive", "hyperbolic", "curvature", "manifold", "adam", "amsgrad", "adagrad", "rsgd", "convergence"] | ABSTRACTSeveral first order stochastic optimization methods commonly used in the Euclideandomain such as stochastic gradient descent ( SGD), accelerated gradient descentor variance reduced methods have already been adapted to certain Riemanniansettings. However, some of the most popular of these optimization tools name... | HkeQ6djp2X | Riemannian Adam/Amsgrad on product manifolds with convergence guarantee but not supported well by experiments. | 7: Good paper, accept | The paper extends Euclidean optimization methods, Adam/Amsgrad, to the Riemannian setting, and provides theoretical convergence analysis which includes the Euclidean versions as a special case. To avoid breaking the sparsity, coordinate-wise updates are performed on product manifolds.
The empirical performance seems ... | 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
Riemannian Adaptive Optimization Methods
### Paper Abstract
Several first order stochastic optimization methods commonly used in the Euclidean domain such as stochastic ... |
Bklzkh0qFm | ICLR.cc/2019/Conference | 2019 | Relational Graph Attention Networks | ["Dan Busbridge", "Dane Sherburn", "Pietro Cavallo", "Nils Y. Hammerla"] | We investigate Relational Graph Attention Networks, a class of models that extends non-relational graph attention mechanisms to incorporate relational information, opening up these methods to a wider variety of problems. A thorough evaluation of these models is performed, and comparisons are made against established be... | ["RGCN", "attention", "graph convolutional networks", "semi-supervised learning", "graph classification", "molecules"] | Under review as a conference paper at ICLR 2019Relational Graph Attention NetworksAnonymous authorsPaper under double-blind reviewAbstractWe investigate Relational Graph Attention Networks, a class of models thatextends non-relational graph attention mechanisms to incorporate relationalinformation, opening up these met... | BkemLDvAn7 | A good submission but not good enough | 4: Ok but not good enough - rejection | This paper presented a relational graph attention networks that could consider both node
features and relational information (edge features) to perform node-level and graph-level
classifications. The basic idea is to combine the graph attention networks (Veličković et
al. 2017) and the relational graph networks (S... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Relational Graph Attention Networks
### Paper Abstract
We investigate Relational Graph Attention Networks, a class of models that extends non-relational graph attention ... |
H1lPUiRcYQ | ICLR.cc/2019/Conference | 2019 | Computing committor functions for the study of rare events using deep learning with importance sampling | ["Qianxiao Li", "Bo Lin", "Weiqing Ren"] | The committor function is a central object of study in understanding transitions between metastable states in complex systems. However, computing the committor function for realistic systems at low temperatures is a challenging task, due to the curse of dimensionality and the scarcity of transition data. In this paper,... | ["committor function", "rare event", "deep learning", "importance sampling"] | ABSTRACTThe committor function is a central object of study in understanding transitionsbetween metastable states in complex systems. However, computing the commit-tor function for realistic systems at low temperatures is a challenging task, due tothe curse of dimensionality and the scarcity of transition data. In this... | Syxwna5-aQ | summary | 5: Marginally below acceptance threshold | The paper proposes to apply neural network to compute the committor function arose from physics, which looks an interesting application problem by employing machine learning algorithms. Typically, I know very well that the BG distribution usually has multi-modes which makes the sampling difficult extremely. The authors... | 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
Computing committor functions for the study of rare events using deep learning with importance sampling
### Paper Abstract
The committor function is a central object of ... |
B1ldb6NKDr | ICLR.cc/2020/Conference | 2020 | Multi-Agent Hierarchical Reinforcement Learning for Humanoid Navigation | ["Glen Berseth", "Brandon haworth", "Seonghyeon Moon", "Mubbasir Kapadia", "Petros Faloutsos"] | Multi-agent reinforcement learning is a particularly challenging problem. Current
methods have made progress on cooperative and competitive environments with
particle-based agents. Little progress has been made on solutions that could op-
erate in the real world with interaction, dynamics, and humanoid robots. In this
... | ["Multi-Agent Reinforcement Learning", "Reinforcement Learning", "Hierarchical Reinforcement Learning"] | ABSTRACTMulti-agent reinforcement learning is a particularly challenging problem. Cur-rent methods have made progress on cooperative and competitive environmentswith particle-based agents using communication and centralized training. Littleprogress has been made on solutions that could operate in the real world with in... | rJgHSAbAFH | Official Blind Review #3 | 3: Weak Reject | This paper proposes a multi-agent hierarchical reinforcement learning algorithm so that multiple humanoid robots can navigate in multi-agent settings (e.g. avoid collisions, collaboration, chase and escape) in a physically simulated environment. The key difference of this paper with the prior work on MARL is that it us... | <|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-Agent Hierarchical Reinforcement Learning for Humanoid Navigation
### Paper Abstract
Multi-agent reinforcement learning is a particularly challenging problem. Curr... | |
SO73JUgks8 | ICLR.cc/2021/Conference | 2021 | AUBER: Automated BERT Regularization | ["Hyun Dong Lee", "Seongmin Lee", "U Kang"] | How can we effectively regularize BERT? Although BERT proves its effectiveness in various downstream natural language processing tasks, it often overfits when there are only a small number of training instances. A promising direction to regularize BERT is based on pruning its attention heads based on a proxy score for h... | ["BERT Regularization", "Reinforcement Learning", "Automated Regularization"] | ABSTRACTHowcanweeffectivelyregularizeBERT?AlthoughBERTprovesitseffective-nessinvariousdownstreamnaturallanguageprocessingtasks,itoftenoverfitswhenthereareonlyasmallnumberoftraininginstances.Apromisingdirec-tiontoregularizeBERTisbasedonpruningitsattentionheadsbasedonaproxyscoreforheadimportance.However,heuristic-basedme... | gOq9tAX8pP | Official Review | 5: Marginally below acceptance threshold | Summary:
The paper focuses on reducing over-fitting for the BERT model by pruning the attention heads. For determining the order of pruning, the paper uses reinforcement learning along with greedy algorithm to get rid of unimportant attention heads. The experimental section shows the analysis on various design choices... | 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
AUBER: Automated BERT Regularization
### Paper Abstract
How can we effectively regularize BERT? Although BERT proves its effectiveness in various downstream natural lang... |
BkGakb9lx | ICLR.cc/2017/conference | 2017 | RenderGAN: Generating Realistic Labeled Data | ["Leon Sixt", "Benjamin Wild", "Tim Landgraf"] | Deep Convolutional Neuronal Networks (DCNNs) are showing remarkable performance on many computer vision tasks. Due to their large parameter space, they require many labeled samples when trained in a supervised setting. The costs of annotating data manually can render the use of DCNNs infeasible. We present a novel fr... | ["Unsupervised Learning", "Computer vision", "Deep learning", "Applications"] | ABSTRACTDeep Convolutional Neuronal Networks (DCNNs) are showing remarkable per-formance on many computer vision tasks. Due to their large parameter space,they require many labeled samples when trained in a supervised setting. The costsof annotating data manually can render the use of DCNNs infeasible. We present anove... | HkosYQzEx | The proposed model has potential merits, but the paper is missing a critical baseline in the evaluation. | 5: Marginally below acceptance threshold | This paper addresses the problem of decoding barcode-like markers depicted in an image. The main insight is to train a CNN from generated data produced from a GAN. The GAN is trained using unlabeled images, and leverages a "3D model" that undergoes learnt image transformations (e.g., blur, lighting, background). 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
RenderGAN: Generating Realistic Labeled Data
### Paper Abstract
Deep Convolutional Neuronal Networks (DCNNs) are showing remarkable performance on many computer vision t... |
rk7YG_4Yg | ICLR.cc/2017/workshop | 2017 | THE PREIMAGE OF RECTIFIER NETWORK ACTIVITIES | ["Stefan Carlsson", "Hossein Azizpour", "Ali Razavian", "Josephine Sullivan", "Kevin Smith"] | We give a procedure for explicitely computing the complete preimage of activities of a layer in a rectifier network with fully connected layers, from knowledge of the weights in the network. The most general characterization of preimages is as piecewise linear manifolds in the input space with possibly multiple branche... | ["preimage", "rectifier network activities", "procedure", "preimages", "complete preimage", "activities", "layer", "rectifier network", "layers", "knowledge"] | ABSTRACTWe give a procedure for explicitly computing the complete preimage of activ-ities of a layer in a rectifier network with fully connected layers, from knowl-edge of the weights in the network. The most general characterisation of preim-ages is as piecewise linear manifolds in the input space with possibly multip... | Sks_Z7xoe | Interesting paper, can be streamlined | 7: Good paper, accept | This paper studies the question: what is the set of points in the input space of a linear + relu layer that map to a given point in output space? This question has been studied before in an empirical manner (for multi-layer networks), but this paper is the first to give a simple linear algebraic answer.
The central co... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
THE PREIMAGE OF RECTIFIER NETWORK ACTIVITIES
### Paper Abstract
We give a procedure for explicitely computing the complete preimage of activities of a layer in a rectifi... |
hkMoYYEkBoI | ICLR.cc/2021/Conference | 2021 | Computational Separation Between Convolutional and Fully-Connected Networks | ["eran malach", "Shai Shalev-Shwartz"] | Convolutional neural networks (CNN) exhibit unmatched performance in a multitude of computer vision tasks. However, the advantage of using convolutional networks over fully-connected networks is not understood from a theoretical perspective. In this work, we show how convolutional networks can leverage locality in the ... | ["Neural Networks", "Deep Learning", "Convolutional Networks", "Fully-Connected Networks", "Gradient Descent"] | ABSTRACTConvolutional neural networks (CNN) exhibit unmatched performance in a mul-titude of computer vision tasks. However, the advantage of using convolutionalnetworks over fully-connected networks is not understood from a theoretical per-spective. In this work, we show how convolutional networks can leverage localit... | QPzAgFN8FMy | Interesting attempt but the argument seems incomplete. | 5: Marginally below acceptance threshold | This paper tries to provide an explanation for the performance advantage of CNNs over FCNs by trying to identify labelling functions where the performance gap between the two for supervised learning can be proven rigorously. The identify an interesting class of such labelling functions : the pseudo-Boolean functions c... | 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
Computational Separation Between Convolutional and Fully-Connected Networks
### Paper Abstract
Convolutional neural networks (CNN) exhibit unmatched performance in a mul... |
S1ln1TNKvH | ICLR.cc/2020/Conference | 2020 | A Harmonic Structure-Based Neural Network Model for Musical Pitch Detection | ["Xian Wang", "Lingqiao Liu", "Qinfeng Shi"] | In this paper we design a harmonic acoustic model for pitch detection. This model arranges conventional convolution and sparse convolution in a way such that the global harmonic patterns captured by sparse convolution are composed of the enough number of local patterns captured by layers of conventional convolution. Wh... | ["musical pitch detection", "automatic music transcription"] | ABSTRACTIn this paper we design a harmonic acoustic model for pitch detection. This modelarranges conventional convolution and sparse convolution in a way such that theglobal harmonic patterns captured by sparse convolution are composed of the largenumber of local patterns captured by layers of conventional convolution... | Byeg1FUmYB | Official Blind Review #3 | 3: Weak Reject | A Harmonic Structure-Based Neural Network Model for Musical Pitch Detection
*Overview*
This paper presents a new architecture for performing frame-wise pitch detection for piano music. Compared to existing models for this task, this study proposes a so-called “sparse convolution layer”, where the distribution of wei... | <|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 Harmonic Structure-Based Neural Network Model for Musical Pitch Detection
### Paper Abstract
In this paper we design a harmonic acoustic model for pitch detection. Thi... | |
HkljioCcFQ | ICLR.cc/2019/Conference | 2019 | MARGINALIZED AVERAGE ATTENTIONAL NETWORK FOR WEAKLY-SUPERVISED LEARNING | ["Yuan Yuan", "Yueming Lyu", "Xi Shen", "Ivor W. Tsang", "Dit-Yan Yeung"] | In weakly-supervised temporal action localization, previous works have failed to locate dense and integral regions for each entire action due to the overestimation of the most salient regions. To alleviate this issue, we propose a marginalized average attentional network (MAAN) to suppress the dominant response of the ... | ["feature aggregation", "weakly supervised learning", "temporal action localization"] | ABSTRACTIn weakly-supervised temporal action localization, previous works have failed tolocate dense and integral regions for each entire action due to the overestimation ofthe most salient regions. To alleviate this issue, we propose a marginalized averageattentional network (MAAN) to suppress the dominant response of... | Syec-wnqn7 | Well executed paper on a reasonable idea | 6: Marginally above acceptance threshold | Summary
This paper proposed a stochastic pooling method over the temporal dimension for weakly-supervised video localization problem. The main motivation is to resolve a problem of discriminative attention that tends to focus on a few discriminative parts of an input data, which is not desirable for the purpose of dens... | 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
MARGINALIZED AVERAGE ATTENTIONAL NETWORK FOR WEAKLY-SUPERVISED LEARNING
### Paper Abstract
In weakly-supervised temporal action localization, previous works have failed ... |
vYVI1CHPaQg | ICLR.cc/2021/Conference | 2021 | A Better Alternative to Error Feedback for Communication-Efficient Distributed Learning | ["Samuel Horv\u00e1th", "Peter Richtarik"] | Modern large-scale machine learning applications require stochastic optimization algorithms to be implemented on distributed computing systems. A key bottleneck of such systems is the communication overhead for exchanging information across the workers, such as stochastic gradients. Among the many techniques proposed t... | ["distributed optimization", "communication efficiency"] | ABSTRACTModern large-scale machine learning applications require stochastic optimizationalgorithms to be implemented on distributed compute systems. A key bottleneckof such systems is the communication overhead for exchanging information (e.g.,stochastic gradients) across the workers. Among the many techniques proposed... | sdv5cFzU_7E | Recommendation to Accept | 9: Top 15% of accepted papers, strong accept | Summary: The paper studies compression of gradients in distributed training with the goal of minimizing communication costs. The main contribution of the paper is the development of an *induced compressor* that takes as input, a possibly biased compressor, and outputs an unbiased compressor with similar error variance... | 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 Better Alternative to Error Feedback for Communication-Efficient Distributed Learning
### Paper Abstract
Modern large-scale machine learning applications require stoch... |
ByxY067CK4 | MIDL.io/2019/Conference/Abstract | 2019 | Cardio-pulmonary Substructure Segmentation of CT images using Convolutional Neural Networks for Clinical Outcome Analysis | ["Rabia Haq", "Alexandra Hotca", "Aditya Apte", "Andreas Rimner", "Joseph O Deasy", "Maria Thor"] | Radiotherapy doses to some cardio-pulmonary substructures may be critical factors in the observed early mortality following radiotherapy for nonsmall cell lung cancer patients. Our goal is to provide an open-source tool to automatically segment cardio-vascular substructures for consistent outcomes analyses. To facilita... | ["Semantic segmentation", "Convolutional Neural Networks", "cardio-pulmonary", "clinical outcomes"] | Proceedings of Machine Learning Research { Under Review:1{5, 2019 Full Paper { MIDL 2019 submissionCardio-pulmonary Substructure Segmentation of CT imagesusing Convolutional Neural Networks for Clinical OutcomeAnalysisRabia Haq1haqr@mskcc.orgAlexandra Hotca1Aditya Apte1Andreas Rimner2Joseph O. Deasy1Maria Thor11Departm... | BJgA529IjE | accept if there is space | 3: Accept | The paper abstract is discussing an relevant and important problem. The proposed approach seems sensible. The initial experimental results also look reasonable. | 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
Cardio-pulmonary Substructure Segmentation of CT images using Convolutional Neural Networks for Clinical Outcome Analysis
### Paper Abstract
Radiotherapy doses to some c... |
bhngY7lHu_ | ICLR.cc/2021/Conference | 2021 | Adaptive N-step Bootstrapping with Off-policy Data | ["Guan Wang", "Dong Yan", "Hang Su", "Jun Zhu"] | The definition of the update target is a crucial design choice in reinforcement learning. Due to the low computation cost and empirical high performance, n-step returns with off-policy data is a widely used update target to bootstrap from scratch. A critical issue of applying n-step returns is to identify... | ["reinforcement learning", "policy evaluation"] | ABSTRACTThe definition of the update target is a crucial design choice in reinforcementlearning. Due to the low computation cost and empirical high performance, n-step returns with off-policy data is a widely used update target to bootstrap fromscratch. A critical issue of applying n-step returns is to identify the opt... | U3QFlQieiEb | Interesting observations with promising experimental result | 5: Marginally below acceptance threshold | ### Evaluation
Although the paper presents interesting observations and some promising experimental results, I recommend rejection. The proposed method is not general, because it is purely based on empirical observations obtained with only SAC and Q-learning. As a result, it is unclear if the proposed method can be com... | 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
Adaptive N-step Bootstrapping with Off-policy Data
### Paper Abstract
The definition of the update target is a crucial design choice in reinforcement learnin... |
hE3JWimujG | ICLR.cc/2021/Conference | 2021 | Cortico-cerebellar networks as decoupled neural interfaces | ["Joseph Pemberton", "Ellen Boven", "Richard Apps", "Rui Ponte Costa"] | The brain solves the credit assignment problem remarkably well. For credit to be correctly assigned across multiple cortical areas a given area should, in principle, wait for others to finish their computation. How the brain deals with this locking problem has remained unclear. Deep learning methods suffer from similar... | ["systems neuroscience", "cerebellum", "neocortex", "decoupled neural interfaces", "deep learning", "decorrelation", "inverse models", "forward models"] | ABSTRACTThe brain solves the credit assignment problem remarkably well. For credit to becorrectly assigned across multiple cortical areas a given area should, in principle,wait for others to finish their computation. How the brain deals with this lockingproblem has remained unclear. Deep learning methods suffer from si... | n0MTBxs_G7 | Sginificant and original hypothesis linking ML and neuroscience; some clarity issues. | 7: Good paper, accept | **General**
The paper presents a very intriguing hypothesis, and I believe that its publication will benefit the community and stimulate fruitful discussion. The model seems to offer a compelling and fairly novel explanation of cerebellar deficits (including non-motor) with a broad significance across the neuroscience... | 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
Cortico-cerebellar networks as decoupled neural interfaces
### Paper Abstract
The brain solves the credit assignment problem remarkably well. For credit to be correctly ... |
ZK6vTvb84s | ICLR.cc/2021/Conference | 2021 | A Trainable Optimal Transport Embedding for Feature Aggregation and its Relationship to Attention | ["Gr\u00e9goire Mialon", "Dexiong Chen", "Alexandre d'Aspremont", "Julien Mairal"] | We address the problem of learning on sets of features, motivated by the need of performing pooling operations in long biological sequences of varying sizes, with long-range dependencies, and possibly few labeled data. To address this challenging task, we introduce a parametrized representation of fixed size, which em... | ["bioinformatics", "optimal transport", "kernel methods", "attention", "transformers"] | ABSTRACTWe address the problem of learning on sets of features, motivated by the need ofperforming pooling operations in long biological sequences of varying sizes, withlong-range dependencies, and possibly few labeled data. To address this challeng-ing task, we introduce a parametrized representation of fixed size, wh... | d7rq4cqrlo | A well-motivated novel embedding/pooling architecture with somewhat weak results | 6: Marginally above acceptance threshold | This paper proposes a kernel embedding for a set/sequence of features, based on the optimal transport distance to a set of references, leading to a fixed-dimensional embedding of variable length sequences.
The set of references can be obtained as cluster centers over the full dataset (unsupervised), or learned based on... | 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 Trainable Optimal Transport Embedding for Feature Aggregation and its Relationship to Attention
### Paper Abstract
We address the problem of learning on sets of featur... |
BkecJjCEuN | ICLR.cc/2019/Workshop/LLD | 2019 | LABEL-EFFICIENT AUDIO CLASSIFICATION THROUGH MULTITASK LEARNING AND SELF-SUPERVISION | ["Tyler Lee", "Ting Gong", "Suchismita Padhy", "Andrew Rouditchenko", "Anthony Ndirango"] | While deep learning has been incredibly successful in modeling tasks with large, carefully curated labeled datasets, its application to problems with limited labeled data remains a challenge. The aim of the present work is to improve the label efficiency of large neural networks operating on audio data through a combin... | ["multitask learning", "self-supervised learning", "end-to-end audio classification"] | ABSTRACTWhile deep learning has been incredibly successful in modeling tasks with large,carefully curated labeled datasets, its application to problems with limited labeleddata remains a challenge. The aim of the present work is to improve the label effi-ciency of large neural networks operating on audio data through a... | Sye52SE_FN | Intriguing approach, but unclear relationship to transfer learning | 3: Marginally above acceptance threshold | The authors propose a technique for improving the performance of audio classification neural networks by simultaneously training them to solve various auxiliary self-supervised tasks. This is achieved by sharing a common "trunk" network with multiple "head" networks, each tailored to solve a specific task. Typically, a... | 2: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
LABEL-EFFICIENT AUDIO CLASSIFICATION THROUGH MULTITASK LEARNING AND SELF-SUPERVISION
### Paper Abstract
While deep learning has been incredibly successful in modeling ta... |
Sklgs0NFvr | ICLR.cc/2020/Conference | 2020 | Learning The Difference That Makes A Difference With Counterfactually-Augmented Data | ["Divyansh Kaushik", "Eduard Hovy", "Zachary Lipton"] | Despite alarm over the reliance of machine learning systems on so-called spurious patterns, the term lacks coherent meaning in standard statistical frameworks. However, the language of causality offers clarity: spurious associations are due to confounding (e.g., a common cause), but not direct or indirect causal effect... | ["humans in the loop", "annotation artifacts", "text classification", "sentiment analysis", "natural language inference"] | ABSTRACTDespite alarm over the reliance of machine learning systems on so-called spuri-ouspatterns, the term lacks coherent meaning in standard statistical frameworks.However, the language of causality offers clarity: spurious associations are due toconfounding (e.g., a common cause), but not direct or indirect causal ... | rylDGP7Zjr | Official Blind Review #1 | 8: Accept | Summary:
The authors take two tasks,sentiment analysis and natural language inference, and identify datasets for them which they counterfactually augment it by asking people over the Amazon Mechanical Turk Platform to change either the sentiment (in the case of sentiment analysis) or the nature of relationship 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
Learning The Difference That Makes A Difference With Counterfactually-Augmented Data
### Paper Abstract
Despite alarm over the reliance of machine learning systems on so... | |
AqkPRUZ-YkO | NeurIPS.cc/2022/Workshop/LaReL | 2022 | Tackling AlfWorld with Action Attention and Common Sense from Language Models | ["Yue Wu", "So Yeon Min", "Yonatan Bisk", "Ruslan Salakhutdinov", "Shrimai Prabhumoye"] | Pre-trained language models (LMs) capture strong prior knowledge about the world. This common sense knowledge can be used in control tasks. However, directly generating actions from LMs may result in a reasonable narrative, but not executable by a low level agent. We propose to instead use the knowledge in LMs to simpl... | ["Language Models", "Zero shot", "AlfWorld", "text games"] | Tackling AlfWorld with Action Attention andCommon Sense from Pretrained LMsYue Wu1, So Yeon Min1, Yonatan Bisk1, Ruslan Salakhutdinov1, Shrimai Prabhumoye1,21Carnegie Mellon University,2Nvidia Research{ywu5,soyeonm,ybisk,rsalakhu}@andrew.cmu.edu, sprabhumoye@nvidia.comAbstractPre-trained language models (LMs) capture s... | NvibP8xFuo | Improvements over Alfworld baseline, need to improve clarity | 6: Marginally above acceptance threshold | The authors improve over the Alfworld benchmark (instruction following domain with text state and action space) via the following two contributions: a) actions are selected via particular action attention block, b) shortening of environment description by filtering out instruction-irrelevant containers through the use ... | 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
Tackling AlfWorld with Action Attention and Common Sense from Language Models
### Paper Abstract
Pre-trained language models (LMs) capture strong prior knowledge about t... |
LXMSvPmsm0g | ICLR.cc/2021/Conference | 2021 | Long Live the Lottery: The Existence of Winning Tickets in Lifelong Learning | ["Tianlong Chen", "Zhenyu Zhang", "Sijia Liu", "Shiyu Chang", "Zhangyang Wang"] | The lottery ticket hypothesis states that a highly sparsified sub-network can be trained in isolation, given the appropriate weight initialization. This paper extends that hypothesis from one-shot task learning, and demonstrates for the first time that such extremely compact and independently trainable sub-networks can... | ["lottery tickets", "winning tickets", "lifelong learning"] | ABSTRACTThe lottery ticket hypothesis states that a highly sparsified sub-network can betrained in isolation, given the appropriate weight initialization. This paper ex-tends that hypothesis from one-shot task learning, and demonstrates for the firsttime that such extremely compact and independently trainable sub-netwo... | S9WnPYbFwl | Long Live the Lottery: The Existence of Winning Tickets in Lifelong Learning | 7: Good paper, accept | ##########################################################################
Summary:
The paper provides an interesting extension of the lottery ticket hypothesis in the lifelong learning setup, showing the existence of these tickets for class incremental learning. The paper also explores top-down and bottom-up ticke... | 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
Long Live the Lottery: The Existence of Winning Tickets in Lifelong Learning
### Paper Abstract
The lottery ticket hypothesis states that a highly sparsified sub-network... |
LvJ8hLSusrv | ICLR.cc/2021/Conference | 2021 | Gradient-based tuning of Hamiltonian Monte Carlo hyperparameters | ["Andrew Campbell", "Wenlong Chen", "Vincent Stimper", "Jos\u00e9 Miguel Hern\u00e1ndez-Lobato", "Yichuan Zhang"] | Hamiltonian Monte Carlo (HMC) is one of the most successful sampling methods in machine learning. However, its performance is significantly affected by the choice of hyperparameter values, which require careful tuning. Existing approaches for automating this task either optimise a proxy for mixing speed or consider the... | ["Hamiltonian Monte Carlo", "HMC", "MCMC", "Variational Inference"] | ABSTRACTHamiltonian Monte Carlo (HMC) is one of the most successful sampling meth-ods in machine learning. However, its performance is significantly affected bythe choice of hyperparameter values, which require careful tuning. Existing ap-proaches for automating this task either optimize a proxy for mixing speed orcons... | 5fGzsvKcotc | an engineering trick with limited practical significance | 4: Ok but not good enough - rejection | The paper proposes a method to optimize the parameters of the Hybrid Monte Carlo (HMC) algorithm (the step size and the diagonal of momentum's covariance matrix). In order to do that, the authors consider the distribution of samples q_T() obtained after T iterations of the algorithm (T accept/reject steps) starting fro... | 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
Gradient-based tuning of Hamiltonian Monte Carlo hyperparameters
### Paper Abstract
Hamiltonian Monte Carlo (HMC) is one of the most successful sampling methods in machi... |
HyevIJStwH | ICLR.cc/2020/Conference | 2020 | Understanding Why Neural Networks Generalize Well Through GSNR of Parameters | ["Jinlong Liu", "Yunzhi Bai", "Guoqing Jiang", "Ting Chen", "Huayan Wang"] | As deep neural networks (DNNs) achieve tremendous success across many application domains, researchers tried to explore in many aspects on why they generalize well. In this paper, we provide a novel perspective on these issues using the gradient signal to noise ratio (GSNR) of parameters during training process of DNNs... | ["DNN", "generalization", "GSNR", "gradient descent"] | ABSTRACTAs deep neural networks (DNNs) achieve tremendous success across many appli-cation domains, researchers tried to explore in many aspects on why they gener-alize well. In this paper, we provide a novel perspective on these issues usingthe gradient signal to noise ratio (GSNR) of parameters during training proces... | B1lyj_jm5B | Official Blind Review #3 | 3: Weak Reject | The paper defines the quantity of "gradient SNR" (GSNR), shows that larger GSNR leads to better generalization, and shows that SGD training of deep networks has large GSNR. It tells a great story on why SGD-trained DNNs have good generalization.
This topic is highly relevant to this conference.
However, I struggle 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
Understanding Why Neural Networks Generalize Well Through GSNR of Parameters
### Paper Abstract
As deep neural networks (DNNs) achieve tremendous success across many app... | |
ryzHXnR5Y7 | ICLR.cc/2019/Conference | 2019 | Select Via Proxy: Efficient Data Selection For Training Deep Networks | ["Cody Coleman", "Stephen Mussmann", "Baharan Mirzasoleiman", "Peter Bailis", "Percy Liang", "Jure Leskovec", "Matei Zaharia"] | At internet scale, applications collect a tremendous amount of data by logging user events, analyzing text, and collecting images. This data powers a variety of machine learning models for tasks such as image classification, language modeling, content recommendation, and advertising. However, training large models over... | ["data selection", "deep learning", "uncertainty sampling"] | ABSTRACTAt internet scale, applications collect a tremendous amount of data by logging userevents, analyzing text, and collecting images. This data powers a variety of ma-chine learning models for tasks such as image classification, language modeling,content recommendation, and advertising. However, training large mode... | rygxBMfw2m | meaningful work, but lacks more supporting evidences | 5: Marginally below acceptance threshold | This paper studies a very simple and intuitive method to boost the training speed of deep neural networks. The authors first train some light weighted proxy models, using these models to rank the data according to its uncertainty, and then pick the most uncertain subset to train the final model. Experiments on CIFAR10/... | 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
Select Via Proxy: Efficient Data Selection For Training Deep Networks
### Paper Abstract
At internet scale, applications collect a tremendous amount of data by logging u... |
B1xRGkHYDS | ICLR.cc/2020/Conference | 2020 | A bi-diffusion based layer-wise sampling method for deep learning in large graphs | ["Yu He", "Shiyang Wen", "Wenjin Wu", "Yan Zhang", "Siran Yang", "Yuan Wei", "Di Zhang", "Guojie Song", "Wei Lin", "Liang Wang", "Bo Zheng"] | The Graph Convolutional Network (GCN) and its variants are powerful models for graph representation learning and have recently achieved great success on many graph-based applications. However, most of them target on shallow models (e.g. 2 layers) on relatively small graphs. Very recently, although many acceleration met... | ["Layerwise Sampling", "Graph Neural Networks", "Attention Mechanism"] | ABSTRACTThe Graph Convolutional Network (GCN) and its variants are powerful models forgraph representation learning and have recently achieved great success on manygraph-based applications. However, most of them target on shallow models (e.g.2 layers) on relatively small graphs. Very recently, although many acceleratio... | HyekiqqfiH | Official Blind Review #4 | 3: Weak Reject | This paper aims at improving the computational efficiency of GCNs to effectively capture information from the larger multi-hop neighborhood. Conventionally, GCNs use information from all the neighbors up to a certain depth; in which case, with consideration of each further hop, the neighborhood size increases exponenti... | <|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 bi-diffusion based layer-wise sampling method for deep learning in large graphs
### Paper Abstract
The Graph Convolutional Network (GCN) and its variants are powerful ... | |
O5MfZRNXTvC | eswc-conferences.org/ESWC/2021/Conference/Research_Track | 2021 | HDT Bitmap Triple Indices for Efficient RDF Data Exploration | ["Maximilian Wenzel", "Birte Glimm", "Thorsten Liebig"] | The exploration of large, unknown RDF data sets is difficult even for users who are familiar with Semantic Web technologies as, e.g., the SPARQL query language. The concept of faceted navigation offers a user-friendly exploration method through filters that are chosen such that no empty result sets occur. However, espe... | ["HDT", "bitmap triples", "indices", "index", "exploration", "faceted navigation", "efficient", "benchmark", "succinct data structures"] | HDT Bitmap Triple Indices for E cient RDF DataExplorationMaximilian Wenzel1, Thorsten Liebig2, and Birte Glimm11Institute of Artificial Intelligence, University of Ulm, Germany,{maximilian.wenzel, birte.glimm}@uni-ulm.de2derivo GmbH, Ulm, Germany, liebig@derivo.deAbstract. The exploration of large, unknown RDF data set... | 5h51TUrVZHv | Solid paper on an important topic, though with modest novelty and some key limitations | 1: Weak Accept | # Summary
The paper describes in-memory indexing approaches to support efficient faceted browsing over RDF graphs. The proposed approach is based on HDT, upon which specialised indexes are built, including a sub-graph index for the current focus of the browsing, indexes on incoming and outgoing edges for particular cl... | 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
HDT Bitmap Triple Indices for Efficient RDF Data Exploration
### Paper Abstract
The exploration of large, unknown RDF data sets is difficult even for users who are famil... |
r1lcM3AcKm | ICLR.cc/2019/Conference | 2019 | RNNs with Private and Shared Representations for Semi-Supervised Sequence Learning | ["Ge Ya Luo", "Jie Fu", "Pengfei Liu", "Zhi Hao Luo", "Chris Pal"] | Training recurrent neural networks (RNNs) on long sequences using backpropagation through time (BPTT) remains a fundamental challenge.
It has been shown that adding a local unsupervised loss term into the optimization objective makes the training of RNNs on long sequences more effective.
While the importance of an un... | ["recurrent neural network", "semi-supervised learning"] | ABSTRACTTraining recurrent neural networks (RNNs) on long sequences using backpropaga-tion through time (BPTT) remains a fundamental challenge. It has been shown thatadding a local unsupervised loss term into the optimization objective makes thetraining of RNNs on long sequences more effective. While the importance of ... | Hkgv_lvT3Q | Review | 3: Clear rejection | Summary of paper: Builds off work from Trinh et al. (2018) that proposed a semi-supervising learning model that combined the main task with an auxiliary task of predicting tokens in the input sequence. This paper proposes improving this two-task setup by splitting the hidden representations into two subsets, one for bo... | 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
RNNs with Private and Shared Representations for Semi-Supervised Sequence Learning
### Paper Abstract
Training recurrent neural networks (RNNs) on long sequences using b... |
r1xYr3C5t7 | ICLR.cc/2019/Conference | 2019 | Neural Message Passing for Multi-Label Classification | ["Jack Lanchantin", "Arshdeep Sekhon", "Yanjun Qi"] | Multi-label classification (MLC) is the task of assigning a set of target labels for a given sample. Modeling the combinatorial label interactions in MLC has been a long-haul challenge. Recurrent neural network (RNN) based encoder-decoder models have shown state-of-the-art performance for solving MLC. However, the sequ... | ["Multi-label Classification", "Graph Neural Networks", "Attention", "Graph Attention"] | ABSTRACTMulti-label classification (MLC) is the task of assigning a set of target labels fora given sample. Modeling the combinatorial label interactions in MLC has beena long-haul challenge. Recurrent neural network (RNN) based encoder-decodermodels have shown state-of-the-art performance for solving MLC. However, the... | Sylw5GMBnm | Graph2Graph without any graph structured inputs or outputs. | 5: Marginally below acceptance threshold | This paper proposes an encoder-decoder model based on the graph representation of inputs and outputs to solve the multi-label classification problem. The proposed model considers the output labels as a fully connected graph where the pair-wise interaction between labels can be modelled.
Overall, although the proposed ... | 2: The reviewer is willing to defend the evaluation, but it is quite likely that the reviewer did not understand central parts of the paper | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Neural Message Passing for Multi-Label Classification
### Paper Abstract
Multi-label classification (MLC) is the task of assigning a set of target labels for a given sam... |
04cII6MumYV | ICLR.cc/2021/Conference | 2021 | A Universal Representation Transformer Layer for Few-Shot Image Classification | ["Lu Liu", "William L. Hamilton", "Guodong Long", "Jing Jiang", "Hugo Larochelle"] | Few-shot classification aims to recognize unseen classes when presented with only a small number of samples. We consider the problem of multi-domain few-shot image classification, where unseen classes and examples come from diverse data sources. This problem has seen growing interest and has inspired the development of... | ["image classification", "unseen classes", "problem", "urt", "classification aims", "small number", "samples", "examples", "diverse data sources"] | ABSTRACTFew-shot classification aims to recognize unseen classes when presented with onlya small number of samples. We consider the problem of multi-domain few-shotimage classification, where unseen classes and examples come from diverse datasources. This problem has seen growing interest and has inspired the developme... | vChIuirDQAs | This work proposed to meta-learn a module for self-attention to select between different domain-specific representation. | 6: Marginally above acceptance threshold | Summary: The paper proposes a meta-trained Universal Representation Transformer (URT) layer, which learn to dynamically re-weight domain-specific representation for classifying given target images. The evaluation on Meta-Dataset shows proposed method achieved competitive performance against compared baselines.
Strengt... | 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 Universal Representation Transformer Layer for Few-Shot Image Classification
### Paper Abstract
Few-shot classification aims to recognize unseen classes when presented... |
aZUGsoLdpa | eswc-conferences.org/ESWC/2021/Workshop/KGCW | 2021 | JenTab: A Toolkit for Semantic Table Annotations | ["Nora Abdelmageed", "Sirko Schindler"] | Tables are a ubiquitous source of structured information. However, their use in automated pipelines is severely affected by conflicts in naming and issues like missing entries or spelling mistakes. The Semantic Web has proven itself a valuable tool in dealing with such issues, allowing the fusion of data from heterogen... | ["knowledge graph", "matching", "tabular data", "semantic annotation"] | JenTab: A Toolkit for Semantic TableAnnotationsNora Abdelmageed14[0000000214056860]and Sirko Schindler2;4[0000000209644457]1Computer Vision Group2Heinz Nixdorf Chair for Distributed Information Systems3Michael Stifel Center Jena, Germany4Friedrich Schiller University Jena, Germanynora.abdelmageed@uni-jena.de, sirko.sch... | SM2-uaEFaY6 | A fair contribution | 7: Good paper, accept | The paper presents JenTab, a new system for the semantic annotation of tables.
The paper is fairly written and clear. JenTab appears to be a useful tool that may be reused by the community. However, the description of the approach is a bit superficial. In particular, section 3.2 should give more details about how map... | 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
JenTab: A Toolkit for Semantic Table Annotations
### Paper Abstract
Tables are a ubiquitous source of structured information. However, their use in automated pipelines i... |
b7ZRqEFXdQ | ICLR.cc/2021/Conference | 2021 | Improving Sequence Generative Adversarial Networks with Feature Statistics Alignment | ["Yekun Chai", "Qiyue Yin", "Junge Zhang"] | Generative Adversarial Networks (GAN) are facing great challenges in synthesizing sequences of discrete elements, such as mode dropping and unstable training. The binary classifier in the discriminator may limit the capacity of learning signals and thus hinder the advance of adversarial training. To address such issues... | ["feature statistics alignment", "signals", "sequence generation", "gan", "great challenges", "sequences", "discrete elements", "mode dropping"] | ABSTRACTGenerative Adversarial Networks (GAN) are facing great challenges in synthesiz-ing sequences of discrete elements, such as mode dropping and unstable training.The binary classifier in the discriminator may limit the capacity of learning signalsand thus hinder the advance of adversarial training. To address such... | CnsHXulT1e | Interesting idea but flawed experiments | 4: Ok but not good enough - rejection | **Main Claim:**
In this work, the authors propose to use the Feature Statistics Alignment paradigm to enrich the learning signal from the discriminator in a sentence generation GAN. The proposed model can generate sentences with better likelihood and BLEU on one synthetic and two real datasets.
**Contributions:**
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
Improving Sequence Generative Adversarial Networks with Feature Statistics Alignment
### Paper Abstract
Generative Adversarial Networks (GAN) are facing great challenges... |
TWDczblpqE | ICLR.cc/2021/Conference | 2021 | Semi-Supervised Audio Representation Learning for Modeling Beehive Strengths | ["Tony Zhang", "Szymon Zmyslony", "Sergei Nozdrenkov", "Matthew Smith", "Brandon Kingsley Hopkins"] | Honey bees are critical to our ecosystem and food security as a pollinator, contributing 35% of our global agriculture yield. In spite of their importance, beekeeping is exclusively dependent on human labor and experience-derived heuristics, while requiring frequent human checkups to ensure the colony is healthy, which... | ["bee", "beehive", "audio", "sound", "computational ethology", "deep learning", "representation learning", "semi-supervised learning", "modeling", "population", "disease"] | ABSTRACTHoney bees are critical to our ecosystem and food security as a pollinator, con-tributing 35% of our global agriculture yield (Klein et al., 2007). In spite of theirimportance, beekeeping is exclusively dependent on human labor and experience-derived heuristics, while requiring frequent human checkups to ensure... | mc9t_77OIKM | Possibly decent contribution in a vertical domain. Uncertain whether it is strong enough for accept | 5: Marginally below acceptance threshold | Authors have implemented semi-supervised model for inferring beehive characteristics (frames, disease severity, and disease type) based on audio sensor and environmental sensor data.
Contributions:
1. Sensor and environmental data collection from beehives together with sparse inspection data. This may be important co... | 3: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Semi-Supervised Audio Representation Learning for Modeling Beehive Strengths
### Paper Abstract
Honey bees are critical to our ecosystem and food security as a pollinato... |
w2Z2OwVNeK | ICLR.cc/2021/Conference | 2021 | Plan-Based Relaxed Reward Shaping for Goal-Directed Tasks | ["Ingmar Schubert", "Ozgur S Oguz", "Marc Toussaint"] | In high-dimensional state spaces, the usefulness of Reinforcement Learning (RL) is limited by the problem of exploration. This issue has been addressed using potential-based reward shaping (PB-RS) previously. In the present work, we introduce Final-Volume-Preserving Reward Shaping (FV-RS). FV-RS relaxes the strict opti... | ["reinforcement learning", "reward shaping", "plan-based reward shaping", "robotics", "robotic manipulation"] | ABSTRACTIn high-dimensional state spaces, the usefulness of Reinforcement Learning (RL)is limited by the problem of exploration. This issue has been addressed usingpotential-based reward shaping (PB-RS) previously. In the present work, we in-troduce Final-V olume-Preserving Reward Shaping (FV-RS). FV-RS relaxes thestri... | Sj34a2kMy7Y | The ideas in the paper are potentially interesting, but the assumptions appear somewhat restrictive, the experiments are unconvincing, and there are a couple of technical details that lack discussion. | 6: Marginally above acceptance threshold | = Overview =
The paper proposes "Asymptotically equivalent reward shaping", a new reward shaping approach that extent potential-based reward shaping to allow for shaping functions that -- even if eventually altering the resulting optimal policy -- retain key elements in the performance of such policy, namely ensuring... | 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
Plan-Based Relaxed Reward Shaping for Goal-Directed Tasks
### Paper Abstract
In high-dimensional state spaces, the usefulness of Reinforcement Learning (RL) is limited b... |
Bye6weHFvB | ICLR.cc/2020/Conference | 2020 | Plan2Vec: Unsupervised Representation Learning by Latent Plans | ["Ge Yang", "Amy Zhang", "Ari Morcos", "Joelle Pineau", "Pieter Abbeel", "Roberto Calandra"] | Creating a useful representation of the world takes more than just rote memorization of individual data samples. This is because fundamentally, we use our internal representation to plan, to solve problems, and to navigate the world. For a representation to be amenable to planning, it is critical for it to embody some ... | ["Unsupervised Learning", "Reinforcement Learning", "Manifold Learning"] | Under review as a conference paper at ICLR 2020Plan2Vec: Unsupervised RepresentationLearning by Latent PlansAnonymous authorsPaper under double-blind reviewAbstractCreating a useful representation of the world takes more than just rotememorization of individual data samples. This is because fundamentally,we use our int... | S1x9U4-I5r | Official Blind Review #1 | 1: Reject | The paper aims at learning a latent representation of images in the setting of sequential decision making. The representation are learned such that there exists a metric that correctly reflects the difference in reachability between points in the neighborhood of the current observation and the goal.
I'm struggeling 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
Plan2Vec: Unsupervised Representation Learning by Latent Plans
### Paper Abstract
Creating a useful representation of the world takes more than just rote memorization of... | |
ByebT3EYvr | ICLR.cc/2020/Conference | 2020 | Single Deep Counterfactual Regret Minimization | ["Eric Steinberger"] | Counterfactual Regret Minimization (CFR) is the most successful algorithm for finding approximate Nash equilibria in imperfect information games. However, CFR's reliance on full game-tree traversals limits its scalability and generality. Therefore, the game's state- and action-space is often abstracted (i.e. simplified... | ["Game Theory", "Deep Reinforcement Learning", "Counterfactual Regret Minimization", "Imperfect Information Games", "Games", "Poker", "Nash Equilibrium"] | ABSTRACTCounterfactual Regret Minimization (CFR) is the most successful algorithm forfinding approximate Nash equilibria in imperfect information games . However,CFR’s reliance on full game-tree traversals limits its scalability and generality.Therefore, the game’s state- and action-space is often abstracted (i.e. simp... | rJe2qeA2tS | Official Blind Review #2 | 6: Weak Accept | The paper builds upon recently introduced Deep CFR, a method to solve large imperfect-information games. Deep CFR, which combines counterfactual regret minimization and deep learning, learns value networks and a strategy network. The paper suggests a modification to avoid learning a strategy network, proving that it is... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Single Deep Counterfactual Regret Minimization
### Paper Abstract
Counterfactual Regret Minimization (CFR) is the most successful algorithm for finding approximate Nash ... | |
SkfQAiA9YX | ICLR.cc/2019/Conference | 2019 | In search of theoretically grounded pruning | ["Filip Svoboda", "Edgar Liberis", "Nicholas D. Lane"] | Deep learning relies on resource-heavy linear algebra operations which can be prohibitively expensive when deploying to constrained embedded and mobile devices, or even when training large-scale networks. One way to reduce a neural network's resource requirements is to sparsify its weight matrices - a process often ref... | ["pruning", "search", "magnitude", "deep learning relies", "linear algebra operations", "expensive", "mobile devices", "networks", "way", "neural network"] | ABSTRACTDeep learning relies on resource-heavy linear algebra operations which can beprohibitively expensive when deploying to constrained embedded and mobile de-vices, or even when training large-scale networks. One way to reduce a neuralnetwork’s resource requirements is to sparsify its weight matrices - a process of... | HygCeTJ5n7 | Good insights on magnitude based pruning but needs more empirical results | 5: Marginally below acceptance threshold | This paper proposes a statistical framework for determining parameter importance beyond magnitude based pruning. The paper derives deep model parameter asymptotic theory to formulate a statistically-grounded pruning criterion which we compare with the magnitude pruning both qualitatively and quantitatively. The paper ... | 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
In search of theoretically grounded pruning
### Paper Abstract
Deep learning relies on resource-heavy linear algebra operations which can be prohibitively expensive when... |
ryECA1kPz | ICLR.cc/2018/Workshop | 2018 | Adaptive Representation Selection in Contextual Bandit with Unlabeled History | ["Baihan Lin", "Guillermo Cecchi", "Djallel Bouneffouf", "Irina Rish"] | We consider an extension of the contextual bandit setting, motivated by several practical applications, where an unlabeled history of contexts can become available for pre-training before the online decision-making begins. We propose an approach for improving the performance of contextual bandit in such setting, via ad... | ["Adaptive Representation", "Embedding Selection", "Machine Learning", "Online Learning", "Reinforcement Learning", "Meta-Learning"] | ABSTRACTWe consider an extension of the contextual bandit setting, motivated by severalpractical applications, where an unlabeled history of contexts can become avail-able for pre-training before the online decision-making begins. We propose anapproach for improving the performance of contextual bandit in such setting,... | HJUlJy1YM | Important problem but lacks crucials precisions | 5: Marginally below acceptance threshold | The papers states an interesting and ambitious problem but I fell that some points would merit some clarifications :
- The exact setting is not very clearly stated (what is the context, what is the reward -here I assume this is good classification of the presented label- , what are the parameters of the dynamics)
- The... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Adaptive Representation Selection in Contextual Bandit with Unlabeled History
### Paper Abstract
We consider an extension of the contextual bandit setting, motivated by ... |
QhOp8oE2Pm | NoDaLiDa/2023/Conference | 2023 | Gamli - Icelandic Oral History Corpus: Design, Collection and Evaluation | ["Luke O'Brien", "Finnur \u00c1g\u00fast Ingimundarson", "J\u00f3n Gu\u00f0nasson", "Stein\u00fe\u00f3r Steingr\u00edmsson"] | We present Gamli, an ASR corpus for Icelandic oral histories, the first of its kind for this language, derived from the Ísmús ethnographic collection. Corpora for oral histories differ in various ways from corpora for general ASR, they contain spontaneous speech, multiple speakers per channel, noisy environments, the e... | ["ASR", "corpus", "oral history", "transcription", "Icelandic", "spontaneous speech"] | Gamli – Icelandic Oral History Corpus:Design, Collection and EvaluationLuke O’BrienTiroluke@tiro.isFinnur Ágúst IngimundarsonThe Árni Magnússon Institutefor Icelandic Studiesfai@hi.isJón GuðnasonUniversityof Reykjavikjg@ru.isSteinþór SteingrímssonThe Árni Magnússon Institutefor Icelandic Studiessteinthor.stein... | Yrvy7OPuWt | speech corpus from oral history | 6: Marginally above acceptance threshold | The paper describes the processing of preexisting data into an ASR corpus for Icelandic.
The paper follows some common processes but might want to cite some of them. For example, the iterative re-training of acoustic models after more material has been aligned has been first outlined for the Spoken Wikipedia Corpora 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
Gamli - Icelandic Oral History Corpus: Design, Collection and Evaluation
### Paper Abstract
We present Gamli, an ASR corpus for Icelandic oral histories, the first of it... |
zFM0Uo_GnYE | ICLR.cc/2021/Conference | 2021 | On the Importance of Looking at the Manifold | ["Nil Adell Mill", "Jannis Born", "Nathaniel Park", "James Hedrick", "Mar\u00eda Rodr\u00edguez Mart\u00ednez", "Matteo Manica"] | Data rarely lies on uniquely Euclidean spaces. Even data typically represented in regular domains, such as images, can have a higher level of relational information, either between data samples or even relations within samples, e.g., how the objects in an image are linked. With this perspective our data points can be e... | ["Topological Learning", "GNN", "VAE"] | ABSTRACTData rarely lies on uniquely Euclidean spaces. Even data typically represented inregular domains, such as images, can have a higher level of relational information,either between data samples or even relations within samples, e.g., how the objectsin an image are linked. With this perspective our data points can... | caC1m8LsR8P | The paper studies the importance of utilising manifold/topology information in prediction tasks. The method is not novel enough and the comparison seems problematic. | 4: Ok but not good enough - rejection | The paper focuses on studying the importance of utilising manifold/topology information for machine learning tasks. To this end, the authors benchmark four different approaches, including VAE, GR-VAE (using graph distances to regularise embedding distances (as shown in Eq. 1)). The paper performs experiments on four ... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
On the Importance of Looking at the Manifold
### Paper Abstract
Data rarely lies on uniquely Euclidean spaces. Even data typically represented in regular domains, such a... |
ryeot3EYPB | ICLR.cc/2020/Conference | 2020 | Bridging the domain gap in cross-lingual document classification | ["Guokun Lai", "Barlas Oguz", "Yiming Yang", "Veselin Stoyanov"] | The scarcity of labeled training data often prohibits the internationalization of NLP models to multiple languages. Cross-lingual understanding has made progress in this area using language universal representations. However, most current approaches focus on the problem as one of aligning language and do not address t... | ["cross-lingual", "document classification", "semi-supervised learning"] | ABSTRACTThe scarcity of labeled training data often prohibits the internationalization of NLPmodels to multiple languages. Cross-lingual understanding has made progressin this area using language universal representations. However, most current ap-proaches focus on the problem as one of aligning language and do not add... | SJxmEE3k9S | Official Blind Review #1 | 3: Weak Reject | The paper considers the problem of cross-lingual document classification with domain gap in source and target languages. In this setting, labeled data is only available in the source language, while the target language has in-domain unlabeled data. The aim of the paper is to tackle the domain and language mismatch prob... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Bridging the domain gap in cross-lingual document classification
### Paper Abstract
The scarcity of labeled training data often prohibits the internationalization of NLP... | |
HyxehhNtvS | ICLR.cc/2020/Conference | 2020 | Why Learning of Large-Scale Neural Networks Behaves Like Convex Optimization | ["Hui Jiang"] | In this paper, we present some theoretical work to explain why simple gradient descent methods are so successful in solving non-convex optimization problems in learning large-scale neural networks (NN). After introducing a mathematical tool called canonical space, we have proved that the objective functions in learnin... | ["function space", "canonical space", "neural networks", "stochastic gradient descent", "disparity matrix"] | ABSTRACTIn this paper, we present some theoretical work to explain why simple gradientdescent methods are so successful in solving non-convex optimization problems inlearning large-scale neural networks (NN). After introducing a mathematical toolcalled canonical space , we have proved that the objective functions in le... | SJxjTucgcr | Official Blind Review #1 | 1: Reject | This paper claims to analyze the global convergence of large-scale neural networks (NNs). Their analysis relies on the idea of “canonical space”, which upon further reading, is nothing more than multi-dimensional Fourier analysis.
While the analysis of NNs is an extremely important problem, it is not clear what result... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Why Learning of Large-Scale Neural Networks Behaves Like Convex Optimization
### Paper Abstract
In this paper, we present some theoretical work to explain why simple gra... | |
HJK6goojf | MIDL.amsterdam/2018/Conference | 2018 | Joint Capsule Segmentation in Ultrasound Images of the Metacarpophalangeal Joint using Convolutional Neural Networks | ["Nelson Martins", "Eva Costa", "Diana Veiga", "Manuel Ferreira", "Miguel Coimbra"] | This work addresses the automatic segmentation of the joint capsule in ultrasound images of the metacarpophalangeal joint using an adapted version of the well known UNet model. These images are used in the diagnosis of rheumatic diseases, one of the main causes of impairment and pain in developed countries. The identif... | ["Semantic Segmentation", "Ultrasound imaging", "Metacarpophalangeal Joint"] | Joint Capsule Segmentation in Ultrasound Images ofthe Metacarpophalangeal Joint using ConvolutionalNeural NetworksNelson MartinsNeadvance, Machine Vision, SABraga, Portugalnmartins@neadvance.comEva CostaNeadvance, Machine Vision, SABraga, Portugalecosta@neadvance.comDiana VeigaNeadvance, Machine Vision, SABraga, Portug... | SkSw29QRz | Using U-net for segmentation of metacarpophalangeal joint | 1: Strong rejection | The authors use an adjusted version of U-net to segment the joint capsule in 2D ultrasound images. The authors introduce five augmentations to the compensate for the reduced number of training samples. The authors also made three adjustments to the U-net introduced by Ronneberger et al.
- Invalid padding to simplify t... | 2: The reviewer is fairly confident that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Joint Capsule Segmentation in Ultrasound Images of the Metacarpophalangeal Joint using Convolutional Neural Networks
### Paper Abstract
This work addresses the automatic... |
BkxdqA4tvB | ICLR.cc/2020/Conference | 2020 | Collapsed amortized variational inference for switching nonlinear dynamical systems | ["Zhe Dong", "Bryan A. Seybold", "Kevin P. Murphy", "Hung H. Bui"] | We propose an efficient inference method for switching nonlinear dynamical systems. The key idea is to learn an inference network which can be used as a proposal distribution for the continuous latent variables, while performing exact marginalization of the discrete latent variables. This allows us to use the reparamet... | ["nonlinear dynamical systems", "amortized variational inference", "efficient inference", "key idea", "inference network", "proposal distribution", "continuous latent variables", "exact marginalization", "discrete latent variables", "reparameterization trick"] | ABSTRACTWe propose an efficient inference method for switching nonlinear dynamical sys-tems. The key idea is to learn an inference network which can be used as a proposaldistribution for the continuous latent variables, while performing exact marginal-ization of the discrete latent variables. This allows us to use the ... | BklPSoBsFB | Official Blind Review #1 | 3: Weak Reject | Thank you for an interesting read.
As far as I understand, the paper claims two contributions:
1. A combination of collapsed variational inference and amortised inference for SNLDS, to make the training pipeline fully differentiable;
2. An improved loss function upon the variational lowerbound (ELBO) to force the mode... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Collapsed amortized variational inference for switching nonlinear dynamical systems
### Paper Abstract
We propose an efficient inference method for switching nonlinear d... | |
ByeGzlrKwH | ICLR.cc/2020/Conference | 2020 | Compression based bound for non-compressed network: unified generalization error analysis of large compressible deep neural network | ["Taiji Suzuki", "Hiroshi Abe", "Tomoaki Nishimura"] | One of the biggest issues in deep learning theory is the generalization ability of networks with huge model size.
The classical learning theory suggests that overparameterized models cause overfitting.
However, practically used large deep models avoid overfitting, which is not well explained by the classical approache... | ["Generalization error", "compression based bound", "local Rademacher complexity"] | ABSTRACTOne of the biggest issues in deep learning theory is the generalization abilityof networks with huge model size. The classical learning theory suggests thatoverparameterized models cause overfitting. However, practically used large deepmodels avoid overfitting, which is not well explained by the classical appro... | Byevc7PaYS | Official Blind Review #2 | 6: Weak Accept | This paper provides generalization bound based on the compression arguments. A key contribution is that instead of bounding the
population risk for the compressed model, this paper manages to give bounds on the non-compressed network following a unified analysis framework. Also, this paper applies the unified framework... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Compression based bound for non-compressed network: unified generalization error analysis of large compressible deep neural network
### Paper Abstract
One of the biggest... | |
rygo9iR9F7 | ICLR.cc/2019/Conference | 2019 | Progressive Weight Pruning Of Deep Neural Networks Using ADMM | ["Shaokai Ye", "Tianyun Zhang", "Kaiqi Zhang", "Jiayu Li", "Kaidi Xu", "Yunfei Yang", "Fuxun Yu", "Jian Tang", "Makan Fardad", "Sijia Liu", "Xiang Chen", "Xue Lin", "Yanzhi Wang"] | Deep neural networks (DNNs) although achieving human-level performance in many domains, have very large model size that hinders their broader applications on edge computing devices. Extensive research work have been conducted on DNN model compression or pruning. However, most of the previous work took heuristic approac... | ["deep learning", "model compression", "optimization", "ADMM", "weight pruning"] | ABSTRACTDeep neural networks (DNNs), although achieving human-level performance inmany domains, have very large model size that hinders their broader applica-tions on edge computing devices. Extensive research work has been conductedon DNN model compression or pruning. However, most of the previous workhas taken heuris... | BJlr4Trw3X | Concerns about proposed method and experiments | 5: Marginally below acceptance threshold | This paper focus on weight pruning for neural network compression. The proposed method is based on ADMM optimization method for neural network loss with constraint on the l_0 norm of weights, proposed in Zhang et al. 2018b. Two improvements, masked retraining and progressive pruning, are introduced. Masked retraining s... | 4: The reviewer is confident but not absolutely certain that the evaluation is correct | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Progressive Weight Pruning Of Deep Neural Networks Using ADMM
### Paper Abstract
Deep neural networks (DNNs) although achieving human-level performance in many domains, ... |
0rfEDZmOlf3 | NeurIPS.cc/2022/Workshop/Federated_Learning | 2022 | Improving Vertical Federated Learning by Efficient Communication with ADMM | ["Chulin Xie", "Pin-Yu Chen", "Ce Zhang", "Bo Li"] | Vertical Federated learning (VFL) allows each client to collect partial features and jointly train the shared model. In this paper, we identified two challenges in VFL: (1) some works directly average the learned feature embeddings and therefore might lose the unique properties of each local feature set; (2) the serve... | ["Federated Learning", "Vertical Federated Learning", "ADMM"] | Improving Vertical Federated Learning by EfficientCommunication with ADMMAnonymous Author(s)AffiliationAddressemailAbstractVertical Federated learning (VFL) allows each client to collect partial features and 1jointly train the shared model. In this paper, we identified two challenges in VFL: 2(1) some works directly av... | 5ljJvZYKlMd | Lack novelty, experimental details and important comparison | This paper considers the multiple linear heads on the server in vertical federated learning and evaluates this design on several datasets.
1. The idea is very straightforward and lacks novelty. Since in vertical federated learning, partial feature sets are stored in different clients, it's obvious to consider two dif... | <|im_start|>system
You are a helpful assistant that reviews papers and provides feedback on their quality.<|im_end|>
<|im_start|>user
### Paper Title
Improving Vertical Federated Learning by Efficient Communication with ADMM
### Paper Abstract
Vertical Federated learning (VFL) allows each client to collect partial fea... | ||
xP37gkVKa_0 | ICLR.cc/2021/Conference | 2021 | Learned Belief Search: Efficiently Improving Policies in Partially Observable Settings | ["Hengyuan Hu", "Adam Lerer", "Noam Brown", "Jakob Nicolaus Foerster"] | Search is an important tool for computing effective policies in single- and multi-agent environments, and has been crucial for achieving superhuman performance in several benchmark fully and partially observable games. However, one major limitation of prior search approaches for partially observable environments is tha... | ["Search", "partially observable games", "multi-agent learning", "Hanabi"] | ABSTRACTSearch is an important tool for computing effective policies in single- and multi-agent environments, and has been crucial for achieving superhuman performancein several benchmark fully and partially observable games. However, one majorlimitation of prior search approaches for partially observable environments ... | j3feMKfzzl | Good paper to read, not convincing presentation. | 5: Marginally below acceptance threshold | This paper proposes a new approach to search in POMDP environments. The main idea is to extend a previous SPARTA search technique with a belief learning ability, called Learned Belief Search (LBS). While the existing work maintains explicit belief representation, LBS is an approximate auto-regressive counterfactual bel... | 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
Learned Belief Search: Efficiently Improving Policies in Partially Observable Settings
### Paper Abstract
Search is an important tool for computing effective policies in... |
HJZvjvJPf | ICLR.cc/2018/Workshop | 2018 | NYSTROM SAMPLING DEPENDS ON THE EIGENSPECTRUM SHAPE OF THE DATA | ["Djallel Bouneffouf"] | Spectral clustering has shown a superior performance in analyzing the cluster structure. However, its computational complexity limits its application in analyzing large-scale data. To address this problem, many low-rank matrix approximating algorithms are proposed, including the Nystrom method – an approach with proven... | ["Clustering", "Nystrom Sampling", "landmark point selection"] | ABSTRACTSpectral clustering has shown a superior performance in analyzing the clusterstructure. However, its computational complexity limits its application in ana-lyzing large-scale data. To address this problem, many low-rank matrix approx-imating algorithms are proposed, including the Nystr ̈om method – an approachw... | HyNPcrxFz | new sampling technique for Nystrom approximation | 5: Marginally below acceptance threshold | The authors consider modifying a prior diversity-based (MS3) landmark sampling scheme for forming Nystrom approximations by oversampling landmarks using MS3, clustering with kmeans, then using the centroids of the clusters as the final landmark points to form a Nystrom approximation (they call this CM3). They further 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
NYSTROM SAMPLING DEPENDS ON THE EIGENSPECTRUM SHAPE OF THE DATA
### Paper Abstract
Spectral clustering has shown a superior performance in analyzing the cluster structur... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.