paper_id
large_stringlengths
9
12
review_id
large_stringlengths
9
12
year
int64
2.02k
2.03k
paper_title
large_stringlengths
6
202
paper_abstract
large_stringlengths
306
4.31k
system_prompt
large_stringclasses
5 values
user_prompt
large_stringclasses
1 value
review_target
large_stringlengths
307
46.1k
zzk231Ms1Ih
TvyIJI1TlM-
2,022
A Theory of Tournament Representations
Real-world tournaments are almost always intransitive. Recent works have noted that parametric models which assume $d$ dimensional node representations can effectively model intransitive tournaments. However, nothing is known about the structure of the class of tournaments that arise out of any fixed $d$ dimensional r...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper studies the theory of tournament representations, i.e. low-rank matrices $M$ whose sign agrees with the sign matrix of a tournament $T$.\n\nThe authors show several properties of such representations, reducing the study to so called $R$-cones, i.e. tournaments where one vertex beats...
zzk231Ms1Ih
mvJh08SL8ke
2,022
A Theory of Tournament Representations
Real-world tournaments are almost always intransitive. Recent works have noted that parametric models which assume $d$ dimensional node representations can effectively model intransitive tournaments. However, nothing is known about the structure of the class of tournaments that arise out of any fixed $d$ dimensional r...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper provides fundamental theories of tournament representations. The authors study two main questions. First they characterize the class of tournaments that can be represented in d dimensions. Second they give lower and upper bounds on the minimum dimension needed to represent a tournam...
zzk231Ms1Ih
C9NhLV0gKlP
2,022
A Theory of Tournament Representations
Real-world tournaments are almost always intransitive. Recent works have noted that parametric models which assume $d$ dimensional node representations can effectively model intransitive tournaments. However, nothing is known about the structure of the class of tournaments that arise out of any fixed $d$ dimensional r...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper studies the relationship between dimensional representation of tournament and their structural characterization. In particular, a relationship is established between rank d tournament and their forbidden configurations in terms of flip classes, introduced by Fisher&Ryan(1995) as a wa...
zzk231Ms1Ih
pUE59TP-ax
2,022
A Theory of Tournament Representations
Real-world tournaments are almost always intransitive. Recent works have noted that parametric models which assume $d$ dimensional node representations can effectively model intransitive tournaments. However, nothing is known about the structure of the class of tournaments that arise out of any fixed $d$ dimensional r...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "A tournament is made by choosing a direction for each of the edges in a complete graph. A tournament can be induced of by skew symmetric matrices M where entries M_{ij} > 0 if and only if (i,j) is an edge. A tournament on n edges can be represented by a set of d-dimensional vectores {h_1, … , ...
zz9hXVhf40
KkJogqpzEwc
2,022
Revisiting Design Choices in Offline Model Based Reinforcement Learning
Offline reinforcement learning enables agents to leverage large pre-collected datasets of environment transitions to learn control policies, circumventing the need for potentially expensive or unsafe online data collection. Significant progress has been made recently in offline model-based reinforcement learning, appro...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "Recent successful model-based offline RL techniques have relied on heuristics to penalize rewards according to the uncertainty of the estimated MDP. This paper reviews the different penalties that have been designed in the literature. The impact and importance of associated hyperparameters suc...
zz9hXVhf40
yfOp35kKfgD
2,022
Revisiting Design Choices in Offline Model Based Reinforcement Learning
Offline reinforcement learning enables agents to leverage large pre-collected datasets of environment transitions to learn control policies, circumventing the need for potentially expensive or unsafe online data collection. Significant progress has been made recently in offline model-based reinforcement learning, appro...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "Model-based offline reinforcement learning algorithms typically involve constructing a pessimistic MDP, which is implemented based on an uncertainty estimation of the learned model. This paper conducts empirical analysis to compare different design choices of the uncertainty estimation in prac...
zz9hXVhf40
L1E57BlqSOY
2,022
Revisiting Design Choices in Offline Model Based Reinforcement Learning
Offline reinforcement learning enables agents to leverage large pre-collected datasets of environment transitions to learn control policies, circumventing the need for potentially expensive or unsafe online data collection. Significant progress has been made recently in offline model-based reinforcement learning, appro...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper provides an evaluation of many of the design choices and hyperparameter decisions made in offline model-based reinforcement learning methods which have emerged recently. Particularly, the empirical study looks at uncertainty penalties used in these methods, as well as hyperparameters...
zz9hXVhf40
pCMrUs15dnj
2,022
Revisiting Design Choices in Offline Model Based Reinforcement Learning
Offline reinforcement learning enables agents to leverage large pre-collected datasets of environment transitions to learn control policies, circumventing the need for potentially expensive or unsafe online data collection. Significant progress has been made recently in offline model-based reinforcement learning, appro...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The authors present an empirical study of several uncertainty quantification heuristics applied to model learning in offline model-based reinforcement learning. Specifically, they consider the basic architecture of MOPO, in which an uncertainty based state-action penalty function is applied on...
zz9hXVhf40
7QJWP1izpM4
2,022
Revisiting Design Choices in Offline Model Based Reinforcement Learning
Offline reinforcement learning enables agents to leverage large pre-collected datasets of environment transitions to learn control policies, circumventing the need for potentially expensive or unsafe online data collection. Significant progress has been made recently in offline model-based reinforcement learning, appro...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper provides detailed analysis of different uncertainty quantifications in Model-based offline RL, from both statistical and empirical perspectives. Further, the paper performs Bayesian optimization to find hyper-parameters and the optimized hyper-parameters perform better empirically. \...
zyrhwrd9EYs
-aEdXoBehI4
2,022
To Impute or Not To Impute? Missing Data in Treatment Effect Estimation
Missing data is a systemic problem in practical scenarios that causes noise and bias when estimating treatment effects. This makes treatment effect estimation from data with missingness a particularly tricky endeavour. A key reason for this is that standard assumptions on missingness are rendered insufficient due to th...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "\"we identify a new missingness mechanism, which we term mixed confounded missingness (MCM), where some missingness determines treatment selection and other missingness is determined by treatment selection.\"\n\nThe author gave a new term called “MCM” which is not something new. MCM is a type ...
zyrhwrd9EYs
Sw15Q7x-eNY
2,022
To Impute or Not To Impute? Missing Data in Treatment Effect Estimation
Missing data is a systemic problem in practical scenarios that causes noise and bias when estimating treatment effects. This makes treatment effect estimation from data with missingness a particularly tricky endeavour. A key reason for this is that standard assumptions on missingness are rendered insufficient due to th...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper studies dealing with missing values in estimating treatment effects. Authors identify a new missingness mechanism, mixed confounded missingness (MCM), including missingness that determines treatment selection and missingness that is determined by treatment selection. The authors sho...
zyrhwrd9EYs
vop8d7q0nC4
2,022
To Impute or Not To Impute? Missing Data in Treatment Effect Estimation
Missing data is a systemic problem in practical scenarios that causes noise and bias when estimating treatment effects. This makes treatment effect estimation from data with missingness a particularly tricky endeavour. A key reason for this is that standard assumptions on missingness are rendered insufficient due to th...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "In this paper, the authors mainly study the problem of missing data in treatment effect estimation and highlight the importance of addressing this problem. The authors propose a selective imputation scheme which is more well suited for addressing missingness in such scenarios. Authors also pre...
zxEfpcmTDnF
zuD4vNVLIxz
2,022
Learning and controlling the source-filter representation of speech with a variational autoencoder
Understanding and controlling latent representations in deep generative models is a challenging yet important problem for analyzing, transforming and generating various types of data. In speech processing, inspiring from the anatomical mechanisms of phonation, the source-filter model considers that speech signals are p...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper shows that the fundamental frequency and formant frequency information is encoded in a speech VAE model.\nThis can be found by using artificially controlled/generated dataset.\nAfter finding how to manipulate the latent space, one can control arbitrary speech samples in a desirable ...
zxEfpcmTDnF
z-B1OnWOtp_
2,022
Learning and controlling the source-filter representation of speech with a variational autoencoder
Understanding and controlling latent representations in deep generative models is a challenging yet important problem for analyzing, transforming and generating various types of data. In speech processing, inspiring from the anatomical mechanisms of phonation, the source-filter model considers that speech signals are p...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper analyzes the latent representations of speech spectrograms learned by unsupervised variational autoencoders (VAEs) and discovers that the VAE learns to model the variation of fundamental frequencies (F0/source) and formant frequencies (filters) using orthogonal subspaces. Based on t...
zxEfpcmTDnF
56VxUr-PO5h
2,022
Learning and controlling the source-filter representation of speech with a variational autoencoder
Understanding and controlling latent representations in deep generative models is a challenging yet important problem for analyzing, transforming and generating various types of data. In speech processing, inspiring from the anatomical mechanisms of phonation, the source-filter model considers that speech signals are p...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper proposes a method for utilizing labeled synthetic data in order to characterize and control the latent space of a VAE trained on individual frames of speech spectrograms.\nKey properties of the data which one might want explicit control over are identified, i.e., pitch and formant fr...
zxEfpcmTDnF
DmvCsvKK-C0
2,022
Learning and controlling the source-filter representation of speech with a variational autoencoder
Understanding and controlling latent representations in deep generative models is a challenging yet important problem for analyzing, transforming and generating various types of data. In speech processing, inspiring from the anatomical mechanisms of phonation, the source-filter model considers that speech signals are p...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper analyzes VAE latent embeddings to extract subspaces that relate to pitch (f0) and formant frequencies (f1 through fN). This is done through first training a frame-synchronous IS-VAE model from clean speech data. The authors pass controlled synthesized speech through the model and o...
zuqcmNVK4c2
4l7GAqeKj4I
2,022
Self-Joint Supervised Learning
Supervised learning is a fundamental framework used to train machine learning systems. A supervised learning problem is often formulated using an i.i.d. assumption that restricts model attention to a single relevant signal at a time when predicting. This contrasts with the human ability to actively use related samples ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The Authors propose a new way of training classifier models. Instead of classifying each i.i.d. example independently during training and inference, they jointly classify a pair (X1, X2) of them, returning a joint distribution of labels (Y1, Y2). The loss function is modified so that the model...
zuqcmNVK4c2
bgq04xhGww
2,022
Self-Joint Supervised Learning
Supervised learning is a fundamental framework used to train machine learning systems. A supervised learning problem is often formulated using an i.i.d. assumption that restricts model attention to a single relevant signal at a time when predicting. This contrasts with the human ability to actively use related samples ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper introduces a pairwise loss function for supervised learning, named self-joint learning. Instead of predicting the conditional distribution of the label given one data sample as in conventional supervised learning, the proposed self-joint learning framework predicts the conditional j...
zuqcmNVK4c2
79tJyc2F8Dg
2,022
Self-Joint Supervised Learning
Supervised learning is a fundamental framework used to train machine learning systems. A supervised learning problem is often formulated using an i.i.d. assumption that restricts model attention to a single relevant signal at a time when predicting. This contrasts with the human ability to actively use related samples ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "Paper propose a simple but effective method of learning. Feeding in a pair of data and learn the join conditional probabilities for predictions. The advantage is that the combination of data goes as the number of ways to partition the original data into pairs. This is a huge number.\n", "main_...
zuDmDfeoB_1
RQljuCcy-hp
2,022
How Does the Task Landscape Affect MAML Performance?
Model-Agnostic Meta-Learning (MAML) has become increasingly popular for training models that can quickly adapt to new tasks via one or few stochastic gradient descent steps. However, the MAML objective is significantly more difficult to optimize compared to standard non-adaptive learning (NAL), and little is understood...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The authors aim to compare theoretically and empirically the ability of MAML and Non-Adaptive Learning (NAL) to different tasks in a multi-task setting, where tasks are sampled independently from the same distribution. They argue that MAML is better suitable for adapting to hard tasks, while N...
zuDmDfeoB_1
XV8T-2YJQ4O
2,022
How Does the Task Landscape Affect MAML Performance?
Model-Agnostic Meta-Learning (MAML) has become increasingly popular for training models that can quickly adapt to new tasks via one or few stochastic gradient descent steps. However, the MAML objective is significantly more difficult to optimize compared to standard non-adaptive learning (NAL), and little is understood...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper analyzes the performance of MAML algorithm under linear regression setting and compares that with the NAL under the same tasks. The authors show that the excess risk is smaller if there is more discrepency in the hardness of tasks and if the optimal solutions of hard tasks locate cl...
zuDmDfeoB_1
nOMar2XTAK-
2,022
How Does the Task Landscape Affect MAML Performance?
Model-Agnostic Meta-Learning (MAML) has become increasingly popular for training models that can quickly adapt to new tasks via one or few stochastic gradient descent steps. However, the MAML objective is significantly more difficult to optimize compared to standard non-adaptive learning (NAL), and little is understood...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper is about finding the conditions under which MAML outperforms standard multi-task learning. In particular, the authors focus on a linear regression setting and show that MAML outperforms NAL under the following two conditions: (i) there must be some discrepancy in hardness among the ...
zuDmDfeoB_1
ofPS6AiV_fA
2,022
How Does the Task Landscape Affect MAML Performance?
Model-Agnostic Meta-Learning (MAML) has become increasingly popular for training models that can quickly adapt to new tasks via one or few stochastic gradient descent steps. However, the MAML objective is significantly more difficult to optimize compared to standard non-adaptive learning (NAL), and little is understood...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper analytically compares the excess risk of two meta-learning methods: MAML and NAL. NAL is the simple baseline where the initialization for optimizing the test tasks is learned as parameter which minimizes the average loss of train tasks. \n\n1. For a particular simple setting of line...
zrdUVVAvcP2
UJH91RO9uJl
2,022
GrASP: Gradient-Based Affordance Selection for Planning
Planning with a learned model is arguably a key component of intelligence. There are several challenges in realizing such a component in large-scale reinforcement learning (RL) problems. One such challenge is dealing effectively with continuous action spaces when using tree-search planning (e.g., it is not feasible to ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper looks incorporating the concept of affordances from ecological psychology into the planning process. \nThe basic premise is that affordances represent the possible relevant actions available to an agent that are potentially moderated by their goals and the state of the world. This h...
zrdUVVAvcP2
76m8u7dGlSC
2,022
GrASP: Gradient-Based Affordance Selection for Planning
Planning with a learned model is arguably a key component of intelligence. There are several challenges in realizing such a component in large-scale reinforcement learning (RL) problems. One such challenge is dealing effectively with continuous action spaces when using tree-search planning (e.g., it is not feasible to ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper proposes a gradient-based method for selecting affordances in planning. Specifically, it includes an affordance module that maps state representations to continuous actions/options. The planning procedure is aided by a learned model with an encoding, dynamics, reward, and value modu...
zrdUVVAvcP2
FCEjZlR6Vs7
2,022
GrASP: Gradient-Based Affordance Selection for Planning
Planning with a learned model is arguably a key component of intelligence. There are several challenges in realizing such a component in large-scale reinforcement learning (RL) problems. One such challenge is dealing effectively with continuous action spaces when using tree-search planning (e.g., it is not feasible to ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper proposes a gradient-based affordance selection, i.e., action/option selection, in addition to the value equivalent modules for tree-expansion procedure(s) in planning.\nThe claim is that GrASP can learn primitive-action and option selection and plan in a continuous state and action s...
zrdUVVAvcP2
Nd4K1G4XxCn
2,022
GrASP: Gradient-Based Affordance Selection for Planning
Planning with a learned model is arguably a key component of intelligence. There are several challenges in realizing such a component in large-scale reinforcement learning (RL) problems. One such challenge is dealing effectively with continuous action spaces when using tree-search planning (e.g., it is not feasible to ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper proposes dealing with tree search planning in continuous action spaces by learning affordances.\nThe authors propose an architecture composed of several modules, where they are able to plan ahead in a tree search manner by learning a module that expands K discrete possible affordance...
zrW-LVXj2k1
aCkK65KHdrt
2,022
On the benefits of maximum likelihood estimation for Regression and Forecasting
We advocate for a practical Maximum Likelihood Estimation (MLE) approach towards designing loss functions for regression and forecasting, as an alternative to the typical approach of direct empirical risk minimization on a specific target metric. The MLE approach is better suited to capture inductive biases such as pri...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper promotes the maximum likelihood estimation over target metric optimization. The main theoretical results is a finite sample error bound. Specific examples and numerical results are provide to further illustrate the investigation.", "main_review": "Strengths:\n\n- It is well known th...
zrW-LVXj2k1
mWSbEDbEHf
2,022
On the benefits of maximum likelihood estimation for Regression and Forecasting
We advocate for a practical Maximum Likelihood Estimation (MLE) approach towards designing loss functions for regression and forecasting, as an alternative to the typical approach of direct empirical risk minimization on a specific target metric. The MLE approach is better suited to capture inductive biases such as pri...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper compares two inferential methods for regression models, the maximum likelihood estimation and the estimation based on loss functions. For that purpose, a quantity is proposed to measure the correctness of an estimator of a regression model. Then it is shown that, under certain condi...
zrW-LVXj2k1
SZL-7j70hED
2,022
On the benefits of maximum likelihood estimation for Regression and Forecasting
We advocate for a practical Maximum Likelihood Estimation (MLE) approach towards designing loss functions for regression and forecasting, as an alternative to the typical approach of direct empirical risk minimization on a specific target metric. The MLE approach is better suited to capture inductive biases such as pri...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper provides theoretical results that favor MLE estimators, in terms of the excess square loss risk, compared to empirical risk estimators under mild assumption. In particular, the paper devises an estimator for Poisson regression and employs an existing estimator for heavy-tail Pareto ...
zq1iJkNk3uN
NfwmpnGvw_s
2,022
Supervision Exists Everywhere: A Data Efficient Contrastive Language-Image Pre-training Paradigm
Recently, large-scale Contrastive Language-Image Pre-training (CLIP) has attracted unprecedented attention for its impressive zero-shot recognition ability and excellent transferability to downstream tasks. However, CLIP is quite data-hungry and requires 400M image-text pairs for pre-training, thereby restricting its a...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper proposes a data-efficient CLIP (DeCLIP) by employing three additional objectives:\n\n- (1) self-supervision within each modality (e.g., vision domain => SimSiam, text domain => masked language model as BERT)\n- (2) Multi-view supervision (e.g., resize random crop for vision modality...
zq1iJkNk3uN
AZrxIVnuhYG
2,022
Supervision Exists Everywhere: A Data Efficient Contrastive Language-Image Pre-training Paradigm
Recently, large-scale Contrastive Language-Image Pre-training (CLIP) has attracted unprecedented attention for its impressive zero-shot recognition ability and excellent transferability to downstream tasks. However, CLIP is quite data-hungry and requires 400M image-text pairs for pre-training, thereby restricting its a...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper proposes DeCLIP to further utilize the data potential by adding three training objectives to CLIP pre-training: 1) inspired by SimSiam and BERT, self-supervised objectives are added to both image and text; 2) they generate different views for both images and text, and apply contrasti...
zq1iJkNk3uN
C7eEUbcCVI
2,022
Supervision Exists Everywhere: A Data Efficient Contrastive Language-Image Pre-training Paradigm
Recently, large-scale Contrastive Language-Image Pre-training (CLIP) has attracted unprecedented attention for its impressive zero-shot recognition ability and excellent transferability to downstream tasks. However, CLIP is quite data-hungry and requires 400M image-text pairs for pre-training, thereby restricting its a...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper proposes DeCLIP with three additions to the original CLIP model: (1) single-modal self supervised pre-training; (2) cross-modal contrastive pretraining across multiple views and (3) cross-modal contrastive pretraining across nearest neighbors in feature space. DeCLIP is able to ach...
zq1iJkNk3uN
NkaP4s_2Iip
2,022
Supervision Exists Everywhere: A Data Efficient Contrastive Language-Image Pre-training Paradigm
Recently, large-scale Contrastive Language-Image Pre-training (CLIP) has attracted unprecedented attention for its impressive zero-shot recognition ability and excellent transferability to downstream tasks. However, CLIP is quite data-hungry and requires 400M image-text pairs for pre-training, thereby restricting its a...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "In this paper, the authors mitigate the data-hungriness of the CLIP model. The authors propose three directions: single-modality self-supervision; multi-view multi-modality contrastive learning, and nearest-neighbor supervision. With the proposed three components, the authors can achieve bette...
ziRLU3Y2PN_
ytBwwamJO4
2,022
Generalized rectifier wavelet covariance models for texture synthesis
State-of-the-art maximum entropy models for texture synthesis are built from statistics relying on image representations defined by convolutional neural networks (CNN). Such representations capture rich structures in texture images, outperforming wavelet-based representations in this regard. However, conversely to neur...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper presents a novel image texture model that is a hybrid of the traditional wavelet constructions, and the more modern machine learning constructions (incorporating only ReLU nonlinearities on the wavelet basis), proves that this class of models is actually akin to phase harmonic functi...
ziRLU3Y2PN_
ScFNCi7rzei
2,022
Generalized rectifier wavelet covariance models for texture synthesis
State-of-the-art maximum entropy models for texture synthesis are built from statistics relying on image representations defined by convolutional neural networks (CNN). Such representations capture rich structures in texture images, outperforming wavelet-based representations in this regard. However, conversely to neur...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper presents a new image representation model based on wavelets and non-linear rectifiers that allows to synthesize complex geometric textures with a better visual quality than previous wavelet-based models. \nThe main interest of the paper is the usage of the mathematical results from ...
ziRLU3Y2PN_
Jp6u6L1toBT
2,022
Generalized rectifier wavelet covariance models for texture synthesis
State-of-the-art maximum entropy models for texture synthesis are built from statistics relying on image representations defined by convolutional neural networks (CNN). Such representations capture rich structures in texture images, outperforming wavelet-based representations in this regard. However, conversely to neur...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper provides an attempt to characterize wavelet-based texture synthesis models contingent on the number of parameters and the nature of the parametrization. Proper baselines are added with respect to other relevant texture synthesis models.", "main_review": "This is a paper that has a g...
ziRLU3Y2PN_
e1_L1peKpQf
2,022
Generalized rectifier wavelet covariance models for texture synthesis
State-of-the-art maximum entropy models for texture synthesis are built from statistics relying on image representations defined by convolutional neural networks (CNN). Such representations capture rich structures in texture images, outperforming wavelet-based representations in this regard. However, conversely to neur...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This work proposed a texture synthesis framework using the rectified wavelet coefficients. The paper claims that the proposed method cand achieve similar quality with the VGG feature based method (Gatys et al. 22015) and random filter based method (RF, Ustyuzhaninov et al 2017) and gets better...
zhynF6JnC4q
56sR7wcCboW
2,022
Adaptive Q-learning for Interaction-Limited Reinforcement Learning
Conventional reinforcement learning (RL) needs an environment to collect fresh data, which is impractical when an online interaction is costly. Offline RL provides an alternative solution by directly learning from the logged dataset. However, it usually yields unsatisfactory performance due to a pessimistic update sche...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper tackles a variation to the offline RL setting, where the agent is allowed some limited number of online interaction steps after learning offline. An algorithm, CGQL, is proposed for this setting and uses the idea that online and offline data should be used in different updates. Expe...
zhynF6JnC4q
O1PzZV75LFC
2,022
Adaptive Q-learning for Interaction-Limited Reinforcement Learning
Conventional reinforcement learning (RL) needs an environment to collect fresh data, which is impractical when an online interaction is costly. Offline RL provides an alternative solution by directly learning from the logged dataset. However, it usually yields unsatisfactory performance due to a pessimistic update sche...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper proposes GCQL, a new RL algorithm that is trained on a mixture of offline data and online interactions. The novel component of the algorithm is a reweighting that balances acting pessimistically on offline data and greedily on online interactions. The authors propose a mixture replay...
zhynF6JnC4q
JCEmdwY_Wjr
2,022
Adaptive Q-learning for Interaction-Limited Reinforcement Learning
Conventional reinforcement learning (RL) needs an environment to collect fresh data, which is impractical when an online interaction is costly. Offline RL provides an alternative solution by directly learning from the logged dataset. However, it usually yields unsatisfactory performance due to a pessimistic update sche...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The authors propose a mixed offline-online RL approach for which they design an algorithm. They propose to maintain 2 separate replay buffers, one for online data and one for offline data, to allow them to sample either an online or offline batch of data when doing an update step, and tailor t...
zhynF6JnC4q
Y8px7UNV0Ma
2,022
Adaptive Q-learning for Interaction-Limited Reinforcement Learning
Conventional reinforcement learning (RL) needs an environment to collect fresh data, which is impractical when an online interaction is costly. Offline RL provides an alternative solution by directly learning from the logged dataset. However, it usually yields unsatisfactory performance due to a pessimistic update sche...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper presents a framework called adaptive Q-learning that integrates the advantage of offline learning and online learning. ", "main_review": "Poor writing. Inadequate experiments design. This paper, at its current status, is not ready to be submitted to conferences like ICLR. \nActionab...
zfmB5vgfaCt
lTB9JBrm6yN
2,022
TransSlowDown: Efficiency Attacks on Neural Machine Translation Systems
Neural machine translation (NMT) systems have received massive attention from academia and industry. Despite a rich set of work focusing on improving NMT systems’ accuracy, the less explored topic of efficiency is also important to NMT systems because of the real-time demand of translation applications. In this paper,...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "TransSlowDown is an attack scheme that aims to targets the computational resources of NMT systems. The authors show that small modifications to benign input sentences can significantly increase the computation NMT needs to process the input, paving the way for denial-of-service attacks on tran...
zfmB5vgfaCt
h5gbQwiJKf
2,022
TransSlowDown: Efficiency Attacks on Neural Machine Translation Systems
Neural machine translation (NMT) systems have received massive attention from academia and industry. Despite a rich set of work focusing on improving NMT systems’ accuracy, the less explored topic of efficiency is also important to NMT systems because of the real-time demand of translation applications. In this paper,...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper proposes an adversarial attack method for NMT, aiming to increase the computation complexity while decoding by introducing perturbations into the input. To achieve that, the authors propose an attacking algorithm that consists of three steps: find the most important source token w.r....
zfmB5vgfaCt
p_UyKCZ2gm
2,022
TransSlowDown: Efficiency Attacks on Neural Machine Translation Systems
Neural machine translation (NMT) systems have received massive attention from academia and industry. Despite a rich set of work focusing on improving NMT systems’ accuracy, the less explored topic of efficiency is also important to NMT systems because of the real-time demand of translation applications. In this paper,...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper focuses on the efficiency of neural machine translation (NMT) systems, proposing a novel attack approach to test the efficiency robustness. The attack approach can be divided into three parts: 1) find the most relevant tokens of inference efficiency; 2) generate adversarial perturba...
zfmB5vgfaCt
V4GCz4JuZKn
2,022
TransSlowDown: Efficiency Attacks on Neural Machine Translation Systems
Neural machine translation (NMT) systems have received massive attention from academia and industry. Despite a rich set of work focusing on improving NMT systems’ accuracy, the less explored topic of efficiency is also important to NMT systems because of the real-time demand of translation applications. In this paper,...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper considers to attack an NMT system from the perspective of efficiency. At a high level, it aims to slightly modify the input such that the target NMT system outputs a long translation, leading to decreased decoding efficiency. To this end, it proposes a method to guide the input modi...
zf_Ll3HZWgy
goW9aCwyBsi
2,022
How Much Can CLIP Benefit Vision-and-Language Tasks?
Most existing Vision-and-Language (V&L) models rely on pre-trained visual encoders, using a relatively small set of manually-annotated data (as compared to web-crawled data), to perceive the visual world. However, it has been observed that large-scale pretraining usually can result in better generalization performance,...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper explores how features from CLIP (Contrastive Language-Image Pre-training, Radford et al., 2021) affect the performance of vision-and-language models across a series of tasks. The authors explore using CLIP as a visual encoder in two settings, plugging its features directly into task...
zf_Ll3HZWgy
2Td7wH2eG1C
2,022
How Much Can CLIP Benefit Vision-and-Language Tasks?
Most existing Vision-and-Language (V&L) models rely on pre-trained visual encoders, using a relatively small set of manually-annotated data (as compared to web-crawled data), to perceive the visual world. However, it has been observed that large-scale pretraining usually can result in better generalization performance,...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper performs empirical analyses of applying the CLIP to various vision-and-language tasks. It demonstrates the potentials of the model in generalizing to different downstream applications, and provides some suggestions for model deployment. Experimental results show that CLIP pretrainin...
zf_Ll3HZWgy
v8_ecERXcu1
2,022
How Much Can CLIP Benefit Vision-and-Language Tasks?
Most existing Vision-and-Language (V&L) models rely on pre-trained visual encoders, using a relatively small set of manually-annotated data (as compared to web-crawled data), to perceive the visual world. However, it has been observed that large-scale pretraining usually can result in better generalization performance,...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper examines how well CLIP's visual encoders are transferred on vision-and-language (VL) tasks. \nThe paper conducted three without-VLP tasks: VQA, image captioning, and vision-and-language navigation, and three with-VLP tasks: VQA, SNLI-VE, and GQA to show CLIP's visual encoders' tran...
zf_Ll3HZWgy
1DmrJVOqcOz
2,022
How Much Can CLIP Benefit Vision-and-Language Tasks?
Most existing Vision-and-Language (V&L) models rely on pre-trained visual encoders, using a relatively small set of manually-annotated data (as compared to web-crawled data), to perceive the visual world. However, it has been observed that large-scale pretraining usually can result in better generalization performance,...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper utilizes recently proposed CLIP-ResNet/ViT visual encoders instead of the standard backbones to conduct a large-scale empirical study for several vision and language tasks. The experimental setup aims to answer the question in the title of this paper by conducting 1) the task-specifi...
zeGpMIt6Pfq
qa2JJrYDc0R
2,022
BioLCNet: Reward-modulated Locally Connected Spiking Neural Networks
Recent studies have shown that convolutional neural networks (CNNs) are not the only feasible solution for image classification. Furthermore, weight sharing and backpropagation used in CNNs do not correspond to the mechanisms present in the primate visual system. To propose a more biologically plausible solution, we de...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper presents a spiking neural network (SNN) architecture for\nimage classification, designed to be more biologically plausible than\ncomparable existing architectures. In particular, it eschews\nconvolutional connectivity in favour of local connectivity, and uses\nSTDP (vanilla and rewa...
zeGpMIt6Pfq
iBhvOdLkInX
2,022
BioLCNet: Reward-modulated Locally Connected Spiking Neural Networks
Recent studies have shown that convolutional neural networks (CNNs) are not the only feasible solution for image classification. Furthermore, weight sharing and backpropagation used in CNNs do not correspond to the mechanisms present in the primate visual system. To propose a more biologically plausible solution, we de...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The authors proposed an SNN model, dubbed BioLCNet, for image classification. The proposed network consists of three layers: 1) input layer, accepting spike trains as input; 2) locally connected (LC) hidden layer; 3) decoding fully connected (FC) output layer. The LC layer is for feature extr...
zeGpMIt6Pfq
N5qHlNdlir0
2,022
BioLCNet: Reward-modulated Locally Connected Spiking Neural Networks
Recent studies have shown that convolutional neural networks (CNNs) are not the only feasible solution for image classification. Furthermore, weight sharing and backpropagation used in CNNs do not correspond to the mechanisms present in the primate visual system. To propose a more biologically plausible solution, we de...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper discusses learning in biological networks, and proposes a model that combines spiking networks, local connections (convolutions without weight sharing) and reward-modulated spike-timing dependent plasticity. All three components are chosen to be more realistic compared to artificial ...
zbZL1s-pBF
5vLBaGy_LYG
2,022
Training-Free Robust Multimodal Learning via Sample-Wise Jacobian Regularization
Multimodal fusion emerges as an appealing technique to improve model performances on many tasks. Nevertheless, the robustness of such fusion methods is rarely involved in the present literature. In this paper, we are the first to propose a training-free robust late-fusion method by exploiting conditional independence a...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper propses a late-fusion algorithm for multimodal learning. This algorithm serves to improve the robustness against adversarial attacks and random corruptions. Assuming that which modality is perturbed, this paper propses to leverage Jacobian regularization and conditional independence...
zbZL1s-pBF
HPDioUk70I6
2,022
Training-Free Robust Multimodal Learning via Sample-Wise Jacobian Regularization
Multimodal fusion emerges as an appealing technique to improve model performances on many tasks. Nevertheless, the robustness of such fusion methods is rarely involved in the present literature. In this paper, we are the first to propose a training-free robust late-fusion method by exploiting conditional independence a...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "In this paper, the authors proposed a training-free late-fusion method for robust multimodal learning. They specifically considering its performance under adversarial attacks and random corruptions which usually confuse the model by introducing noise to the input data. To promote the multimoda...
zbZL1s-pBF
eNjwwnLrxRR
2,022
Training-Free Robust Multimodal Learning via Sample-Wise Jacobian Regularization
Multimodal fusion emerges as an appealing technique to improve model performances on many tasks. Nevertheless, the robustness of such fusion methods is rarely involved in the present literature. In this paper, we are the first to propose a training-free robust late-fusion method by exploiting conditional independence a...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper proposed a training-free robust multimodal learning late-fusion methods via sample-wise Jacobian regularization. The key idea of the work is to minimize the Frobenius norm of a Jacobian matrix, so that the multimodal prediction is stabilized. The paper demonstrate\nthe good efficacy ...
zaALYtvbRlH
XoUHF965-h8
2,022
SpanDrop: Simple and Effective Counterfactual Learning for Long Sequences
Distilling supervision signal from a long sequence to make predictions is a challenging task in machine learning, especially when not all elements in the input sequence contribute equally to the desired output. In this paper, we propose SpanDrop, a simple and effective data augmentation technique that helps models iden...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "In this paper, we instead investigate learning problems for long sequences where not all input elements contribute equally to the desired output. SCANDROP is a simple algorithm to randomly drop segments in a sequence. The authors first establish that when the number of contributing segments is...
zaALYtvbRlH
FzzbMZ14-LE
2,022
SpanDrop: Simple and Effective Counterfactual Learning for Long Sequences
Distilling supervision signal from a long sequence to make predictions is a challenging task in machine learning, especially when not all elements in the input sequence contribute equally to the desired output. In this paper, we propose SpanDrop, a simple and effective data augmentation technique that helps models iden...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper focuses on distilling supervision signal from long sequences. They focus on cases where the input is a long sequence of length n, but the target prediction is determined by a small subset of size m of sequence fragments, where m << n. The authors propose augmenting data by randomly d...
zaALYtvbRlH
9kJd9LkP5k
2,022
SpanDrop: Simple and Effective Counterfactual Learning for Long Sequences
Distilling supervision signal from a long sequence to make predictions is a challenging task in machine learning, especially when not all elements in the input sequence contribute equally to the desired output. In this paper, we propose SpanDrop, a simple and effective data augmentation technique that helps models iden...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This work proposes SPANDROP, a simple variant of dropout, working on the spans of long sequences. SPANDROP randomly ablates parts of a sequence at a time and asks the model to perform the same task to emulate counterfactual learning and achieve input attribution. The method is tested on both t...
zaALYtvbRlH
5SONjJQCFcx
2,022
SpanDrop: Simple and Effective Counterfactual Learning for Long Sequences
Distilling supervision signal from a long sequence to make predictions is a challenging task in machine learning, especially when not all elements in the input sequence contribute equally to the desired output. In this paper, we propose SpanDrop, a simple and effective data augmentation technique that helps models iden...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper proposes a data augmentation method, SpanDrop (and its variant Beta-SpanDrop), for long sequence data, especially where supporting facts take small portions. They provide a theoretical background on their method and evaluate the method on a synthetic task (FindCats) and four real na...
zXne1klXIQ
_ChCHY2Eycq
2,022
Improving Out-of-Distribution Robustness via Selective Augmentation
Machine learning algorithms typically assume that training and test examples are drawn from the same distribution. However, distribution shifts is a common problem in real-world applications and can cause models to perform dramatically worse at test time. In this paper, we specifically consider the problems of domain s...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "Authors introduced approaches aimed at learning invariant predictors across data sources. Rather than using distribution/risk matching schemes as often done by previous work, they propose to train models against mixtures of data points as a means to avoid that models rely on spurious correlati...
zXne1klXIQ
AIjTwOXgDz_
2,022
Improving Out-of-Distribution Robustness via Selective Augmentation
Machine learning algorithms typically assume that training and test examples are drawn from the same distribution. However, distribution shifts is a common problem in real-world applications and can cause models to perform dramatically worse at test time. In this paper, we specifically consider the problems of domain s...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper considers the model robustness under distribution shift brought by domains and subpopulations. Specifically, based on the interpolation scheme in mixup, the authors propose two selection strategies to perform data augmentation, aim at eliminating the spurious correlations and learnin...
zXne1klXIQ
ZU9Dq3GC-Xp
2,022
Improving Out-of-Distribution Robustness via Selective Augmentation
Machine learning algorithms typically assume that training and test examples are drawn from the same distribution. However, distribution shifts is a common problem in real-world applications and can cause models to perform dramatically worse at test time. In this paper, we specifically consider the problems of domain s...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper propose a mixup-style data augmentation method under the data distribution shift context. In particular, data distributions are formulated as mixture of distributions (i.e., domains), and two distribution shift scenarios are considered: (1) domain shift, where the test domain and tr...
zXM0b4hi5_B
XEhGsithtlc
2,022
On the relation between statistical learning and perceptual distances
It has been demonstrated many times that the behavior of the human visual system is connected to the statistics of natural images. Since machine learning relies on the statistics of training data as well, the above connection has interesting implications when using perceptual distances (which mimic the behavior of the ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper presents a mainly theoretical explication of the relationship between natural image statistics and perceptual distances for small image distortions. The paper presents a number of observations linking distances in natural images, autoencoders, and perceptual similarity for humans. T...
zXM0b4hi5_B
1ug4LiQ4hLn
2,022
On the relation between statistical learning and perceptual distances
It has been demonstrated many times that the behavior of the human visual system is connected to the statistics of natural images. Since machine learning relies on the statistics of training data as well, the above connection has interesting implications when using perceptual distances (which mimic the behavior of the ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The authors present few relation between statistical learning and perceptual distances. Specifically, they explain that\n\n1. perceptual distances correlate with image likelihoods;\n2. auto-encoder latent space induced distances of natural images are correlated with training data probability a...
zXM0b4hi5_B
h2Z0NhoA5CM
2,022
On the relation between statistical learning and perceptual distances
It has been demonstrated many times that the behavior of the human visual system is connected to the statistics of natural images. Since machine learning relies on the statistics of training data as well, the above connection has interesting implications when using perceptual distances (which mimic the behavior of the ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "\nThe work presented in this paper aims to analyze the relationships between the probability distribution of the data, perceptual\ndistances, and unsupervised machine learning. Perceptual sensitivity is correlated with the probability of an image \nin its close neighborhood. The paper also exp...
zXM0b4hi5_B
O92p9ItcWyE
2,022
On the relation between statistical learning and perceptual distances
It has been demonstrated many times that the behavior of the human visual system is connected to the statistics of natural images. Since machine learning relies on the statistics of training data as well, the above connection has interesting implications when using perceptual distances (which mimic the behavior of the ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "Interesting (but debatable) observations between image prior, image quality, and perceptual distances.", "main_review": "Strengths:\n\nThe authors investigate a very interesting and important problem that lies in the intersection between perception and statistics.\n\nWeaknesses:\n\nFor Observa...
zU2v47WF0Ku
etJIqhCK7Sw
2,022
Implicit Bias of Linear Equivariant Networks
Group equivariant convolutional neural networks (G-CNNs) are generalizations of convolutional neural networks (CNNs) which excel in a wide range of scientific and technical applications by explicitly encoding particular group symmetries, such as rotations and permutations, in their architectures. Although the success o...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "Group Equivariant CNNs (G-CNNs) generalize CNNs from the translation group to arbitrary groups. It has been shown that overparameterized/undetermined linear transformations and linear CNNs learned by gradient descent are implicitly biased, in that they find zero train loss solutions that minim...
zU2v47WF0Ku
7gKDTygP2n-
2,022
Implicit Bias of Linear Equivariant Networks
Group equivariant convolutional neural networks (G-CNNs) are generalizations of convolutional neural networks (CNNs) which excel in a wide range of scientific and technical applications by explicitly encoding particular group symmetries, such as rotations and permutations, in their architectures. Although the success o...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper describes a theoretical analysis of the implicit bias in G-CNNs. The results show that for linear G-CNNs trained on linearly separable data using GD converge to sparse solutions in the Fourier domain (equivalently dense solutions in the real domain). The theory is theoretically confi...
zU2v47WF0Ku
Xevn5xmUQ7f
2,022
Implicit Bias of Linear Equivariant Networks
Group equivariant convolutional neural networks (G-CNNs) are generalizations of convolutional neural networks (CNNs) which excel in a wide range of scientific and technical applications by explicitly encoding particular group symmetries, such as rotations and permutations, in their architectures. Although the success o...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper studies the induced bias of gradient descent on L-layer linear group-equivariant convolutional neural networks. Their main result is: gradient descent implicitly controls the 2/L-Schatten norm of the Fourier transform of the (linear) predictor. More precisely, they show that gradien...
zU2v47WF0Ku
tsWTCG-eUEJ
2,022
Implicit Bias of Linear Equivariant Networks
Group equivariant convolutional neural networks (G-CNNs) are generalizations of convolutional neural networks (CNNs) which excel in a wide range of scientific and technical applications by explicitly encoding particular group symmetries, such as rotations and permutations, in their architectures. Although the success o...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "Because of the explicit inductive bias of G-CNN, other inductive biases have not been discussed much.\nIn this paper, the author focuses on and analyzes the non-explicit inductive bias of G-CNNs.\nTechnically, they present the non-explicit bias in terms of Fourier matrices by using a group Fou...
zRb7IWkTZAU
ageyidlqlb1
2,022
Zero-Shot Reward Specification via Grounded Natural Language
Reward signals in reinforcement learning can be expensive signals in many tasks and often require access to direct state. The alternative to reward signals are usually demonstrations or goal images which can be labor intensive to collect. Goal text description is a low effort way of communicating the desired task. Goal...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper presents a hybrid learning/rule-based approach to language-conditioned reward specification, particularly for robotic control tasks from images. The core contribution of the work is in specifying the reward 0-shot. Specifically, the method parses a language instruction into an object...
zRb7IWkTZAU
XRmBKWKTXYY
2,022
Zero-Shot Reward Specification via Grounded Natural Language
Reward signals in reinforcement learning can be expensive signals in many tasks and often require access to direct state. The alternative to reward signals are usually demonstrations or goal images which can be labor intensive to collect. Goal text description is a low effort way of communicating the desired task. Goal...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper demonstrates how to use pretrained CLIP model as a reward function for an RL agent. This approach enables flexible goal specification using language. The proposed reward generation method uses CLIP to identify the relevant objects and a separate module to compute the reward based on...
zRb7IWkTZAU
OsW_FKM_UuA
2,022
Zero-Shot Reward Specification via Grounded Natural Language
Reward signals in reinforcement learning can be expensive signals in many tasks and often require access to direct state. The alternative to reward signals are usually demonstrations or goal images which can be labor intensive to collect. Goal text description is a low effort way of communicating the desired task. Goal...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper proposes a framework to train policies for tasks that are specified via language text without the use of any expert trajectories or underlying state information to engineer reward functions. The authors leverage a state-of-the-art visual grounding model (CLIP) to ground object nouns ...
zRb7IWkTZAU
Sf9J7hCAU5h
2,022
Zero-Shot Reward Specification via Grounded Natural Language
Reward signals in reinforcement learning can be expensive signals in many tasks and often require access to direct state. The alternative to reward signals are usually demonstrations or goal images which can be labor intensive to collect. Goal text description is a low effort way of communicating the desired task. Goal...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper proposes an approach for zero-shot reward specification using natural language groundings without expert demos and state information. The work proposes using CLIP’s image+language encoder to find the saliency maps and separately encoding the target locations for the object. These obj...
zRJu6mU2BaE
fyl7sNm1Rm
2,022
ConFeSS: A Framework for Single Source Cross-Domain Few-Shot Learning
Most current few-shot learning methods train a model from abundantly labeled base category data and then transfer and adapt the model to sparsely labeled novel category data. These methods mostly generalize well on novel categories from the same domain as the base categories but perform poorly for distant domain catego...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The problem is to learn by few shots (FSL) when the final domain is highly distant from the source base (i.e. natural, medical and satellite), dubbed single source cross-domain few-shot learning. In short, the approach consists of three steps: 1) train a feature extracting backbone with the co...
zRJu6mU2BaE
RH5ZPeNj99_
2,022
ConFeSS: A Framework for Single Source Cross-Domain Few-Shot Learning
Most current few-shot learning methods train a model from abundantly labeled base category data and then transfer and adapt the model to sparsely labeled novel category data. These methods mostly generalize well on novel categories from the same domain as the base categories but perform poorly for distant domain catego...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper proposes a framework, terms as ConFeSS, for dealing with cross-domain few-shot learning problems. Specifically, it firstly trains a feature extracting backbone with the contrastive loss on the base category data for learning better features. Then, it trains a masking module to selec...
zRJu6mU2BaE
tHfwJoFGnPr
2,022
ConFeSS: A Framework for Single Source Cross-Domain Few-Shot Learning
Most current few-shot learning methods train a model from abundantly labeled base category data and then transfer and adapt the model to sparsely labeled novel category data. These methods mostly generalize well on novel categories from the same domain as the base categories but perform poorly for distant domain catego...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "In this paper, the authors propose a generic framework for cross-domain few-shot learning. There are three steps: 1. Pre-training the backbone unsupervisedly using a self-supervised contrastive loss. 2. Select relevant features via a mask module. 3. Fine-tune the network with the selected feat...
zRJu6mU2BaE
0fOk7rrmJxG
2,022
ConFeSS: A Framework for Single Source Cross-Domain Few-Shot Learning
Most current few-shot learning methods train a model from abundantly labeled base category data and then transfer and adapt the model to sparsely labeled novel category data. These methods mostly generalize well on novel categories from the same domain as the base categories but perform poorly for distant domain catego...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper proposes a framework for cross-domain few-shot learning. The framework consists of three steps: 1) pre-training backbone; 2) meta-learn the feature masking network; 3) fine-tuning on target domain. Experiments on the CDFSL dataset show this framework outperforms SOTA methods. ", "ma...
zPLQSnfd14w
bAqMulVpkzN
2,022
Two Regimes of Generalization for Non-Linear Metric Learning
A common approach to metric learning is to seek an embedding of the input data that behaves well with respect to the labels. While generalization bounds for linear embeddings are known, the non-linear case is not well understood. In this work we fill this gap by providing uniform generalization guarantees for the cas...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "In this paper, the authors look at the Rademacher complexity of the family of Euclidean metrics learned on a data set via an $L$ layer network where the activations functions are Lipschitz. The idea behind the proof is to use the bounds for $\\epsilon$ net for the embedding network from Barlet...
zPLQSnfd14w
7MwP7ZcuVDI
2,022
Two Regimes of Generalization for Non-Linear Metric Learning
A common approach to metric learning is to seek an embedding of the input data that behaves well with respect to the labels. While generalization bounds for linear embeddings are known, the non-linear case is not well understood. In this work we fill this gap by providing uniform generalization guarantees for the cas...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The authors consider the setting of metric learning. They prove bounds on the Radermacher complexity of embeddings-composed-with-distance-functions for a particular architecture of neural network, with different results for the 'dense' and 'sparse' regimes. \n", "main_review": "**Caveats:**\n-...
zPLQSnfd14w
RMMuZkxCsrB
2,022
Two Regimes of Generalization for Non-Linear Metric Learning
A common approach to metric learning is to seek an embedding of the input data that behaves well with respect to the labels. While generalization bounds for linear embeddings are known, the non-linear case is not well understood. In this work we fill this gap by providing uniform generalization guarantees for the cas...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper tries to provide uniform guarantees for the DNN type metric embeddings.", "main_review": "Strengths: \nIt is indeed a promising research direction to regard neural networks as a special nonlinear metric embedding.\n\nWeakness:\n1.\tThis paper only uses metric embedding to tell a sto...
zPLQSnfd14w
bmXMP79ttn_
2,022
Two Regimes of Generalization for Non-Linear Metric Learning
A common approach to metric learning is to seek an embedding of the input data that behaves well with respect to the labels. While generalization bounds for linear embeddings are known, the non-linear case is not well understood. In this work we fill this gap by providing uniform generalization guarantees for the cas...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper provides two new generalization bounds for non-linear metric learning with deep neural networks, by extending results of Bartlett et al. 2017 to the metric learning setting. The two bounds have been called the 'sparse' and 'non-sparse' bounds and differ in the norm used for the last ...
zNR43c03lRy
g1hhvDiMHVn
2,022
Learning to Annotate Part Segmentation with Gradient Matching
The success of state-of-the-art deep neural networks heavily relies on the presence of large-scale labelled datasets, which are extremely expensive and time-consuming to annotate. This paper focuses on tackling semi-supervised part segmentation tasks by generating high-quality images with a pre-trained GAN and labellin...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "In this paper, the authors mainly propose a gradient-matching-based method for part segmentation to reduce the annotation cost. Based on the DatasetGAN, the proposed model also used the Style GAN family to generate high-quality images and remove the human annotations on a handful of synthesize...
zNR43c03lRy
K1hwLF4A5cS
2,022
Learning to Annotate Part Segmentation with Gradient Matching
The success of state-of-the-art deep neural networks heavily relies on the presence of large-scale labelled datasets, which are extremely expensive and time-consuming to annotate. This paper focuses on tackling semi-supervised part segmentation tasks by generating high-quality images with a pre-trained GAN and labellin...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This work deals with the problem of training a part segmentation network by automatically synthesizing pairs of images and annotations. The authors propose a training method of an annotator given the well-pretrained generator model. The formulation is well-motivated and converged to the gradie...
zNR43c03lRy
HI5dVMRG868
2,022
Learning to Annotate Part Segmentation with Gradient Matching
The success of state-of-the-art deep neural networks heavily relies on the presence of large-scale labelled datasets, which are extremely expensive and time-consuming to annotate. This paper focuses on tackling semi-supervised part segmentation tasks by generating high-quality images with a pre-trained GAN and labellin...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper proposes a method to solve the problem that generally requires large-scale labeled datasets to train deep learning models. The proposed method is designed to solve the nested-loop optimization problem (with an annotator that generates a label and a student network that predicts a la...
zNR43c03lRy
_VXGf3EX4zm
2,022
Learning to Annotate Part Segmentation with Gradient Matching
The success of state-of-the-art deep neural networks heavily relies on the presence of large-scale labelled datasets, which are extremely expensive and time-consuming to annotate. This paper focuses on tackling semi-supervised part segmentation tasks by generating high-quality images with a pre-trained GAN and labellin...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The ultimate goal of the paper is to train an unsupervised generative model, in a semi-supervised manner, as a synthetic training example generator for training segmentation models. Prior works (eg. DatasetGAN) show that one can label a few synthetic images from a pre-trained generative model,...
zNHzqZ9wrRB
Dwa3Jbqmshg
2,022
Equivariant Transformers for Neural Network based Molecular Potentials
The prediction of quantum mechanical properties is historically plagued by a trade-off between accuracy and speed. Machine learning potentials have previously shown great success in this domain, reaching increasingly better accuracy while maintaining computational efficiency comparable with classical force fields. In t...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper proposed equivariant transformers --- a neural network based algorithm to predict properties of molecules. The architecture is built upon the traditional transformer architecture, combined with some modifications specific to molecular property prediction tasks, such as exponential n...
zNHzqZ9wrRB
TjdbneJCYa4
2,022
Equivariant Transformers for Neural Network based Molecular Potentials
The prediction of quantum mechanical properties is historically plagued by a trade-off between accuracy and speed. Machine learning potentials have previously shown great success in this domain, reaching increasingly better accuracy while maintaining computational efficiency comparable with classical force fields. In t...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper describes an equivariant neural network (ET) with attention mechanisms that is applied to the prediction of molecular properties. The ET performs competitive with previous approaches on commonly used benchmark data.", "main_review": "The paper is well structured and the proposed meth...
zNHzqZ9wrRB
UPawlAhAAYq
2,022
Equivariant Transformers for Neural Network based Molecular Potentials
The prediction of quantum mechanical properties is historically plagued by a trade-off between accuracy and speed. Machine learning potentials have previously shown great success in this domain, reaching increasingly better accuracy while maintaining computational efficiency comparable with classical force fields. In t...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The authors introduce a novel architecture for ML force fields, the Equivariant transformer (ET). It is based on the Transformer approach and can be used to predict energies (and forces) and other molecular properties (e.g., QM targets). The performance on standard benchmarks such as QM9 and M...
zNHzqZ9wrRB
18PH99BmXaW
2,022
Equivariant Transformers for Neural Network based Molecular Potentials
The prediction of quantum mechanical properties is historically plagued by a trade-off between accuracy and speed. Machine learning potentials have previously shown great success in this domain, reaching increasingly better accuracy while maintaining computational efficiency comparable with classical force fields. In t...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper presents an equivariant transformer model for predicting quantum mechanical properties from an atomic graph. The model obtains SOTA or near-SOTA results on three popular datasets while maintaining good computational efficiency. The primary novelty in their method is a new way to comp...
zLb9oSWy933
_uymrH5KAs6
2,022
Fast Finite Width Neural Tangent Kernel
The Neural Tangent Kernel (NTK), defined as the outer product of the neural network (NN) Jacobians, $\Theta_\theta(x_1, x_2) = \left[\partial f(\theta, x_1)\big/\partial \theta\right] \left[\partial f(\theta, x_2)\big/\partial \theta\right]^T$, has emerged as a central object of study in deep learning. In the infinite ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This work aims to solve the computation problem of the finite-width neural tangent kernel (NTK), which is a central object in deep learning. The authors analyze the computation and memory requirements for finite-width NTK and propose two novel algorithms that can improve efficiency. Open-sourc...
zLb9oSWy933
WPVoHUX2aO8
2,022
Fast Finite Width Neural Tangent Kernel
The Neural Tangent Kernel (NTK), defined as the outer product of the neural network (NN) Jacobians, $\Theta_\theta(x_1, x_2) = \left[\partial f(\theta, x_1)\big/\partial \theta\right] \left[\partial f(\theta, x_2)\big/\partial \theta\right]^T$, has emerged as a central object of study in deep learning. In the infinite ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "Experiments are convincing. The provided code is helpful for researchers who need a fast computation of NTK. Though the ideological (mathematical part) is very simple.", "main_review": "The paper describes three methods for computing the neural tangent kernel for a given a batch of vectors and...
zLb9oSWy933
FnWeeyLcfkv
2,022
Fast Finite Width Neural Tangent Kernel
The Neural Tangent Kernel (NTK), defined as the outer product of the neural network (NN) Jacobians, $\Theta_\theta(x_1, x_2) = \left[\partial f(\theta, x_1)\big/\partial \theta\right] \left[\partial f(\theta, x_2)\big/\partial \theta\right]^T$, has emerged as a central object of study in deep learning. In the infinite ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper studies the practical compute and memory requirements to computing the Neural Tangent Kernel (NTK), introducing two new approaches to doing so for standard NN primitives (structured derivative and NTK-vector product) which each have advantages (in terms of variables like batch size,...
zLb9oSWy933
BYbuvXI-lDD
2,022
Fast Finite Width Neural Tangent Kernel
The Neural Tangent Kernel (NTK), defined as the outer product of the neural network (NN) Jacobians, $\Theta_\theta(x_1, x_2) = \left[\partial f(\theta, x_1)\big/\partial \theta\right] \left[\partial f(\theta, x_2)\big/\partial \theta\right]^T$, has emerged as a central object of study in deep learning. In the infinite ...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper studies an in-depth analysis of runtime and memory requirements for computing the finite-width NTK. The authors analyze computing costs of Jacobian-vector products (and vice versa) for both fully-connected and convolutional neural networks. They also improve the NTK computation cost...
zKbMQ2NY1y
HElz5vYaGT
2,022
Aug-ILA: More Transferable Intermediate Level Attacks with Augmented References
An intriguing property of deep neural networks is that adversarial attacks can transfer across different models. Existing methods such as the Intermediate Level Attack (ILA) further improve black-box transferability by fine-tuning a reference adversarial attack, so as to maximize the perturbation on a pre-specified lay...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The authors proposed / used three techniques to enhance the transferability of ILA on undefended models. The techniques include image augmentation, reverse adversarial update on the clean example and reference attack update through interpolation with an automatic parameter selection scheme. Th...
zKbMQ2NY1y
YPCMte9btvk
2,022
Aug-ILA: More Transferable Intermediate Level Attacks with Augmented References
An intriguing property of deep neural networks is that adversarial attacks can transfer across different models. Existing methods such as the Intermediate Level Attack (ILA) further improve black-box transferability by fine-tuning a reference adversarial attack, so as to maximize the perturbation on a pre-specified lay...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper presents a transfer-based attack method based on Intermediate Level Attack(ILA). Image augementation and reverse adversarial updateare applied to ILA input, to make diverse adversarial references. Moreover, the interpolation of cumlative attacks can maintain a better transfer direct...
zKbMQ2NY1y
QZjkSfSxk81
2,022
Aug-ILA: More Transferable Intermediate Level Attacks with Augmented References
An intriguing property of deep neural networks is that adversarial attacks can transfer across different models. Existing methods such as the Intermediate Level Attack (ILA) further improve black-box transferability by fine-tuning a reference adversarial attack, so as to maximize the perturbation on a pre-specified lay...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "This paper introduces improvements over Transferable Intermediate-level Attacks (ILA), by incorporating data augmentation into the attack tuning stage. The data augmentations introduced include 3 kinds: simple transformations (cropping), reverse-adversarial update, and attack interpolation. An...
zKbMQ2NY1y
x5diiyntoan
2,022
Aug-ILA: More Transferable Intermediate Level Attacks with Augmented References
An intriguing property of deep neural networks is that adversarial attacks can transfer across different models. Existing methods such as the Intermediate Level Attack (ILA) further improve black-box transferability by fine-tuning a reference adversarial attack, so as to maximize the perturbation on a pre-specified lay...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The paper proposed the augmented Intermediate Level Attack (ILA) algorithm to strengthen the transferability of adversarial examples. Also, it claimed that increasing the diversity of input references could improve the generalization of adversarial examples when attacking different models. Spe...
zIUyj55nXR
fdClrbLgAga
2,022
Frame Averaging for Invariant and Equivariant Network Design
Many machine learning tasks involve learning functions that are known to be invariant or equivariant to certain symmetries of the input data. However, it is often challenging to design neural network architectures that respect these symmetries while being expressive and computationally efficient. For example, Euclidean...
You are an expert academic peer reviewer for ICLR 2022. Read the paper provided in the next message and write a complete review as a JSON object with the following fields: - "summary_of_the_paper" (string): Provide a brief summary of the paper and its contributions. - "main_review" (string): Please list both the stren...
{paper_markdown}
{"summary_of_the_paper": "The authors introduce Frame Averaging (FA), a general framework for adapting known architectures to become invariant or equivariant with respect to a general group by using group averaging operator. The idea of FA is to replace the averaging operator over the entire group by the averaging over...