ICLR 2021 Schedule

Filter Events

MON 3 MAY

midnight

Remarks:

Opening Remarks

(ends 1:00 AM)

1 a.m.

Poster Session 01 [1:00-3:00]

Mutual Information State Intrinsic Control

Does enhanced shape bias improve neural network robustness to common corruptions?

Rethinking the Role of Gradient-based Attribution Methods for Model Interpretability

Progressive Skeletonization: Trimming more fat from a network at initialization

Towards Robust Neural Networks via Close-loop Control

On the Universality of the Double Descent Peak in Ridgeless Regression

Implicit Normalizing Flows

Gauge Equivariant Mesh CNNs: Anisotropic convolutions on geometric graphs

Rapid Neural Architecture Search by Learning to Generate Graphs from Datasets

A Good Image Generator Is What You Need for High-Resolution Video Synthesis

QPLEX: Duplex Dueling Multi-Agent Q-Learning

Wasserstein-2 Generative Networks

Neural Jump Ordinary Differential Equations: Consistent Continuous-Time Prediction and Filtering

Randomized Ensembled Double Q-Learning: Learning Fast Without a Model

SaliencyMix: A Saliency Guided Data Augmentation Strategy for Better Regularization

WaNet - Imperceptible Warping-based Backdoor Attack

Domain Generalization with MixStyle

SALD: Sign Agnostic Learning with Derivatives

PSTNet: Point Spatio-Temporal Convolution on Point Cloud Sequences

Isometric Transformation Invariant and Equivariant Graph Convolutional Networks

LEAF: A Learnable Frontend for Audio Classification

Wasserstein Embedding for Graph Learning

Learning N:M Fine-grained Structured Sparse Neural Networks From Scratch

Interpreting and Boosting Dropout from a Game-Theoretic View

The Unreasonable Effectiveness of Patches in Deep Convolutional Kernels Methods

Overfitting for Fun and Profit: Instance-Adaptive Data Compression

Trusted Multi-View Classification

Batch Reinforcement Learning Through Continuation Method

Solving Compositional Reinforcement Learning Problems via Task Reduction

A Unified Approach to Interpreting and Boosting Adversarial Transferability

Improve Object Detection with Feature-based Knowledge Distillation: Towards Accurate and Efficient Detectors

Parameter-Based Value Functions

Towards Impartial Multi-task Learning

Set Prediction without Imposing Structure as Conditional Density Estimation

Deciphering and Optimizing Multi-Task Learning: a Random Matrix Approach

Targeted Attack against Deep Neural Networks via Flipping Limited Weight Bits

Towards Robustness Against Natural Language Word Substitutions

Scalable Transfer Learning with Expert Models

On InstaHide, Phase Retrieval, and Sparse Matrix Factorization

What Makes Instance Discrimination Good for Transfer Learning?

MODALS: Modality-agnostic Automated Data Augmentation in the Latent Space

MELR: Meta-Learning via Modeling Episode-Level Relationships for Few-Shot Learning

Training with Quantization Noise for Extreme Model Compression

Spatially Structured Recurrent Modules

Tomographic Auto-Encoder: Unsupervised Bayesian Recovery of Corrupted Data

Uncertainty Estimation and Calibration with Finite-State Probabilistic RNNs

Predicting Infectiousness for Proactive Contact Tracing

Signatory: differentiable computations of the signature and logsignature transforms, on both CPU and GPU

Temporally-Extended ε-Greedy Exploration

Semantic Re-tuning with Contrastive Tension

Is Label Smoothing Truly Incompatible with Knowledge Distillation: An Empirical Study

ResNet After All: Neural ODEs and Their Numerical Solution

Noise against noise: stochastic label noise helps combat inherent label noise

FairFil: Contrastive Neural Debiasing Method for Pretrained Text Encoders

Neural Approximate Sufficient Statistics for Implicit Models

Meta-GMVAE: Mixture of Gaussian VAE for Unsupervised Meta-Learning

On the Transfer of Disentangled Representations in Realistic Settings

Exploring Balanced Feature Spaces for Representation Learning

MetaNorm: Learning to Normalize Few-Shot Batches Across Domains

Stabilized Medical Image Attacks

Revisiting Locally Supervised Learning: an Alternative to End-to-end Training

On Learning Universal Representations Across Languages

(ends 3:00 AM)

3 a.m.

Oral Session 1 [3:00-6:05]

Ozan Sener, Yutian Chen, Blake Richards

Oral s 3:00-3:30

[3:00] Dataset Condensation with Gradient Matching

[3:15] Free Lunch for Few-shot Learning: Distribution Calibration

Spotlight s 3:30-3:50

[3:30] Deciphering and Optimizing Multi-Task Learning: a Random Matrix Approach

[3:40] Generalization in data-driven models of primary visual cortex

Q&A s 3:50-4:00

[3:50] Q&A

Oral s 4:00-4:30

[4:00] Towards Nonlinear Disentanglement in Natural Data with Temporal Sparse Coding

[4:15] A Distributional Approach to Controlled Text Generation

Spotlight s 4:30-4:50

[4:30] The Intrinsic Dimension of Images and Its Impact on Learning

[4:40] How Benign is Benign Overfitting ?

Q&A s 4:50-5:00

[4:50] Q&A

Oral s 5:00-5:45

[5:00] Geometry-aware Instance-reweighted Adversarial Training

[5:15] Do 2D GANs Know 3D Shape? Unsupervised 3D Shape Reconstruction from 2D Image GANs

[5:30] Rethinking the Role of Gradient-based Attribution Methods for Model Interpretability

Spotlight s 5:45-5:55

[5:45] Contrastive Divergence Learning is a Time Reversal Adversarial Game

Q&A s 5:55-6:05

[5:55] Q&A

(ends 6:05 AM)

6:15 a.m.

BREAK:

Please visit the Sponsor Hall, the Socials, and Mentorships.

8 a.m.

Invited Talk:

Moving beyond the fairness rhetoric in machine learning

Timnit Gebru

(ends 9:00 AM)

9 a.m.

Poster Session 02 [9:00-11:00]

Revisiting Few-sample BERT Fine-tuning

Universal approximation power of deep residual neural networks via nonlinear control theory

Plan-Based Relaxed Reward Shaping for Goal-Directed Tasks

The Traveling Observer Model: Multi-task Learning Through Spatial Variable Embeddings

Language-Agnostic Representation Learning of Source Code from Structure and Context

Training GANs with Stronger Augmentations via Contrastive Discriminator

Overparameterisation and worst-case generalisation: friend or foe?

What Can You Learn From Your Muscles? Learning Visual Representation from Human Interactions

Into the Wild with AudioScope: Unsupervised Audio-Visual Separation of On-Screen Sounds

Self-Supervised Policy Adaptation during Deployment

Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixels

On the Stability of Fine-tuning BERT: Misconceptions, Explanations, and Strong Baselines

Multi-Time Attention Networks for Irregularly Sampled Time Series

Scaling Symbolic Methods using Gradients for Neural Model Explanation

Shapley Explanation Networks

Shape-Texture Debiased Neural Network Training

PAC Confidence Predictions for Deep Neural Network Classifiers

MoVie: Revisiting Modulated Convolutions for Visual Counting and Beyond

The role of Disentanglement in Generalisation

Single-Photon Image Classification

Share or Not? Learning to Schedule Language-Specific Capacity for Multilingual Translation

Learning Structural Edits via Incremental Tree Transformations

The Risks of Invariant Risk Minimization

Teaching Temporal Logics to Neural Networks

Saliency is a Possible Red Herring When Diagnosing Poor Generalization

Adaptive Federated Optimization

On the role of planning in model-based deep reinforcement learning

Extracting Strong Policies for Robotics Tasks from Zero-Order Trajectory Optimizers

Predicting Classification Accuracy When Adding New Unobserved Classes

Disentangling 3D Prototypical Networks for Few-Shot Concept Learning

Sparse encoding for more-interpretable feature-selecting representations in probabilistic matrix factorization

Unsupervised Meta-Learning through Latent-Space Interpolation in Generative Models

Planning from Pixels using Inverse Dynamics Models

LiftPool: Bidirectional ConvNet Pooling

Rapid Task-Solving in Novel Environments

Learning "What-if" Explanations for Sequential Decision-Making

Graph Convolution with Low-rank Learnable Local Filters

Grounding Physical Concepts of Objects and Events Through Dynamic Visual Reasoning

Vector-output ReLU Neural Network Problems are Copositive Programs: Convex Analysis of Two Layer Networks and Polynomial-time Algorithms

LambdaNetworks: Modeling long-range Interactions without Attention

Symmetry-Aware Actor-Critic for 3D Molecular Design

Selectivity considered harmful: evaluating the causal impact of class selectivity in DNNs

Learning explanations that are hard to vary

Seq2Tens: An Efficient Representation of Sequences by Low-Rank Tensor Projections

Structured Prediction as Translation between Augmented Natural Languages

Learning Hyperbolic Representations of Topological Features

Understanding the failure modes of out-of-distribution generalization

Representation learning for improved interpretability and classification accuracy of clinical factors from EEG

Learning-based Support Estimation in Sublinear Time

Gradient Projection Memory for Continual Learning

A statistical theory of cold posteriors in deep neural networks

MultiModalQA: complex question answering over text, tables and images

Multi-Level Local SGD: Distributed SGD for Heterogeneous Hierarchical Networks

On the Universality of Rotation Equivariant Point Cloud Networks

On Statistical Bias In Active Learning: How and When to Fix It

On the Impossibility of Global Convergence in Multi-Loss Optimization

Learning from others' mistakes: Avoiding dataset biases without modeling them

Predicting Inductive Biases of Pre-Trained Models

Accelerating Convergence of Replica Exchange Stochastic Gradient MCMC via Variance Reduction

Effective Distributed Learning with Random Features: Improved Bounds and Algorithms

Parameter Efficient Multimodal Transformers for Video Representation Learning

Trajectory Prediction using Equivariant Continuous Convolution

Learning with AMIGo: Adversarially Motivated Intrinsic Goals

Shapley explainability on the data manifold

Open Question Answering over Tables and Text

Zero-Cost Proxies for Lightweight NAS

GraPPa: Grammar-Augmented Pre-Training for Table Semantic Parsing

Intrinsic-Extrinsic Convolution and Pooling for Learning on 3D Protein Structures

What Should Not Be Contrastive in Contrastive Learning

Conditional Negative Sampling for Contrastive Learning of Visual Representations

Fast convergence of stochastic subgradient method under interpolation

Pruning Neural Networks at Initialization: Why Are We Missing the Mark?

NeMo: Neural Mesh Models of Contrastive Features for Robust 3D Pose Estimation

X2T: Training an X-to-Text Typing Interface with Online Learning from User Feedback

Learning Invariant Representations for Reinforcement Learning without Reconstruction

InfoBERT: Improving Robustness of Language Models from An Information Theoretic Perspective

Using latent space regression to analyze and leverage compositionality in GANs

Towards Nonlinear Disentanglement in Natural Data with Temporal Sparse Coding

Rethinking Embedding Coupling in Pre-trained Language Models

Uncertainty Sets for Image Classifiers using Conformal Prediction

Reset-Free Lifelong Learning with Skill-Space Planning

Primal Wasserstein Imitation Learning

WrapNet: Neural Net Inference with Ultra-Low-Precision Arithmetic

Off-Dynamics Reinforcement Learning: Training for Transfer with Domain Classifiers

(ends 11:00 AM)

11 a.m.

Oral Session 2 [11:00-2:23]

Oral s 11:00-11:45

[11:00] Federated Learning Based on Dynamic Regularization

[11:15] Gradient Projection Memory for Continual Learning

[11:30] Growing Efficient Deep Networks by Structured Continuous Sparsification

Spotlight s 11:45-11:55

[11:45] Geometry-Aware Gradient Algorithms for Neural Architecture Search

Q&A s 11:55-12:05

[11:55] Q&A

Spotlight s 12:05-1:05

[12:05] Generalization bounds via distillation

[12:15] On the Theory of Implicit Deep Learning: Global Convergence with Implicit Layers

[12:25] Sharpness-aware Minimization for Efficiently Improving Generalization

[12:35] Systematic generalisation with group invariant predictions

[12:45] On Statistical Bias In Active Learning: How and When to Fix It

[12:55] Very Deep VAEs Generalize Autoregressive Models and Can Outperform Them on Images

Q&A s 1:05-1:20

[1:05] Q&A

Spotlight s 1:20-2:10

[1:20] Uncertainty Sets for Image Classifiers using Conformal Prediction

[1:30] PMI-Masking: Principled masking of correlated spans

[1:40] Gradient Vaccine: Investigating and Improving Multi-task Optimization in Massively Multilingual Models

[1:50] Watch-And-Help: A Challenge for Social Perception and Human-AI Collaboration

[2:00] Predicting Infectiousness for Proactive Contact Tracing

Q&A s 2:10-2:23

[2:10] Q&A

(ends 2:23 PM)

2:30 p.m.

BREAK:

Please visit the Sponsor Hall, the Socials, and Mentorships.

4 p.m.

Invited Talk:

Commonsense AI: Myth and Truth

Yejin Choi

(ends 5:00 PM)

5 p.m.

Poster Session 03 [5:00-7:00]

Random Feature Attention

Zero-shot Synthesis with Group-Supervised Learning

Remembering for the Right Reasons: Explanations Reduce Catastrophic Forgetting

When Optimizing $f$-Divergence is Robust with Label Noise

The Intrinsic Dimension of Images and Its Impact on Learning

Regularized Inverse Reinforcement Learning

Robust Curriculum Learning: from clean label detection to noisy label self-correction

Benefit of deep learning with non-convex noisy gradient descent: Provable excess risk bound and superiority to kernel methods

Deberta: Decoding-Enhanced Bert With Disentangled Attention

Sequential Density Ratio Estimation for Simultaneous Optimization of Speed and Accuracy

Representation Learning for Sequence Data with Deep Autoencoding Predictive Components

Latent Skill Planning for Exploration and Transfer

SOLAR: Sparse Orthogonal Learned and Random Embeddings

When does preconditioning help or hurt generalization?

Learning Energy-Based Models by Diffusion Recovery Likelihood

MONGOOSE: A Learnable LSH Framework for Efficient Neural Network Training

Decentralized Attribution of Generative Models

Partitioned Learned Bloom Filters

SAFENet: A Secure, Accurate and Fast Neural Network Inference

Deep Partition Aggregation: Provable Defenses against General Poisoning Attacks

MALI: A memory efficient and reverse accurate integrator for Neural ODEs

Very Deep VAEs Generalize Autoregressive Models and Can Outperform Them on Images

The Role of Momentum Parameters in the Optimal Convergence of Adaptive Polyak's Heavy-ball Methods

PlasticineLab: A Soft-Body Manipulation Benchmark with Differentiable Physics

Semi-supervised Keypoint Localization

Federated Learning Based on Dynamic Regularization

On the geometry of generalization and memorization in deep neural networks

Unlearnable Examples: Making Personal Data Unexploitable

Proximal Gradient Descent-Ascent: Variable Convergence under KŁ Geometry

UPDeT: Universal Multi-agent RL via Policy Decoupling with Transformers

On Fast Adversarial Robustness Adaptation in Model-Agnostic Meta-Learning

DeLighT: Deep and Light-weight Transformer

Undistillable: Making A Nasty Teacher That CANNOT teach students

Tilted Empirical Risk Minimization

Rethinking Architecture Selection in Differentiable NAS

What are the Statistical Limits of Offline RL with Linear Function Approximation?

Bypassing the Ambient Dimension: Private SGD with Gradient Subspace Identification

Spatio-Temporal Graph Scattering Transform

MoPro: Webly Supervised Learning with Momentum Prototypes

VA-RED$^2$: Video Adaptive Redundancy Reduction

Score-Based Generative Modeling through Stochastic Differential Equations

SCoRe: Pre-Training for Context Representation in Conversational Semantic Parsing

Robust Reinforcement Learning on State Observations with Learned Optimal Adversary

Neural Attention Distillation: Erasing Backdoor Triggers from Deep Neural Networks

On Dyadic Fairness: Exploring and Mitigating Bias in Graph Connections

WaveGrad: Estimating Gradients for Waveform Generation

Online Adversarial Purification based on Self-supervised Learning

Meta-Learning with Neural Tangent Kernels

Learning a Latent Simplex in Input Sparsity Time

Taking Notes on the Fly Helps Language Pre-Training

Parrot: Data-Driven Behavioral Priors for Reinforcement Learning

Learning A Minimax Optimizer: A Pilot Study

The Deep Bootstrap Framework: Good Online Learners are Good Offline Generalizers

MixKD: Towards Efficient Distillation of Large-scale Language Models

Incorporating Symmetry into Deep Dynamics Models for Improved Generalization

Optimal Regularization can Mitigate Double Descent

Explaining the Efficacy of Counterfactually Augmented Data

Selective Classification Can Magnify Disparities Across Groups

Why resampling outperforms reweighting for correcting sampling bias with stochastic gradients

Contextual Transformation Networks for Online Continual Learning

Layer-adaptive Sparsity for the Magnitude-based Pruning

Long Live the Lottery: The Existence of Winning Tickets in Lifelong Learning

Regularization Matters in Policy Optimization - An Empirical Study on Continuous Control

PseudoSeg: Designing Pseudo Labels for Semantic Segmentation

One Network Fits All? Modular versus Monolithic Task Formulations in Neural Networks

Improved Estimation of Concentration Under $\ell_p$-Norm Distance Metrics Using Half Spaces

Offline Model-Based Optimization via Normalized Maximum Likelihood Estimation

SenSeI: Sensitive Set Invariance for Enforcing Individual Fairness

Optimizing Memory Placement using Evolutionary Graph Reinforcement Learning

Personalized Federated Learning with First Order Model Optimization

Rethinking Positional Encoding in Language Pre-training

Variational Intrinsic Control Revisited

Self-training For Few-shot Transfer Across Extreme Task Differences

Model Patching: Closing the Subgroup Performance Gap with Data Augmentation

Robust and Generalizable Visual Representation Learning via Random Convolutions

(ends 7:00 PM)

7 p.m.

Oral Session 3 [7:00-10:16]

Oral s 7:00-7:45

[7:00] SMiRL: Surprise Minimizing Reinforcement Learning in Unstable Environments

[7:15] Contrastive Explanations for Reinforcement Learning via Embedded Self Predictions

[7:30] Parrot: Data-Driven Behavioral Priors for Reinforcement Learning

Spotlight s 7:45-8:05

[7:45] Structured Prediction as Translation between Augmented Natural Languages

[7:55] Mathematical Reasoning via Self-supervised Skip-tree Training

Q&A s 8:05-8:18

[8:05] Q&A

Spotlight s 8:18-9:08

[8:18] Improving Adversarial Robustness via Channel-wise Activation Suppressing

[8:28] Fast Geometric Projections for Local Robustness Certification

[8:38] Information Laundering for Model Privacy

[8:48] Dataset Inference: Ownership Resolution in Machine Learning

[8:58] HW-NAS-Bench: Hardware-Aware Neural Architecture Search Benchmark

Q&A s 9:08-9:21

[9:08] Q&A

Oral s 9:21-9:36

[9:21] How Neural Networks Extrapolate: From Feedforward to Graph Neural Networks

Spotlight s 9:36-10:06

[9:36] Graph Convolution with Low-rank Learnable Local Filters

[9:46] The Traveling Observer Model: Multi-task Learning Through Spatial Variable Embeddings

[9:56] Meta-GMVAE: Mixture of Gaussian VAE for Unsupervised Meta-Learning

Q&A s 10:06-10:16

[10:06] Q&A

(ends 10:16 PM)

10:30 p.m.

BREAK:

Please visit the Sponsor Hall, the Socials, and Mentorships.

TUE 4 MAY

midnight

Invited Talk:

Geometric Deep Learning: the Erlangen Programme of ML

Michael Bronstein

(ends 1:00 AM)

1 a.m.

Poster Session 04 [1:00-3:00]

Scaling the Convex Barrier with Active Sets

Complex Query Answering with Neural Link Predictors

Contemplating Real-World Object Classification

Do 2D GANs Know 3D Shape? Unsupervised 3D Shape Reconstruction from 2D Image GANs

Lossless Compression of Structured Convolutional Models via Lifting

MiCE: Mixture of Contrastive Experts for Unsupervised Image Clustering

Capturing Label Characteristics in VAEs

Disambiguating Symbolic Expressions in Informal Documents

SkipW: Resource Adaptable RNN with Strict Upper Computational Limit

Sample-Efficient Automated Deep Reinforcement Learning

Identifying nonlinear dynamical systems with multiple time scales and long-range dependencies

Probing BERT in Hyperbolic Spaces

Efficient Certified Defenses Against Patch Attacks on Image Classifiers

A Distributional Approach to Controlled Text Generation

Class Normalization for (Continual)? Generalized Zero-Shot Learning

Exemplary Natural Images Explain CNN Activations Better than State-of-the-Art Feature Visualization

PDE-Driven Spatiotemporal Disentanglement

Group Equivariant Stand-Alone Self-Attention For Vision

Bayesian Context Aggregation for Neural Processes

Generalization in data-driven models of primary visual cortex

Effective Abstract Reasoning with Dual-Contrast Network

Generalized Multimodal ELBO

Large-width functional asymptotics for deep Gaussian neural networks

Policy-Driven Attack: Learning to Query for Hard-label Black-box Adversarial Examples

Auxiliary Learning by Implicit Differentiation

Calibration tests beyond classification

Prototypical Contrastive Learning of Unsupervised Representations

Computational Separation Between Convolutional and Fully-Connected Networks

Rao-Blackwellizing the Straight-Through Gumbel-Softmax Gradient Estimator

Coping with Label Shift via Distributionally Robust Optimisation

Generalized Energy Based Models

Learning Incompressible Fluid Dynamics from Scratch - Towards Fast, Differentiable Fluid Models that Generalize

Activation-level uncertainty in deep neural networks

Accurate Learning of Graph Representations with Graph Multiset Pooling

GraphCodeBERT: Pre-training Code Representations with Data Flow

Image GANs meet Differentiable Rendering for Inverse Graphics and Interpretable 3D Neural Rendering

Spatial Dependency Networks: Neural Layers for Improved Generative Image Modeling

Learning Subgoal Representations with Slow Dynamics

Learning Accurate Entropy Model with Global Reference for Image Compression

Model-based micro-data reinforcement learning: what are the crucial model properties and which model to choose?

Group Equivariant Conditional Neural Processes

Multiscale Score Matching for Out-of-Distribution Detection

Winning the L2RPN Challenge: Power Grid Management via Semi-Markov Afterstate Actor-Critic

Intraclass clustering: an implicit learning ability that regularizes DNNs

Practical Real Time Recurrent Learning with a Sparse Approximation

Distributed Momentum for Byzantine-resilient Stochastic Gradient Descent

Refining Deep Generative Models via Discriminator Gradient Flow

Hyperbolic Neural Networks++

IDF++: Analyzing and Improving Integer Discrete Flows for Lossless Compression

A Universal Representation Transformer Layer for Few-Shot Image Classification

Improving Transformation Invariance in Contrastive Representation Learning

Learning the Pareto Front with Hypernetworks

FedMix: Approximation of Mixup under Mean Augmented Federated Learning

A Trainable Optimal Transport Embedding for Feature Aggregation and its Relationship to Attention

Prediction and generalisation over directed actions by grid cells

Conformation-Guided Molecular Representation with Hamiltonian Neural Networks

not-MIWAE: Deep Generative Modelling with Missing not at Random Data

BOIL: Towards Representation Change for Few-shot Learning

Risk-Averse Offline Reinforcement Learning

A Block Minifloat Representation for Training Deep Neural Networks

Learning Better Structured Representations Using Low-rank Adaptive Label Smoothing

What they do when in doubt: a study of inductive biases in seq2seq learners

Neurally Augmented ALISTA

On Self-Supervised Image Representations for GAN Evaluation

AdaGCN: Adaboosting Graph Convolutional Networks into Deep Models

Deep Repulsive Clustering of Ordered Data Based on Order-Identity Decomposition

Monte-Carlo Planning and Learning with Language Action Value Estimates

(ends 3:00 AM)

3 a.m.

Oral Session 4 [3:00-5:58]

Oral s 3:00-3:15

[3:00] End-to-end Adversarial Text-to-Speech

Spotlight s 3:15-3:55

[3:15] Autoregressive Entity Retrieval

[3:25] Interpreting Graph Neural Networks for NLP With Differentiable Edge Masking

[3:35] Expressive Power of Invariant and Equivariant Graph Neural Networks

[3:45] Gauge Equivariant Mesh CNNs: Anisotropic convolutions on geometric graphs

Q&A s 3:55-4:08

[3:55] Q&A

Oral s 4:08-4:38

[4:08] Rao-Blackwellizing the Straight-Through Gumbel-Softmax Gradient Estimator

[4:23] Scalable Learning and MAP Inference for Nonsymmetric Determinantal Point Processes

Spotlight s 4:38-4:58

[4:38] Multivariate Probabilistic Time Series Forecasting via Conditioned Normalizing Flows

[4:48] Noise against noise: stochastic label noise helps combat inherent label noise

Q&A s 4:58-5:08

[4:58] Q&A

Spotlight s 5:08-5:48

[5:08] Mutual Information State Intrinsic Control

[5:18] Learning Incompressible Fluid Dynamics from Scratch - Towards Fast, Differentiable Fluid Models that Generalize

[5:28] Identifying nonlinear dynamical systems with multiple time scales and long-range dependencies

[5:38] Fidelity-based Deep Adiabatic Scheduling

Q&A s 5:48-5:58

[5:48] Q&A

(ends 5:58 AM)

6 a.m.

BREAK:

Please visit the Sponsor Hall, the Socials, and Mentorships.

8 a.m.

Invited Talk:

AI in Finance: Scope and Examples

Manuela Veloso

(ends 9:00 AM)

9 a.m.

Poster Session 05 [9:00-11:00]

Tent: Fully Test-Time Adaptation by Entropy Minimization

Text Generation by Learning from Demonstrations

VAEBM: A Symbiosis between Variational Autoencoders and Energy-based Models

Rank the Episodes: A Simple Approach for Exploration in Procedurally-Generated Environments

DC3: A learning method for optimization with hard constraints

Towards Resolving the Implicit Bias of Gradient Descent for Matrix Factorization: Greedy Low-Rank Learning

Robust Pruning at Initialization

Unsupervised Object Keypoint Learning using Local Spatial Predictability

Direction Matters: On the Implicit Bias of Stochastic Gradient Descent with Moderate Learning Rate

The geometry of integration in text classification RNNs

Recurrent Independent Mechanisms

Large Associative Memory Problem in Neurobiology and Machine Learning

Reducing the Computational Cost of Deep Generative Models with Binary Neural Networks

Representation Learning via Invariant Causal Mechanisms

DINO: A Conditional Energy-Based GAN for Domain Translation

Are wider nets better given the same number of parameters?

Sharper Generalization Bounds for Learning with Gradient-dominated Objective Functions

Shape or Texture: Understanding Discriminative Features in CNNs

Systematic generalisation with group invariant predictions

Quantifying Differences in Reward Functions

Understanding Over-parameterization in Generative Adversarial Networks

Reinforcement Learning with Random Delays

Training BatchNorm and Only BatchNorm: On the Expressive Power of Random Features in CNNs

Iterative Empirical Game Solving via Single Policy Best Response

Learning a Latent Search Space for Routing Problems using Variational Autoencoders

Taming GANs with Lookahead-Minmax

Rethinking Attention with Performers

Tradeoffs in Data Augmentation: An Empirical Study

Physics-aware, probabilistic model order reduction with guaranteed stability

Characterizing signal propagation to close the performance gap in unnormalized ResNets

NOVAS: Non-convex Optimization via Adaptive Stochastic Search for End-to-end Learning and Control

Self-Supervised Learning of Compressed Video Representations

Ringing ReLUs: Harmonic Distortion Analysis of Nonlinear Feedforward Networks

How Benign is Benign Overfitting ?

VTNet: Visual Transformer Network for Object Goal Navigation

Mapping the Timescale Organization of Neural Language Models

Support-set bottlenecks for video-text representation learning

Clairvoyance: A Pipeline Toolkit for Medical Time Series

Uncertainty Estimation in Autoregressive Structured Prediction

Distance-Based Regularisation of Deep Networks for Fine-Tuning

Learning from Protein Structure with Geometric Vector Perceptrons

Data-Efficient Reinforcement Learning with Self-Predictive Representations

Nearest Neighbor Machine Translation

Transformer protein language models are unsupervised structure learners

Provable Rich Observation Reinforcement Learning with Combinatorial Latent States

SSD: A Unified Framework for Self-Supervised Outlier Detection

Discovering a set of policies for the worst case reward

Getting a CLUE: A Method for Explaining Uncertainty Estimates

DialoGraph: Incorporating Interpretable Strategy-Graph Networks into Negotiation Dialogues

Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrieval

Uncertainty-aware Active Learning for Optimal Bayesian Classifier

Learning Parametrised Graph Shift Operators

UMEC: Unified model and embedding compression for efficient recommendation systems

Vulnerability-Aware Poisoning Mechanism for Online RL with Unknown Dynamics

Scalable Bayesian Inverse Reinforcement Learning

SMiRL: Surprise Minimizing Reinforcement Learning in Unstable Environments

Teaching with Commentaries

On the Theory of Implicit Deep Learning: Global Convergence with Implicit Layers

Transient Non-stationarity and Generalisation in Deep Reinforcement Learning

Auction Learning as a Two-Player Game

On the Dynamics of Training Attention Models

Single-Timescale Actor-Critic Provably Finds Globally Optimal Policy

Fair Mixup: Fairness via Interpolation

FairBatch: Batch Selection for Model Fairness

Towards Faster and Stabilized GAN Training for High-fidelity Few-shot Image Synthesis

C-Learning: Horizon-Aware Cumulative Accessibility Estimation

On the Origin of Implicit Regularization in Stochastic Gradient Descent

Meta-learning Symmetries by Reparameterization

Learning Value Functions in Deep Policy Gradients using Residual Variance

Global optimality of softmax policy gradient with single hidden layer neural networks in the mean-field regime

Supervised Contrastive Learning for Pre-trained Language Model Fine-tuning

Anatomy of Catastrophic Forgetting: Hidden Representations and Task Semantics

Statistical inference for individual fairness

Learning Robust State Abstractions for Hidden-Parameter Block MDPs

On the mapping between Hopfield networks and Restricted Boltzmann Machines

Decoupling Global and Local Representations via Invertible Generative Flows

Unsupervised Representation Learning for Time Series with Temporal Neighborhood Coding

Learning Neural Event Functions for Ordinary Differential Equations

Interpreting Knowledge Graph Relation Representation from Word Embeddings

(ends 11:00 AM)

11 a.m.

Oral Session 5 [11:00-1:56]

Oral s 11:00-11:30

[11:00] Iterated learning for emergent systematicity in VQA

[11:15] Learning Generalizable Visual Representations via Interactive Gameplay

Spotlight s 11:30-11:50

[11:30] How Does Mixup Help With Robustness and Generalization?

[11:40] Recurrent Independent Mechanisms

Q&A s 11:50-12:00

[11:50] Q&A

Oral s 12:00-12:30

[12:00] Randomized Automatic Differentiation

[12:15] Image GANs meet Differentiable Rendering for Inverse Graphics and Interpretable 3D Neural Rendering

Spotlight s 12:30-1:00

[12:30] Mind the Pad -- CNNs Can Develop Blind Spots

[12:40] Implicit Convex Regularizers of CNN Architectures: Convex Optimization of Two- and Three-Layer Networks in Polynomial Time

[12:50] Learning from Protein Structure with Geometric Vector Perceptrons

Q&A s 1:00-1:13

[1:00] Q&A

Oral s 1:13-1:28

[1:13] On the mapping between Hopfield networks and Restricted Boltzmann Machines

Spotlight s 1:28-1:48

[1:28] Learning-based Support Estimation in Sublinear Time

[1:38] Long-tail learning via logit adjustment

Q&A s 1:48-1:56

[1:48] Q&A

(ends 1:56 PM)

2 p.m.

Expo Talk Panel:

Interpretability with skeptical and user-centric mind

(ends 3:00 PM)

3 p.m.

BREAK:

Please visit the Sponsor Hall, the Socials, and Mentorships.

4 p.m.

Town Hall:

Town Hall

(ends 5:00 PM)

5 p.m.

Poster Session 06 [5:00-7:00]

Lipschitz Recurrent Neural Networks

A Discriminative Gaussian Mixture Model with Sparsity

Diverse Video Generation using a Gaussian Process Trigger

Mirostat: A Neural Text Decoding Algorithm That Directly Controls Perplexity

Drop-Bottleneck: Learning Discrete Compressed Representation for Noise-Robust Exploration

Fuzzy Tiling Activations: A Simple Approach to Learning Sparse Representations Online

Usable Information and Evolution of Optimal Representations During Training

Topology-Aware Segmentation Using Discrete Morse Theory

SEDONA: Search for Decoupled Neural Networks toward Greedy Block-wise Learning

The Importance of Pessimism in Fixed-Dataset Policy Optimization

Hierarchical Autoregressive Modeling for Neural Video Compression

Debiasing Concept-based Explanations with Causal Analysis

Fourier Neural Operator for Parametric Partial Differential Equations

Autoregressive Dynamics Models for Offline Policy Evaluation and Optimization

Can a Fruit Fly Learn Word Embeddings?

Implicit Convex Regularizers of CNN Architectures: Convex Optimization of Two- and Three-Layer Networks in Polynomial Time

Deep Equals Shallow for ReLU Networks in Kernel Regimes

DrNAS: Dirichlet Neural Architecture Search

Universal Weakly Supervised Segmentation by Pixel-to-Segment Contrastive Learning

On Graph Neural Networks versus Graph-Augmented MLPs

DOP: Off-Policy Multi-Agent Decomposed Policy Gradients

A unifying view on implicit bias in training linear neural networks

Monotonic Kronecker-Factored Lattice

Attentional Constellation Nets for Few-Shot Learning

CcGAN: Continuous Conditional Generative Adversarial Networks for Image Generation

Mathematical Reasoning via Self-supervised Skip-tree Training

Hopper: Multi-hop Transformer for Spatiotemporal Reasoning

CompOFA – Compound Once-For-All Networks for Faster Multi-Platform Deployment

CoDA: Contrast-enhanced and Diversity-promoting Data Augmentation for Natural Language Understanding

How to Find Your Friendly Neighborhood: Graph Attention Design with Self-Supervision

Dataset Inference: Ownership Resolution in Machine Learning

Concept Learners for Few-Shot Learning

Generating Adversarial Computer Programs using Optimized Obfuscations

Knowledge Distillation as Semiparametric Inference

Fantastic Four: Differentiable and Efficient Bounds on Singular Values of Convolution Layers

BUSTLE: Bottom-Up Program Synthesis Through Learning-Guided Exploration

Federated Semi-Supervised Learning with Inter-Client Consistency & Disjoint Learning

DDPNOpt: Differential Dynamic Programming Neural Optimizer

RNNLogic: Learning Logic Rules for Reasoning on Knowledge Graphs

A Temporal Kernel Approach for Deep Learning with Continuous-time Information

Denoising Diffusion Implicit Models

Individually Fair Rankings

Memory Optimization for Deep Networks

RMSprop converges with proper hyper-parameter

Achieving Linear Speedup with Partial Worker Participation in Non-IID Federated Learning

Why Are Convolutional Nets More Sample-Efficient than Fully-Connected Nets?

Co-Mixup: Saliency Guided Joint Mixup with Supermodular Diversity

Implicit Under-Parameterization Inhibits Data-Efficient Deep Reinforcement Learning

CopulaGNN: Towards Integrating Representational and Correlational Roles of Graphs in Graph Neural Networks

Behavioral Cloning from Noisy Demonstrations

Understanding the role of importance weighting for deep learning

Aligning AI With Shared Human Values

Large Batch Simulation for Deep Reinforcement Learning

Do Wide and Deep Networks Learn the Same Things? Uncovering How Neural Network Representations Vary with Width and Depth

Enjoy Your Editing: Controllable GANs for Image Editing via Latent Space Navigation

Multi-resolution modeling of a discrete stochastic process identifies causes of cancer

Generalized Variational Continual Learning

Contextual Dropout: An Efficient Sample-Dependent Dropout Module

Bowtie Networks: Generative Modeling for Joint Few-Shot Recognition and Novel-View Synthesis

Projected Latent Markov Chain Monte Carlo: Conditional Sampling of Normalizing Flows

Learning Safe Multi-agent Control with Decentralized Neural Barrier Certificates

Discovering Non-monotonic Autoregressive Orderings with Variational Inference

HalentNet: Multimodal Trajectory Forecasting with Hallucinative Intents

Learning to Reach Goals via Iterated Supervised Learning

Neural Mechanics: Symmetry and Broken Conservation Laws in Deep Learning Dynamics

Unsupervised Discovery of 3D Physical Objects

How Neural Networks Extrapolate: From Feedforward to Graph Neural Networks

Are Neural Rankers still Outperformed by Gradient Boosted Decision Trees?

Information Laundering for Model Privacy

Viewmaker Networks: Learning Views for Unsupervised Representation Learning

Linear Mode Connectivity in Multitask and Continual Learning

A Hypergradient Approach to Robust Regression without Correspondence

Discrete Graph Structure Learning for Forecasting Multiple Time Series

FedBE: Making Bayesian Model Ensemble Applicable to Federated Learning

Convex Potential Flows: Universal Probability Distributions with Optimal Transport and Convex Optimization

Gradient Descent on Neural Networks Typically Occurs at the Edge of Stability

(ends 7:00 PM)

7 p.m.

Oral Session 6 [7:00-10:16]

Oral s 7:00-7:15

[7:00] Deep symbolic regression: Recovering mathematical expressions from data via risk-seeking policy gradients

Spotlight s 7:15-7:45

[7:15] DDPNOpt: Differential Dynamic Programming Neural Optimizer

[7:25] Orthogonalizing Convolutional Layers with the Cayley Transform

[7:35] Model-Based Visual Planning with Self-Supervised Functional Distances

Q&A s 7:45-7:55

[7:45] Q&A

Oral s 7:55-8:10

[7:55] Global Convergence of Three-layer Neural Networks in the Mean Field Regime

Spotlight s 8:10-8:50

[8:10] Minimum Width for Universal Approximation

[8:20] Async-RED: A Provably Convergent Asynchronous Block Parallel Stochastic Method using Deep Denoising Priors

[8:30] Individually Fair Gradient Boosting

[8:40] Are Neural Rankers still Outperformed by Gradient Boosted Decision Trees?

Q&A s 8:50-9:03

[8:50] Q&A

Oral s 9:03-9:33

[9:03] Co-Mixup: Saliency Guided Joint Mixup with Supermodular Diversity

[9:18] MONGOOSE: A Learnable LSH Framework for Efficient Neural Network Training

Spotlight s 9:33-10:03

[9:33] Locally Free Weight Sharing for Network Width Search

[9:43] Memory Optimization for Deep Networks

[9:53] Neural Topic Model via Optimal Transport

Q&A s 10:03-10:16

[10:03] Q&A

(ends 10:16 PM)

10:30 p.m.

BREAK:

Please visit the Sponsor Hall, the Socials, and Mentorships.

WED 5 MAY

midnight

Invited Talk:

Perceiving the 3D World from Images and Video

Lourdes Agapito

(ends 1:00 AM)

1 a.m.

Poster Session 07 [1:00-3:00]

Graph Edit Networks

Neural ODE Processes

Dance Revolution: Long-Term Dance Generation with Music via Curriculum Learning

Removing Undesirable Feature Contributions Using Out-of-Distribution Data

Interpretable Neural Architecture Search via Bayesian Optimisation with Weisfeiler-Lehman Kernels

Return-Based Contrastive Representation Learning for Reinforcement Learning

Differentiable Segmentation of Sequences

Expressive Power of Invariant and Equivariant Graph Neural Networks

A Panda? No, It's a Sloth: Slowdown Attacks on Adaptive Multi-Exit Neural Network Inference

Robust Learning of Fixed-Structure Bayesian Networks in Nearly-Linear Time

Deployment-Efficient Reinforcement Learning via Model-Based Offline Optimization

Bag of Tricks for Adversarial Training

Negative Data Augmentation

Neural Delay Differential Equations

Efficient Continual Learning with Modular Networks and Task-Driven Priors

Improving Relational Regularized Autoencoders with Spherical Sliced Fused Gromov Wasserstein

Isometric Propagation Network for Generalized Zero-shot Learning

Simple Spectral Graph Convolution

An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

Active Contrastive Learning of Audio-Visual Video Representations

Net-DNF: Effective Deep Modeling of Tabular Data

Separation and Concentration in Deep Networks

Locally Free Weight Sharing for Network Width Search

CoCon: A Self-Supervised Approach for Controlled Text Generation

Neural gradients are near-lognormal: improved quantized and sparse training

Communication in Multi-Agent Reinforcement Learning: Intention Sharing

A Wigner-Eckart Theorem for Group Equivariant Convolution Kernels

Learning Associative Inference Using Fast Weight Memory

Reweighting Augmented Samples by Minimizing the Maximal Expected Loss

Bidirectional Variational Inference for Non-Autoregressive Text-to-Speech

Knowledge distillation via softmax regression representation learning

Gradient Origin Networks

Explainable Deep One-Class Classification

Discovering Diverse Multi-Agent Strategic Behavior via Reward Randomization

Acting in Delayed Environments with Non-Stationary Markov Policies

Self-supervised Adversarial Robustness for the Low-label, High-data Regime

IEPT: Instance-Level and Episode-Level Pretext Tasks for Few-Shot Learning

Explaining by Imitating: Understanding Decisions by Interpretable Policy Learning

Identifying Physical Law of Hamiltonian Systems via Meta-Learning

Degree-Quant: Quantization-Aware Training for Graph Neural Networks

Improving Adversarial Robustness via Channel-wise Activation Suppressing

A Better Alternative to Error Feedback for Communication-Efficient Distributed Learning

Augmenting Physical Models with Deep Networks for Complex Dynamics Forecasting

Long Range Arena : A Benchmark for Efficient Transformers

Learning from Demonstration with Weakly Supervised Disentanglement

Grounded Language Learning Fast and Slow

Byzantine-Resilient Non-Convex Stochastic Gradient Descent

Deep Neural Network Fingerprinting by Conferrable Adversarial Examples

Fidelity-based Deep Adiabatic Scheduling

Disentangled Recurrent Wasserstein Autoencoder

FOCAL: Efficient Fully-Offline Meta-Reinforcement Learning via Distance Metric Learning and Behavior Regularization

Geometry-aware Instance-reweighted Adversarial Training

High-Capacity Expert Binary Networks

BRECQ: Pushing the Limit of Post-Training Quantization by Block Reconstruction

Relating by Contrasting: A Data-efficient Framework for Multimodal Generative Models

Neural networks with late-phase weights

Deep Learning meets Projective Clustering

No Cost Likelihood Manipulation at Test Time for Making Better Mistakes in Deep Networks

New Bounds For Distributed Mean Estimation and Variance Reduction

DICE: Diversity in Deep Ensembles via Conditional Redundancy Adversarial Estimation

Fooling a Complete Neural Network Verifier

On Data-Augmentation and Consistency-Based Semi-Supervised Learning

Learning Task Decomposition with Ordered Memory Policy Network

Auxiliary Task Update Decomposition: The Good, the Bad and the Neutral

(ends 3:00 AM)

3 a.m.

Oral Session 7 [3:00-6:08]

Oral s 3:00-3:45

[3:00] An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

[3:15] Rethinking Attention with Performers

[3:30] Share or Not? Learning to Schedule Language-Specific Capacity for Multilingual Translation

Spotlight s 3:45-3:55

[3:45] Support-set bottlenecks for video-text representation learning

Q&A s 3:55-4:05

[3:55] Q&A

Oral s 4:05-4:20

[4:05] Getting a CLUE: A Method for Explaining Uncertainty Estimates

Spotlight s 4:20-4:50

[4:20] Influence Estimation for Generative Adversarial Networks

[4:30] Stabilized Medical Image Attacks

[4:40] Deep Neural Network Fingerprinting by Conferrable Adversarial Examples

Q&A s 4:50-5:00

[4:50] Q&A

Oral s 5:00-5:15

[5:00] Learning Cross-Domain Correspondence for Control with Dynamics Cycle-Consistency

Spotlight s 5:15-5:55

[5:15] Benefit of deep learning with non-convex noisy gradient descent: Provable excess risk bound and superiority to kernel methods

[5:25] Tent: Fully Test-Time Adaptation by Entropy Minimization

[5:35] Neural Approximate Sufficient Statistics for Implicit Models

[5:45] Implicit Normalizing Flows

Q&A s 5:55-6:08

[5:55] Q&A

(ends 6:08 AM)

6:15 a.m.

BREAK:

Please visit the Sponsor Hall, the Socials, and Mentorships.

8 a.m.

Invited Talk:

Is My Dataset Biased?

Kate Saenko

(ends 9:00 AM)

9 a.m.

Poster Session 08 [9:00-11:00]

Anytime Sampling for Autoregressive Models via Ordered Autoencoding

Learning advanced mathematical computations from examples

Witches' Brew: Industrial Scale Data Poisoning via Gradient Matching

Multiplicative Filter Networks

OPAL: Offline Primitive Discovery for Accelerating Offline Reinforcement Learning

Are Neural Nets Modular? Inspecting Functional Modularity Through Differentiable Weight Masks

Learning Task-General Representations with Generative Neuro-Symbolic Modeling

Learning to Represent Action Values as a Hypergraph on the Action Vertices

Faster Binary Embeddings for Preserving Euclidean Distances

On the Bottleneck of Graph Neural Networks and its Practical Implications

You Only Need Adversarial Supervision for Semantic Image Synthesis

Pre-training Text-to-Text Transformers for Concept-centric Common Sense

Multivariate Probabilistic Time Series Forecasting via Conditioned Normalizing Flows

Self-supervised Visual Reinforcement Learning with Object-centric Representations

Iterated learning for emergent systematicity in VQA

Average-case Acceleration for Bilinear Games and Normal Matrices

Continuous Wasserstein-2 Barycenter Estimation without Minimax Optimization

Long-tail learning via logit adjustment

Perceptual Adversarial Robustness: Defense Against Unseen Threat Models

IsarStep: a Benchmark for High-level Mathematical Reasoning

Differentiable Trust Region Layers for Deep Reinforcement Learning

Federated Learning via Posterior Averaging: A New Perspective and Practical Algorithms

Watch-And-Help: A Challenge for Social Perception and Human-AI Collaboration

Entropic gradient descent algorithms and wide flat minima

Coupled Oscillatory Recurrent Neural Network (coRNN): An accurate and (gradient) stable architecture for learning long time dependencies

Geometry-Aware Gradient Algorithms for Neural Architecture Search

How Does Mixup Help With Robustness and Generalization?

Orthogonalizing Convolutional Layers with the Cayley Transform

Unbiased Teacher for Semi-Supervised Object Detection

PC2WF: 3D Wireframe Reconstruction from Raw Point Clouds

RODE: Learning Roles to Decompose Multi-Agent Tasks

Graph Traversal with Tensor Functionals: A Meta-Algorithm for Scalable Learning

HyperDynamics: Meta-Learning Object and Agent Dynamics with Hypernetworks

Exploring the Uncertainty Properties of Neural Networks’ Implicit Priors in the Infinite-Width Limit

Theoretical bounds on estimation error for meta-learning

A Gradient Flow Framework For Analyzing Network Pruning

Creative Sketch Generation

Deep symbolic regression: Recovering mathematical expressions from data via risk-seeking policy gradients

DeepAveragers: Offline Reinforcement Learning By Solving Derived Non-Parametric MDPs

Learning Generalizable Visual Representations via Interactive Gameplay

Probabilistic Numeric Convolutional Neural Networks

My Body is a Cage: the Role of Morphology in Graph-Based Incompatible Control

Sliced Kernelized Stein Discrepancy

Ask Your Humans: Using Human Instructions to Improve Generalization in Reinforcement Learning

Property Controllable Variational Autoencoder via Invertible Mutual Dependence

Graph Information Bottleneck for Subgraph Recognition

Sharpness-aware Minimization for Efficiently Improving Generalization

Boost then Convolve: Gradient Boosting Meets Graph Neural Networks

Neural Networks for Learning Counterfactual G-Invariances from Single Environments

Chaos of Learning Beyond Zero-sum and Coordination via Game Decompositions

Optimism in Reinforcement Learning with Generalized Linear Function Approximation

Benchmarks for Deep Off-Policy Evaluation

Modeling the Second Player in Distributionally Robust Optimization

Mastering Atari with Discrete World Models

Variational State-Space Models for Localisation and Dense 3D Mapping in 6 DoF

DARTS-: Robustly Stepping out of Performance Collapse Without Indicators

Unsupervised Audiovisual Synthesis via Exemplar Autoencoders

Human-Level Performance in No-Press Diplomacy via Equilibrium Search

TropEx: An Algorithm for Extracting Linear Terms in Deep Neural Networks

Few-Shot Bayesian Optimization with Deep Kernel Surrogates

HyperGrid Transformers: Towards A Single Model for Multiple Tasks

Interpreting Graph Neural Networks for NLP With Differentiable Edge Masking

Evaluation of Neural Architectures Trained With Square Loss vs Cross-Entropy in Classification Tasks

For self-supervised learning, Rationality implies generalization, provably

Learning Mesh-Based Simulation with Graph Networks

Provably robust classification of adversarial examples with detection

Growing Efficient Deep Networks by Structured Continuous Sparsification

Early Stopping in Deep Networks: Double Descent and How to Eliminate it

NAS-Bench-ASR: Reproducible Neural Architecture Search for Speech Recognition

Anchor & Transform: Learning Sparse Embeddings for Large Vocabularies

SEED: Self-supervised Distillation For Visual Representation

More or Less: When and How to Build Convolutional Neural Network Ensembles

Generative Time-series Modeling with Fourier Flows

Neural Synthesis of Binaural Speech From Mono Audio

Training independent subnetworks for robust prediction

(ends 11:00 AM)

11 a.m.

Oral Session 8 [11:00-2:21]

Oral s 11:00-12:00

[11:00] Human-Level Performance in No-Press Diplomacy via Equilibrium Search

[11:15] Learning to Reach Goals via Iterated Supervised Learning

[11:30] Learning Invariant Representations for Reinforcement Learning without Reconstruction

[11:45] Evolving Reinforcement Learning Algorithms

Spotlight s 12:00-12:10

[12:00] Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixels

Q&A s 12:10-12:23

[12:10] Q&A

Oral s 12:23-12:38

[12:23] Coupled Oscillatory Recurrent Neural Network (coRNN): An accurate and (gradient) stable architecture for learning long time dependencies

Spotlight s 12:38-1:08

[12:38] Sequential Density Ratio Estimation for Simultaneous Optimization of Speed and Accuracy

[12:48] LambdaNetworks: Modeling long-range Interactions without Attention

[12:58] Grounded Language Learning Fast and Slow

Q&A s 1:08-1:18

[1:08] Q&A

Spotlight s 1:18-2:08

[1:18] Unsupervised Object Keypoint Learning using Local Spatial Predictability

[1:28] VAEBM: A Symbiosis between Variational Autoencoders and Energy-based Models

[1:38] Dynamic Tensor Rematerialization

[1:48] A Gradient Flow Framework For Analyzing Network Pruning

[1:58] Differentially Private Learning Needs Better Features (or Much More Data)

Q&A s 2:08-2:21

[2:08] Q&A

(ends 2:21 PM)

2 p.m.

Expo Talk Panel:

Live Panel - Academics@ Presents: Representation Learning at Amazon

(ends 3:00 PM)

3 p.m.

BREAK:

Please visit the Sponsor Hall, the Socials, and Mentorships.

4 p.m.

Outstanding Paper Session 1 [4:00-5:05]

Oral s 4:00-4:45

[4:00] Neural Synthesis of Binaural Speech From Mono Audio

[4:15] EigenGame: PCA as a Nash Equilibrium

[4:30] Score-Based Generative Modeling through Stochastic Differential Equations

Spotlight s 4:45-4:55

[4:45] Learning Mesh-Based Simulation with Graph Networks

Q&A s 4:55-5:05

[4:55] Q&A

(ends 5:05 PM)

5 p.m.

Poster Session 09 [5:00-7:00]

Efficient Wasserstein Natural Gradients for Reinforcement Learning

CoCo: Controllable Counterfactuals for Evaluating Dialogue State Trackers

Deep Encoder, Shallow Decoder: Reevaluating Non-autoregressive Machine Translation

Learning and Evaluating Representations for Deep One-Class Classification

CPR: Classifier-Projection Regularization for Continual Learning

Empirical or Invariant Risk Minimization? A Sample Complexity Perspective

Efficient Reinforcement Learning in Factored MDPs with Application to Constrained RL

Graph-Based Continual Learning

Evolving Reinforcement Learning Algorithms

Learning with Instance-Dependent Label Noise: A Sample Sieve Approach

Adaptive Universal Generalized PageRank Graph Neural Network

Learning to Deceive Knowledge Graph Augmented Models via Targeted Perturbation

In-N-Out: Pre-Training and Self-Training using Auxiliary Information for Out-of-Distribution Robustness

A Geometric Analysis of Deep Generative Image Models and Its Applications

AutoLRS: Automatic Learning-Rate Schedule by Bayesian Optimization on the Fly

Simple Augmentation Goes a Long Way: ADRL for DNN Quantization

BERTology Meets Biology: Interpreting Attention in Protein Language Models

Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective

Evaluations and Methods for Explanation through Robustness Analysis

Mind the Pad -- CNNs Can Develop Blind Spots

Beyond Fully-Connected Layers with Quaternions: Parameterization of Hypercomplex Multiplications with $1/n$ Parameters

Is Attention Better Than Matrix Decomposition?

Filtered Inner Product Projection for Crosslingual Embedding Alignment

Measuring Massive Multitask Language Understanding

Robust Overfitting may be mitigated by properly learned smoothening

Economic Hyperparameter Optimization With Blended Search Strategy

Evaluating the Disentanglement of Deep Generative Models through Manifold Topology

Modelling Hierarchical Structure between Dialogue Policy and Natural Language Generator with Option Framework for Task-oriented Dialogue System

Meta Back-Translation

Adaptive Procedural Task Generation for Hard-Exploration Problems

Conservative Safety Critics for Exploration

Learning with Feature-Dependent Label Noise: A Progressive Approach

ALFWorld: Aligning Text and Embodied Environments for Interactive Learning

Learning Manifold Patch-Based Representations of Man-Made Shapes

Protecting DNNs from Theft using an Ensemble of Diverse Models

Efficient Empowerment Estimation for Unsupervised Stabilization

Deep Neural Tangent Kernel and Laplace Kernel Have the Same RKHS

Wandering within a world: Online contextualized few-shot learning

Local Convergence Analysis of Gradient Descent Ascent with Finite Timescale Separation

Fast And Slow Learning Of Recurrent Independent Mechanisms

Efficient Conformal Prediction via Cascaded Inference with Expanded Admission

Empirical Analysis of Unlabeled Entity Problem in Named Entity Recognition

Estimating Lipschitz constants of monotone deep equilibrium models

Effective and Efficient Vote Attack on Capsule Networks

Task-Agnostic Morphology Evolution

Control-Aware Representations for Model-based Reinforcement Learning

Explainable Subgraph Reasoning for Forecasting on Temporal Knowledge Graphs

Improved Autoregressive Modeling with Distribution Smoothing

AdaFuse: Adaptive Temporal Fusion Network for Efficient Action Recognition

DiffWave: A Versatile Diffusion Model for Audio Synthesis

Better Fine-Tuning by Reducing Representational Collapse

Learning to Generate 3D Shapes with Generative Cellular Automata

Combining Physics and Machine Learning for Network Flow Estimation

NBDT: Neural-Backed Decision Tree

AdaSpeech: Adaptive Text to Speech for Custom Voice

Interactive Weak Supervision: Learning Useful Heuristics for Data Labeling

Generating Furry Cars: Disentangling Object Shape and Appearance across Multiple Domains

Influence Functions in Deep Learning Are Fragile

Beyond Categorical Label Representations for Image Classification

Mixed-Features Vectors and Subspace Splitting

Emergent Symbols through Binding in External Memory

INT: An Inequality Benchmark for Evaluating Generalization in Theorem Proving

Learning Long-term Visual Dynamics with Region Proposal Interaction Networks

PolarNet: Learning to Optimize Polar Keypoints for Keypoint Based Object Detection

Individually Fair Gradient Boosting

In Search of Lost Domain Generalization

On the Critical Role of Conventions in Adaptive Human-AI Collaboration

gradSim: Differentiable simulation for system identification and visuomotor control

GANs Can Play Lottery Tickets Too

A PAC-Bayesian Approach to Generalization Bounds for Graph Neural Networks

Estimating informativeness of samples with Smooth Unique Information

Revisiting Dynamic Convolution via Matrix Decomposition

Model-Based Visual Planning with Self-Supervised Functional Distances

(ends 7:00 PM)

7 p.m.

Oral Session 9 [7:00-10:05]

Oral s 7:00-7:15

[7:00] Improved Autoregressive Modeling with Distribution Smoothing

Spotlight s 7:15-7:45

[7:15] GAN "Steerability" without optimization

[7:25] Large Scale Image Completion via Co-Modulated Generative Adversarial Networks

[7:35] Emergent Symbols through Binding in External Memory

Q&A s 7:45-7:55

[7:45] Q&A

Oral s 7:55-8:10

[7:55] Deformable DETR: Deformable Transformers for End-to-End Object Detection

Spotlight s 8:10-9:00

[8:10] Graph-Based Continual Learning

[8:20] Understanding the role of importance weighting for deep learning

[8:30] Towards Robustness Against Natural Language Word Substitutions

[8:40] Undistillable: Making A Nasty Teacher That CANNOT teach students

[8:50] CPT: Efficient Deep Neural Network Training via Cyclic Precision

Q&A s 9:00-9:15

[9:00] Q&A

Spotlight s 9:15-9:55

[9:15] PlasticineLab: A Soft-Body Manipulation Benchmark with Differentiable Physics

[9:25] Regularization Matters in Policy Optimization - An Empirical Study on Continuous Control

[9:35] Regularized Inverse Reinforcement Learning

[9:45] Behavioral Cloning from Noisy Demonstrations

Q&A s 9:55-10:05

[9:55] Q&A

(ends 10:05 PM)

10:15 p.m.

BREAK:

Please visit the Sponsor Hall, the Socials, and Mentorships.

THU 6 MAY

midnight

Outstanding Paper Session 2 [12:00-1:05]

Oral s 12:00-12:45

[12:00] Rethinking Architecture Selection in Differentiable NAS

[12:15] Complex Query Answering with Neural Link Predictors

[12:30] Optimal Rates for Averaged Stochastic Gradient Descent under Neural Tangent Kernel Regime

Spotlight s 12:45-12:55

[12:45] Beyond Fully-Connected Layers with Quaternions: Parameterization of Hypercomplex Multiplications with $1/n$ Parameters

Q&A s 12:55-1:05

[12:55] Q&A

(ends 1:05 AM)

1 a.m.

Poster Session 10 [1:00-3:00]

Network Pruning That Matters: A Case Study on Retraining Variants

Certify or Predict: Boosting Certified Robustness with Compositional Architectures

Learning Reasoning Paths over Semantic Graphs for Video-grounded Dialogues

Interpretable Models for Granger Causality Using Self-explaining Neural Networks

Influence Estimation for Generative Adversarial Networks

Do not Let Privacy Overbill Utility: Gradient Embedding Perturbation for Private Learning

Incremental few-shot learning via vector quantization in deep embedded space

Colorization Transformer

Genetic Soft Updates for Policy Evolution in Deep Reinforcement Learning

Understanding the effects of data parallelism and sparsity on neural network training

Adaptive Extra-Gradient Methods for Min-Max Optimization and Games

Sparse Quantized Spectral Clustering

Revisiting Hierarchical Approach for Persistent Long-Term Video Prediction

Private Image Reconstruction from System Side Channels Using Generative Models

Grounding Language to Autonomously-Acquired Skills via Goal Generation

The inductive bias of ReLU networks on orthogonally separable data

AdamP: Slowing Down the Slowdown for Momentum Optimizers on Scale-invariant Weights

Categorical Normalizing Flows via Continuous Transformations

Kanerva++: Extending the Kanerva Machine With Differentiable, Locally Block Allocated Latent Memory

Continual learning in recurrent neural networks

Collective Robustness Certificates: Exploiting Interdependence in Graph Neural Networks

An Unsupervised Deep Learning Approach for Real-World Image Denoising

ChipNet: Budget-Aware Pruning with Heaviside Continuous Approximations

Balancing Constraints and Rewards with Meta-Gradient D4PG

Learning Deep Features in Instrumental Variable Regression

Emergent Road Rules In Multi-Agent Driving Environments

Hopfield Networks is All You Need

CausalWorld: A Robotic Manipulation Benchmark for Causal Structure and Transfer Learning

Neural Topic Model via Optimal Transport

Distilling Knowledge from Reader to Retriever for Question Answering

R-GAP: Recursive Gradient Attack on Privacy

Contrastive Explanations for Reinforcement Learning via Embedded Self Predictions

Adaptive and Generative Zero-Shot Learning

GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding

Latent Convergent Cross Mapping

Contrastive Learning with Adversarial Perturbations for Conditional Text Generation

Go with the flow: Adaptive control for Neural ODEs

Free Lunch for Few-shot Learning: Distribution Calibration

GAN "Steerability" without optimization

Adversarially Guided Actor-Critic

IOT: Instance-wise Layer Reordering for Transformer Structures

Efficient Generalized Spherical CNNs

Retrieval-Augmented Generation for Code Summarization via Hybrid GNN

Mind the Gap when Conditioning Amortised Inference in Sequential Latent-Variable Models

Learning continuous-time PDEs from sparse data with graph neural networks

Efficient Inference of Flexible Interaction in Spiking-neuron Networks

A Diffusion Theory For Deep Learning Dynamics: Stochastic Gradient Descent Exponentially Favors Flat Minima

Scalable Learning and MAP Inference for Nonsymmetric Determinantal Point Processes

Learnable Embedding sizes for Recommender Systems

What Matters for On-Policy Deep Actor-Critic Methods? A Large-Scale Study

Representation Balancing Offline Model-based Reinforcement Learning

Contrastive Divergence Learning is a Time Reversal Adversarial Game

Conditional Generative Modeling via Learning the Latent Space

Deformable DETR: Deformable Transformers for End-to-End Object Detection

Learning What To Do by Simulating the Past

Repurposing Pretrained Models for Robust Out-of-domain Few-Shot Learning

Counterfactual Generative Networks

Autoregressive Entity Retrieval

Optimal Conversion of Conventional Artificial Neural Networks to Spiking Neural Networks

Learning Neural Generative Dynamics for Molecular Conformation Generation

Loss Function Discovery for Object Detection via Convergence-Simulation Driven Search

Practical Massively Parallel Monte-Carlo Tree Search Applied to Molecular Design

Optimal Rates for Averaged Stochastic Gradient Descent under Neural Tangent Kernel Regime

Inductive Representation Learning in Temporal Networks via Causal Anonymous Walks

Impact of Representation Learning in Linear Bandits

(ends 3:00 AM)

3 a.m.

Oral Session 10 [3:00-5:40]

Oral s 3:00-3:15

[3:00] What Matters for On-Policy Deep Actor-Critic Methods? A Large-Scale Study

Spotlight s 3:15-4:05

[3:15] Winning the L2RPN Challenge: Power Grid Management via Semi-Markov Afterstate Actor-Critic

[3:25] UPDeT: Universal Multi-agent RL via Policy Decoupling with Transformers

[3:35] Quantifying Differences in Reward Functions

[3:45] Iterative Empirical Game Solving via Single Policy Best Response

[3:55] Discovering a set of policies for the worst case reward

Q&A s 4:05-4:20

[4:05] Q&A

Oral s 4:20-4:35

[4:20] Augmenting Physical Models with Deep Networks for Complex Dynamics Forecasting

Spotlight s 4:35-5:25

[4:35] Unlearnable Examples: Making Personal Data Unexploitable

[4:45] Self-supervised Visual Reinforcement Learning with Object-centric Representations

[4:55] On Self-Supervised Image Representations for GAN Evaluation

[5:05] Retrieval-Augmented Generation for Code Summarization via Hybrid GNN

[5:15] Practical Real Time Recurrent Learning with a Sparse Approximation

Q&A s 5:25-5:40

[5:25] Q&A

(ends 5:40 AM)

6 a.m.

BREAK:

Please visit the Sponsor Hall, the Socials, and Mentorships.

8 a.m.

Invited Talk:

Soft bodied robots for human centered design of robots for everyday life

Kyu Jin Cho

(ends 9:00 AM)

9 a.m.

Poster Session 11 [9:00-11:00]

Uncertainty in Gradient Boosting via Ensembles

Deep Networks and the Multiple Manifold Problem

Flowtron: an Autoregressive Flow-based Generative Network for Text-to-Speech Synthesis

Analyzing the Expressive Power of Graph Neural Networks in a Spectral Perspective

Graph Coarsening with Neural Networks

BSQ: Exploring Bit-Level Sparsity for Mixed-Precision Neural Network Quantization

Dual-mode ASR: Unify and Improve Streaming ASR with Full-context Modeling

Contrastive Learning with Hard Negative Samples

On Position Embeddings in BERT

Contrastive Behavioral Similarity Embeddings for Generalization in Reinforcement Learning

CaPC Learning: Confidential and Private Collaborative Learning

Hierarchical Reinforcement Learning by Discovering Intrinsic Options

Integrating Categorical Semantics into Unsupervised Domain Translation

Dynamic Tensor Rematerialization

Meta-learning with negative learning rates

Heating up decision boundaries: isocapacitory saturation, adversarial scenarios and generalization bounds

Generative Language-Grounded Policy in Vision-and-Language Navigation with Bayes' Rule

Neural Learning of One-of-Many Solutions for Combinatorial Problems in Structured Output Spaces

Enforcing robust control guarantees within neural network policies

Local Search Algorithms for Rank-Constrained Convex Optimization

Rethinking Soft Labels for Knowledge Distillation: A Bias–Variance Tradeoff Perspective

Conditionally Adaptive Multi-Task Learning: Improving Transfer Learning in NLP Using Fewer Parameters & Less Data

Initialization and Regularization of Factorized Neural Layers

Variational Information Bottleneck for Effective Low-Resource Fine-Tuning

Differentially Private Learning Needs Better Features (or Much More Data)

Lifelong Learning of Compositional Structures

Directed Acyclic Graph Neural Networks

Private Post-GAN Boosting

Implicit Gradient Regularization

Correcting experience replay for multi-agent communication

Learning to Recombine and Resample Data For Compositional Generalization

A teacher-student framework to distill future trajectories

Efficient Transformers in Reinforcement Learning using Actor-Learner Distillation

Learning Cross-Domain Correspondence for Control with Dynamics Cycle-Consistency

Robust early-learning: Hindering the memorization of noisy labels

EigenGame: PCA as a Nash Equilibrium

Linear Last-iterate Convergence in Constrained Saddle-point Optimization

Model-Based Offline Planning

PMI-Masking: Principled masking of correlated spans

EEC: Learning to Encode and Regenerate Images for Continual Learning

Meta-Learning of Structured Task Distributions in Humans and Machines

End-to-End Egospheric Spatial Memory

Understanding and Improving Encoder Layer Fusion in Sequence-to-Sequence Learning

Improving VAEs' Robustness to Adversarial Attack

Cut out the annotator, keep the cutout: better segmentation with weak supervision

Adversarial score matching and improved sampling for image generation

Neural Spatio-Temporal Point Processes

C-Learning: Learning to Achieve Goals via Recursive Classification

Gradient Vaccine: Investigating and Improving Multi-task Optimization in Massively Multilingual Models

Blending MPC & Value Function Approximation for Efficient Reinforcement Learning

Improving Zero-Shot Voice Style Transfer via Disentangled Representation Learning

Dataset Meta-Learning from Kernel Ridge-Regression

VCNet and Functional Targeted Regularization For Learning Causal Effects of Continuous Treatments

A Critique of Self-Expressive Deep Subspace Clustering

Noise or Signal: The Role of Image Backgrounds in Object Recognition

A Mathematical Exploration of Why Language Models Help Solve Downstream Tasks

Dataset Condensation with Gradient Matching

Deconstructing the Regularization of BatchNorm

Learning to Set Waypoints for Audio-Visual Navigation

Domain-Robust Visual Imitation Learning with Mutual Information Constraints

Bayesian Few-Shot Classification with One-vs-Each Pólya-Gamma Augmented Gaussian Processes

End-to-end Adversarial Text-to-Speech

Adversarially-Trained Deep Nets Transfer Better: Illustration on Image Classification

Multi-Class Uncertainty Calibration via Mutual Information Maximization-based Binning

Learning to live with Dale's principle: ANNs with separate excitatory and inhibitory units

BREEDS: Benchmarks for Subpopulation Shift

Understanding and Improving Lexical Choice in Non-Autoregressive Translation

(ends 11:00 AM)

11 a.m.

Oral Session 11 [11:00-2:13]

Oral s 11:00-12:00

[11:00] VCNet and Functional Targeted Regularization For Learning Causal Effects of Continuous Treatments

[11:15] SenSeI: Sensitive Set Invariance for Enforcing Individual Fairness

[11:30] When Do Curricula Work?

[11:45] Why Are Convolutional Nets More Sample-Efficient than Fully-Connected Nets?

Q&A s 12:00-12:10

[12:00] Q&A

Spotlight s 12:10-12:50

[12:10] Correcting experience replay for multi-agent communication

[12:20] Contrastive Behavioral Similarity Embeddings for Generalization in Reinforcement Learning

[12:30] DeepAveragers: Offline Reinforcement Learning By Solving Derived Non-Parametric MDPs

[12:40] Data-Efficient Reinforcement Learning with Self-Predictive Representations

Q&A s 12:50-1:00

[12:50] Q&A

Oral s 1:00-1:30

[1:00] DiffWave: A Versatile Diffusion Model for Audio Synthesis

[1:15] Self-training For Few-shot Transfer Across Extreme Task Differences

Spotlight s 1:30-2:00

[1:30] A Panda? No, It's a Sloth: Slowdown Attacks on Adaptive Multi-Exit Neural Network Inference

[1:40] BUSTLE: Bottom-Up Program Synthesis Through Learning-Guided Exploration

[1:50] Disentangled Recurrent Wasserstein Autoencoder

Q&A s 2:00-2:13

[2:00] Q&A

(ends 2:13 PM)

2 p.m.

Expo Talk Panel:

AI Model Efficiency Toolkit talk & demo

(ends 3:00 PM)

3 p.m.

BREAK:

Please visit the Sponsor Hall, the Socials, and Mentorships.

4 p.m.

Invited Talk:

Self-Supervision for Learning from the Bottom Up

Alexei Efros

(ends 5:00 PM)

5 p.m.

Poster Session 12 [5:00-7:00]

Generative Scene Graph Networks

Neural Pruning via Growing Regularization

Few-Shot Learning via Learning the Representation, Provably

CO2: Consistent Contrast for Unsupervised Visual Representation Learning

Molecule Optimization by Explainable Evolution

MARS: Markov Molecular Sampling for Multi-objective Drug Discovery

Distributional Sliced-Wasserstein and Applications to Generative Modeling

Combining Ensembles and Data Augmentation Can Harm Your Calibration

HeteroFL: Computation and Communication Efficient Federated Learning for Heterogeneous Clients

Non-asymptotic Confidence Intervals of Off-policy Evaluation: Primal and Dual Bounds

Large Scale Image Completion via Co-Modulated Generative Adversarial Networks

Contrastive Syn-to-Real Generalization

Linear Convergent Decentralized Optimization with Compression

In Defense of Pseudo-Labeling: An Uncertainty-Aware Pseudo-label Selection Framework for Semi-Supervised Learning

ANOCE: Analysis of Causal Effects with Multiple Mediators via Constrained Structural Learning

Minimum Width for Universal Approximation

Multi-timescale Representation Learning in LSTM Language Models

HW-NAS-Bench: Hardware-Aware Neural Architecture Search Benchmark

Multi-Prize Lottery Ticket Hypothesis: Finding Accurate Binary Neural Networks by Pruning A Randomly Weighted Network

CT-Net: Channel Tensorization Network for Video Classification

Factorizing Declarative and Procedural Knowledge in Structured, Dynamical Environments

Fast and Complete: Enabling Complete Neural Network Verification with Rapid and Massively Parallel Incomplete Verifiers

Learning perturbation sets for robust machine learning

The Recurrent Neural Tangent Kernel

BiPointNet: Binary Neural Network for Point Clouds

No MCMC for me: Amortized sampling for fast and stable training of energy-based models

Neural representation and generation for RNA secondary structures

On the Curse of Memory in Recurrent Neural Networks: Approximation and Optimization Analysis

Evaluation of Similarity-based Explanations

Long-tailed Recognition by Routing Diverse Distribution-Aware Experts

FedBN: Federated Learning on Non-IID Features via Local Batch Normalization

Extreme Memorization via Scale of Initialization

Async-RED: A Provably Convergent Asynchronous Block Parallel Stochastic Method using Deep Denoising Priors

Nonseparable Symplectic Neural Networks

Learning to Sample with Local and Global Contexts in Experience Replay Buffer

Group Equivariant Generative Adversarial Networks

A Design Space Study for LISTA and Beyond

ARMOURED: Adversarially Robust MOdels using Unlabeled data by REgularizing Diversity

Randomized Automatic Differentiation

Representing Partial Programs with Blended Abstract Semantics

Neural Thompson Sampling

Self-supervised Representation Learning with Relative Predictive Coding

$i$-Mix: A Domain-Agnostic Strategy for Contrastive Representation Learning

LowKey: Leveraging Adversarial Attacks to Protect Social Media Users from Facial Recognition

Adapting to Reward Progressivity via Spectral Reinforcement Learning

Self-supervised Learning from a Multi-view Perspective

Calibration of Neural Networks using Splines

Greedy-GQ with Variance Reduction: Finite-time Analysis and Improved Complexity

Fast Geometric Projections for Local Robustness Certification

Prototypical Representation Learning for Relation Extraction

Cross-Attentional Audio-Visual Fusion for Weakly-Supervised Action Localization

Convex Regularization behind Neural Reconstruction

CPT: Efficient Deep Neural Network Training via Cyclic Precision

Auto Seg-Loss: Searching Metric Surrogates for Semantic Segmentation

Stochastic Security: Adversarial Defense Using Long-Run Dynamics of Energy-Based Models

Theoretical Analysis of Self-Training with Deep Networks on Unlabeled Data

FastSpeech 2: Fast and High-Quality End-to-End Text to Speech

Global Convergence of Three-layer Neural Networks in the Mean Field Regime

Clustering-friendly Representation Learning via Instance Discrimination and Feature Decorrelation

Answering Complex Open-Domain Questions with Multi-Hop Dense Retrieval

Estimating and Evaluating Regression Predictive Uncertainty in Deep Object Detectors

A Learning Theoretic Perspective on Local Explainability

Generalization bounds via distillation

DynaTune: Dynamic Tensor Program Optimization in Deep Neural Network Compilation

Fully Unsupervised Diversity Denoising with Convolutional Variational Autoencoders

Heteroskedastic and Imbalanced Deep Learning with Adaptive Regularization

How Much Over-parameterization Is Sufficient to Learn Deep ReLU Networks?

Isotropy in the Contextual Embedding Space: Clusters and Manifolds

Combining Label Propagation and Simple Models out-performs Graph Neural Networks

Learning Energy-Based Generative Models via Coarse-to-Fine Expanding and Sampling

When Do Curricula Work?

Learning to Make Decisions via Submodular Regularization

GAN2GAN: Generative Noise Learning for Blind Denoising with Single Noisy Images

(ends 7:00 PM)

7 p.m.

Oral Session 12 [7:00-9:48]

Oral s 7:00-7:15

[7:00] Theoretical Analysis of Self-Training with Deep Networks on Unlabeled Data

Spotlight s 7:15-7:45

[7:15] Long-tailed Recognition by Routing Diverse Distribution-Aware Experts

[7:25] Self-Supervised Policy Adaptation during Deployment

[7:35] What are the Statistical Limits of Offline RL with Linear Function Approximation?

Q&A s 7:45-7:55

[7:45] Q&A

Spotlight s 7:55-8:45

[7:55] RMSprop converges with proper hyper-parameter

[8:05] A Good Image Generator Is What You Need for High-Resolution Video Synthesis

[8:15] Random Feature Attention

[8:25] Learning with Feature-Dependent Label Noise: A Progressive Approach

[8:35] Sparse Quantized Spectral Clustering

Q&A s 8:45-8:58

[8:45] Q&A

Spotlight s 8:58-9:38

[8:58] Learning a Latent Simplex in Input Sparsity Time

[9:08] Topology-Aware Segmentation Using Discrete Morse Theory

[9:18] MARS: Markov Molecular Sampling for Multi-objective Drug Discovery

[9:28] Distributional Sliced-Wasserstein and Applications to Generative Modeling

Q&A s 9:38-9:48

[9:38] Q&A

(ends 9:48 PM)

10 p.m.

BREAK:

Please visit the Sponsor Hall, the Socials, and Mentorships.

Remarks:

Closing Remarks

(ends 11:00 PM)

FRI 7 MAY

2:30 a.m.

Workshop:

Science and Engineering of Deep Learning

(ends 11:30 AM)

3:30 a.m.

Workshop:

Neural Compression: From Information Theory to Applications

(ends 1:30 PM)

4:45 a.m.

Workshop:

Hardware-Aware Efficient Training of Deep Learning Models

(ends 12:00 PM)

5 a.m.

Workshop:

S2D-OLAD: From shallow to deep, overcoming limited and adverse data

(ends 3:00 PM)

Workshop:

Geometric and Topological Representation Learning

(ends 4:00 PM)

5:15 a.m.

Workshop:

Beyond Static Papers: Rethinking How We Share Scientific Understanding in ML

(ends 10:00 AM)

5:45 a.m.

Workshop:

Self-Supervision for Reinforcement Learning

(ends 4:30 PM)

5:50 a.m.

Workshop:

Energy-Based Models: Current Perspectives, Challenges, and Opportunities

(ends 4:10 PM)

5:55 a.m.

Workshop:

The Role of Mathematical Reasoning in General Artificial Intelligence

(ends 2:35 PM)

Workshop:

AI for Public Health

(ends 3:00 PM)

6 a.m.

Workshop:

AIMOCC -- AI: Modeling Oceans and Climate Change

(ends 10:00 AM)

BREAK:

Please visit the Sponsor Hall, the Socials, and Mentorships.

Workshop:

A Roadmap to Never-Ending RL

(ends 1:55 PM)

Workshop:

Workshop on Neural Architecture Search

(ends 3:00 PM)

6:30 a.m.

Workshop:

How Can Findings About The Brain Improve AI Systems?

(ends 5:30 PM)

6:45 a.m.

Workshop:

Responsible AI (RAI)

(ends 7:00 PM)

7 a.m.

Workshop:

2nd Workshop on Practical ML for Developing Countries: Learning Under Limited/low Resource Scenarios

(ends 12:15 PM)

Workshop:

Generalization beyond the training distribution in brains and machines

(ends 3:55 PM)

Workshop:

Workshop on Weakly Supervised Learning

(ends 6:00 PM)

Workshop:

Neural Conversational AI: Bridging the Gap Between Research and Real World (NeuCAIR)

(ends 8:30 PM)

Workshop:

Synthetic Data Generation: Quality, Privacy, Bias

(ends 2:30 PM)

Workshop:

Workshop on Learning to Learn

(ends 12:45 PM)

7:45 a.m.

Workshop:

Workshop on Enormous Language Models: Perspectives and Benchmarks

(ends 5:00 PM)

7:55 a.m.

Workshop:

ICLR 2021 Workshop on Embodied Multimodal Learning (EML)

(ends 3:35 PM)

8 a.m.

Workshop:

Robust and reliable machine learning in the real world

(ends 5:05 PM)

8:30 a.m.

Workshop:

Workshop on Distributed and Private Machine Learning

(ends 1:00 PM)

8:45 a.m.

Workshop:

Deep Learning for Simulation

(ends 5:00 PM)

Workshop:

Machine Learning for Preventing and Combating Pandemics

(ends 5:50 PM)

Workshop:

Security and Safety in Machine Learning Systems

(ends 6:00 PM)

2 p.m.

BREAK:

Please visit the Sponsor Hall, the Socials, and Mentorships.