Machine Learning

Project Overview

We are working on core models and algorithms in machine learning. One long-term interest is in deep network architectures. We are interested both in new layers and operators, and in new macroscopic structures and connectivity patterns. Another interest is the development of deep learning techniques for discrete structures, such as graphs and sets, and combinatorial optimization problems. Another line of work concerns settings such as continual learning, which depart from conventional offline batch supervised learning.


Non-deep Networks

Domain Generalization without Excess Empirical Risk

Neural Deep Equilibrium Solvers

Online Continual Learning with Natural Distribution Shifts: An Empirical Study with Visual Data

Stabilizing Equilibrium Models by Jacobian Regularization

Training Graph Neural Networks with 1000 Layers

Drinking from a Firehose: Continual Learning with Web-scale Natural Language

Auto-decoding Graphs

Multiscale Deep Equilibrium Models

Learning to Guide Random Search

Deep Equilibrium Models

The Limited Multi-Label Projection Layer

Trellis Networks for Sequence Modeling

Deep Layers as Stochastic Solvers

Multi-Task Learning as Multi-Objective Optimization

Combinatorial Optimization with Graph Convolutional Networks and Guided Tree Search

Deep Fundamental Matrix Estimation

Deep Continuous Clustering

An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling

Robust Continuous Clustering

Parameter Learning and Convergent Inference for Dense Random Fields

Efficient Inference in Fully Connected CRFs with Gaussian Edge Potentials