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.

Publications

Non-deep Networks

Ankit Goyal, Alexey Bochkovskiy, Jia Deng, and Vladlen Koltun
Advances in Neural Information Processing Systems (NeurIPS), 2022

Domain Generalization without Excess Empirical Risk

Ozan Sener and Vladlen Koltun
Advances in Neural Information Processing Systems (NeurIPS), 2022

Neural Deep Equilibrium Solvers

Shaojie Bai, Vladlen Koltun, and J. Zico Kolter
International Conference on Learning Representations (ICLR), 2022

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

Zhipeng Cai, Ozan Sener, and Vladlen Koltun
International Conference on Computer Vision (ICCV), 2021

Stabilizing Equilibrium Models by Jacobian Regularization

Shaojie Bai, Vladlen Koltun, and J. Zico Kolter
International Conference on Machine Learning (ICML), 2021

Training Graph Neural Networks with 1000 Layers

Guohao Li, Matthias Müller, Bernard Ghanem, and Vladlen Koltun
International Conference on Machine Learning (ICML), 2021

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

Hexiang Hu, Ozan Sener, Fei Sha, and Vladlen Koltun
IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(5), 2023

Auto-decoding Graphs

Sohil Atul Shah and Vladlen Koltun
Pre-print, arXiv:2006.02879, 2020

Multiscale Deep Equilibrium Models

Shaojie Bai, Vladlen Koltun, and J. Zico Kolter
Advances in Neural Information Processing Systems (NeurIPS), 2020
(Selected for full oral presentation)

Learning to Guide Random Search

Ozan Sener and Vladlen Koltun
International Conference on Learning Representations (ICLR), 2020

Deep Equilibrium Models

Shaojie Bai, J. Zico Kolter, and Vladlen Koltun
Advances in Neural Information Processing Systems (NeurIPS), 2019
(Selected for spotlight oral presentation)

The Limited Multi-Label Projection Layer

Brandon Amos, Vladlen Koltun, and J. Zico Kolter
Technical Report, arXiv:1906.08707, 2019

Trellis Networks for Sequence Modeling

Shaojie Bai, J. Zico Kolter, and Vladlen Koltun
International Conference on Learning Representations (ICLR), 2019

Deep Layers as Stochastic Solvers

Adel Bibi, Bernard Ghanem, Vladlen Koltun, and René Ranftl
International Conference on Learning Representations (ICLR), 2019

Multi-Task Learning as Multi-Objective Optimization

Ozan Sener and Vladlen Koltun
Advances in Neural Information Processing Systems (NIPS), 2018

Combinatorial Optimization with Graph Convolutional Networks and Guided Tree Search

Zhuwen Li, Qifeng Chen, and Vladlen Koltun
Advances in Neural Information Processing Systems (NIPS), 2018

Deep Fundamental Matrix Estimation

René Ranftl and Vladlen Koltun
European Conference on Computer Vision (ECCV), 2018

Deep Continuous Clustering

Sohil Atul Shah and Vladlen Koltun
Technical Report, arXiv:1803.01449, 2018

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

Shaojie Bai, J. Zico Kolter, and Vladlen Koltun
Technical Report, arXiv:1803.01271, 2018

Robust Continuous Clustering

Sohil Atul Shah and Vladlen Koltun
Proceedings of the National Academy of Sciences (PNAS), 114(37), 2017

Parameter Learning and Convergent Inference for Dense Random Fields

Philipp Krähenbühl and Vladlen Koltun
International Conference on Machine Learning (ICML), 2013

Efficient Inference in Fully Connected CRFs with Gaussian Edge Potentials

Philipp Krähenbühl and Vladlen Koltun
Advances in Neural Information Processing Systems, 2011 (Oustanding Student Paper Award)