Deep Learning and Combinatorial Optimization - IPAM

Deep Learning and Combinatorial Optimization

February 22 - 25, 2021

Schedule

All times in this Schedule are Pacific Time (PT)

Monday, February 22, 2021

Morning Session

07:55-07:55
SESSION CHAIR: Louis-Martin Rousseau (École Polytechnique de Montréal)
07:55-08:00
Welcome & Opening Remarks: Dean Miguel García-Garibay (Dean of Physical Sciences, UCLA) and Dima Shlyakhtenko (Director, IPAM)
08:00-08:25
Oktay Gunluk (Cornell University)
Fair and Interpretable Decision Rules for Binary Classification
08:35-09:00
Ron Kimmel (Technion - Israel Institute of Technology)
Learning Geometry
09:10-09:25
Break
09:25-09:50
Sanjeeb Dash (IBM Watson Research Center)
A method to learn compact logical rules for knowledge graph reasoning
10:00-10:25
Xavier Bresson (National University of Singapore)
The Transformer Network for the Traveling Salesman Problem
10:35-10:50
Break
10:50-11:15
Kyle Cranmer (University of Wisconsin-Madison)
Quarks, hierarchical clustering, and combinatorial optimization
11:25-11:50
Azalia Mirhoseini (Google AI)
Learning to Solve Combinatorial Optimization Problems in Systems and Chip Design

Tuesday, February 23, 2021

Morning Session

08:00-08:00
SESSION CHAIR: Wouter Kool (University of Amsterdam)
08:00-08:25
Sebastian Pokutta (Konrad-Zuse-Zentrum für Informationstechnik (ZIB))
Structured ML Training via Conditional Gradients
08:35-09:00
Thiago Serra (Bucknell University)
Scaling Up Exact Neural Network Compression by ReLU Stability
09:10-09:25
Break
09:25-09:50
Yunhao Tang (Columbia University)
Reinforcement Learning for Integer Programming: Learning to Cut
10:00-10:25
Joseph Huchette (Rice University)
Neural network verification as piecewise linear optimization
10:35-10:50
Break
10:50-11:15
Emma Frejinger (University of Montreal)
Can ML Help in Solving Cargo Capacity Management Booking Control Problems?
11:25-11:50
Le Song (Georgia Institute of Technology)
A Framework For Differentiable Discovery Of Graph Algorithms

Afternoon Session

12:00-12:25
Petar Veličković (DeepMind Technologies)
Reasoning on Natural Inputs

Wednesday, February 24, 2021

Morning Session

08:00-08:00
SESSION CHAIR: Kyle Cranmer (New York University)
08:00-08:25
Vinod Nair (Google Brain )
Solving Mixed Integer Programs Using Neural Networks
08:35-09:00
Joan Bruna (New York University)
On Sparse Linear Programming and (simple) neural networks
09:10-09:25
Break
09:25-09:50
Laurent Charlin (HEC Montréal)
Exact Combinatorial Optimization with Graph Convolutional Neural Networks
10:00-10:25
Santanu Dey (Georgia Institute of Technology)
Solving SDPs by using sparse PCA
10:35-10:50
Break
10:50-11:15
Maxime Gasse (École Polytechnique de Montréal)
Ecole: A Gym-like Library for Machine Learning in Combinatorial Optimization Solvers
11:25-11:50
Wouter Kool (University of Amsterdam)
Deep Learning for Combinatorial Optimization: count your flops and make your flops count!

Afternoon Session

12:00-12:25
Ellen Vitercik (Stanford University)
How much data is sufficient to learn high-performing algorithms?

Thursday, February 25, 2021

Morning Session

08:00-08:00
SESSION CHAIR: Petar Veličković (DeepMind Technologies)
08:00-08:25
Zico Kolter (Carnegie Mellon University)
Fast semidefinite programming for (differentiable) combinatorial optimization
08:35-09:00
Stefano Gualandi (Università di Pavia)
Discrete Optimal Transport by Parallel Network Simplex
09:10-09:25
Break
09:25-09:50
Bistra Dilkina (University of Southern California (USC))
Decision-focused learning: integrating downstream combinatorics in ML
10:00-10:25
Stefanie Jegelka (Massachusetts Institute of Technology)
Task structure and generalization in graph neural networks
10:35-10:50
Break
10:50-11:15
Louis-Martin Rousseau (École Polytechnique de Montréal)
Combining Reinforcement Learning and Constraint Programming for Combinatorial Optimization
11:25-11:50
Tina Eliassi-Rad (Northeastern University)
Task-driven Network Discovery via Deep Reinforcement Learning on Embedded Spaces

Afternoon Session

12:00-12:25
Sanjeev Arora (Princeton University)
How to allow deep learning on your data without revealing your data
Printable Version