Elise van der Pol

Senior Researcher, Microsoft Research Amsterdam

• E-mail: evanderpol[at][company][dot]com
• Twitter
• Mastodon
• Google Scholar
• Github

My research interests lie in AI4Science, and structure, symmetry, and equivariance in reinforcement learning and machine learning. I'm a Senior Researcher at Microsoft Research Amsterdam, working on reinforcement learning and deep learning for molecular simulation. Before joining MSR, I completed my PhD on the topic of Symmetry and Structure in Deep Reinforcement Learning in the Amsterdam Machine Learning Lab, working with Max Welling (UvA), Frans Oliehoek (TU Delft) and Herke van Hoof (UvA). Before my PhD, I studied Artificial Intelligence at the University of Amsterdam where I obtained my Master's degree (cum laude). My MSc thesis was on the subject of Deep Reinforcement Learning for Coordinated Traffic Light Controllers. I was also involved in Inclusive AI.

Jump to papers.

News

2023

• Book: I was featured in the PhD narratives book Why would I want to do a PhD in computer science?.
• Talk: 25th October 2023 I gave an invited talk at the Chalmers Structured Learning Workshop.
• Talk: 10th October 2023 I gave a talk at the WAI NL Summit.
• Nomination: I was nominated for the WAI NL Researcher Award.
• Paper: One-shot Imitation Learning via Interaction Warping was accepted to CoRL 2023.
• Defense: 12th July 2023 I successfully defended my PhD thesis entitled Symmetry and Structure in Deep Reinforcement Learning and obtained my PhD degree.
• Organization: We organized a workshop on Symmetries in Robot Learning at RSS 2023.
• Lecture: 5th July 2023 I gave a lecture on Symmetries in RL in the Barcelona Reinforcement Learning Summer School
• Paper: New preprint: One-shot Imitation Learning via Interaction Warping.
• Organization: We organized a workshop on Physics for Machine Learning at ICLR 2023.
• Talk: 2nd February 2023 I gave an invited talk in the MILA Geometric Deep Learning Reading Group.
• Profile: I was profiled by Women in Machine Learning.

2022

• Paper: Maximum Class Separation as Inductive Bias in One Matrix was accepted to NeurIPS 2022 as an oral.
• Paper: Equivariant Networks for Zero-Shot Coordination was accepted to NeurIPS 2022.
• Talk: 14 December 2022 I gave an invited talk at the icarl seminars at Imperial College London.
• Paper: Stochastic Parallelizable Eigengap Dilation for Large Graph Clustering was accepted to the ICML 2022 Workshop on Topology, Algebra, and Geometry in Machine Learning 2022.
• Lecture: November 2022 I gave a guest lecture Introduction to Reinforcement Learning in University of Amsterdam's MSc Data Science course Applied Machine Learning.
• Paper: New preprint: Equivariant Networks for Zero-Shot Coordination.
• Paper: New preprint: Maximum Class Separation as Inductive Bias in One Matrix.
• Paper: New preprint: Stochastic Parallelizable Eigengap Dilation for Large Graph Clustering.
• Job: I've joined Microsoft Research Amsterdam as a Senior Researcher. I'll be working on reinforcement learning and deep learning for molecular simulation.
• Lecture: 14th July 2022 I'm giving a lecture on Symmetries in Reinforcement Learning at the Amsterdam Reinforcement Learning Summer School.
• Talk: 03 June 2022 I'm giving an invited talk at BeneRL 2022.
• Talk: 16 May 2022 I'm giving an invited talk at Stanford University.
• Talk: 19 April 2022 I gave an invited talk at University of Oxford.
• Blog: I was invited to contribute my thoughts on GNNs in RL to What does 2022 hold for Geometric and Graph ML? by Michael Bronstein and Petar Veličković.
• Paper: Multi-Agent MDP Homomorphic Networks was accepted to ICLR 2022.
• Paper: Geometric and Physical Quantities improve E(3) Equivariant Message Passing was accepted to ICLR 2022 as a spotlight.
• Organization: We're organizing a workshop on gamification and multi-agent solutions at ICLR 2022.

2021

• Lecture: 30 November 2021 I gave a guest lecture Introduction to Reinforcement Learning in University of Amsterdam's MSc Data Science course Applied Machine Learning.
• Paper: New preprint: Multi-Agent MDP Homomorphic Networks.
• Paper: New preprint: Geometric and Physical Quantities improve E(3) Equivariant Message Passing.
• Organization: We're organizing a workshop on data-centric reinforcement learning at NeurIPS 2021.
• Talk: 15 July 2021 I gave a talk on Geometric Deep Learning and Reinforcement Learning for the Geometric Deep Learning course in the African Masters in Machine Intelligence.
• Paper: The Impact of Negative Sampling on Contrastive Structured World Models was accepted to the ICML 2021 workshop on Self-Supervised Learning for Reasoning and Perception.
• Visit: I was a research scientist intern in DeepMind's Multi-Agent team.
• Organization: We organized an ICLR social which took place 6 May 2021.
• Talk: 7 May 2021 I gave a keynote talk on Equivariance in MDPs at the ICLR workshop on Self-supervision for Reinforcement Learning.

2020

• Talk: 15 December 2020 I gave a talk on MDP Homomorphic Networks for Deep Reinforcement Learning at the University of Tuebingen.
• Lecture: 23 November 2020 I gave a guest lecture Introduction to Reinforcement Learning in University of Amsterdam's MSc Data Science course Applied Machine Learning.
• Talk: 13 November 2020 I gave a talk on Equivariance, deep reinforcement learning, and making it work at Leiden University.
• Talk: I gave a talk on Data-efficient Reinforcement Learning as part of TU Delft's AI in Practice course.
• Paper: MDP Homomorphic Networks: Group Symmetries in Reinforcement Learning was accepted to NeurIPS 2020.
• Paper: Plannable Approximations to MDP Homomorphisms: Equivariance under Actions was accepted to AAMAS 2020.
• Paper: Contrastive Learning of Structured World Models was accepted to ICLR 2020 as a long talk.

2019

• Paper: Hyperspherical Prototype Networks was accepted to NeurIPS 2019.

Publications and Preprints

2023

• One-shot Imitation Learning via Interaction Warping - O. Biza, S. Thompson, K.R. Pagidi, A. Kumar, E. van der Pol, R. Walters, T. Kipf, J. van de Meent, L.L.S. Wong, R. Platt, accepted to CoRL 2023
  · PDF
• Symmetry and Structure in Deep Reinforcement Learning - E. van der Pol, PhD Thesis, University of Amsterdam
  · PDF · Cover · On the university website

2022

• Maximum Class Separation as Inductive Bias in One Matrix - T. Kasarla, G. J. Burghouts, M. van Spengler, E. van der Pol, R. Cucchiara, P. Mettes, NeurIPS 2022 (Oral)
  · PDF · Overview · Code
• Equivariant Networks for Zero-Shot Coordination - D. Muglich, C. Schroeder de Witt, E. van der Pol, S. Whiteson, J. Foerster, NeurIPS 2022
  · PDF
• Stochastic Parallelizable Eigengap Dilation for Large Graph Clustering - E. van der Pol, I. Gemp, Y. Bachrach, R. Everett, ICML 2022 Workshop on Topology, Algebra, and Geometry in Machine Learning
  · PDF · Overview
• Multi-Agent MDP Homomorphic Networks - E. van der Pol, H. van Hoof, F.A. Oliehoek, M. Welling, ICLR 2022
  · PDF · Code
• Geometric and Physical Quantities improve E(3) Equivariant Message Passing - J. Brandstetter, R. Hesselink, E. van der Pol, E.J. Bekkers, M. Welling, ICLR 2022 (Spotlight)
  · PDF · Code

2021

• The Impact of Negative Sampling on Contrastive Structured World Models - O. Biza, E. van der Pol, T. Kipf, ICML 2021 workshop on Self-Supervised Learning for Reasoning and Perception
  · PDF · Code

2020

• MDP Homomorphic Networks: Group Symmetries in Reinforcement Learning - E. van der Pol, D.E. Worrall, H. van Hoof, F.A. Oliehoek, M. Welling, NeurIPS 2020
  · PDF · Overview · Video · Code · Supplementary
• Plannable Approximations to MDP Homomorphisms: Equivariance under Actions - E. van der Pol, T. Kipf, F.A. Oliehoek, M. Welling, AAMAS 2020
  · PDF · Overview · Video · Code
• Contrastive Learning of Structured World Models - T. Kipf, E. van der Pol, M. Welling, ICLR 2020 (Oral)
  · PDF · Video · Code

2019

• Hyperspherical Prototype Networks - P. Mettes, E. van der Pol, C.G.M. Snoek, NeurIPS 2019
  · PDF · Code

2018

• Beyond Local Nash Equilibria for Adversarial Networks - F.A. Oliehoek, R. Savani, J. Gallego-Posada, E. van der Pol, R. Groß, Benelearn 2018
  · PDF
• Visual Rationalizations in Deep Reinforcement Learning for Atari Games - L. Weitkamp, E. van der Pol, Z. Akata, BNAIC 2018 (Oral)
  · PDF

2017

• GANGs: Generative Adversarial Network Games - F.A. Oliehoek, R. Savani, J. Gallego-Posada, E. van der Pol, E.D. de Jong, R. Groß, Arxiv Preprint
  · PDF

2016

• Deep Reinforcement Learning for Coordination in Traffic Light Control - E. van der Pol, Master's Thesis, University of Amsterdam
  · PDF
• Coordinated Deep Reinforcement Learners for Traffic Light Control - E. van der Pol, F.A. Oliehoek, NIPS 2016 Workshop on Learning, Inference and Control of Multi-Agent Systems
  · PDF
• Video Demo: Deep Reinforcement Learning for Coordination in Traffic Light Control - E. van der Pol, F.A. Oliehoek, BNAIC 2016
  · PDF · Video
• Linguistic Style Accommodation in Disagreements - E. van der Pol, S. Gieske, R. Fernández, *SEM 2016
  · PDF

2015

• Empirical Evaluation of Collective Rationality for Quota Rules in Judgment Aggregation - S. Gieske, E. van der Pol, U. Endriss, BNAIC 2015
  · PDF