Phillip Lippe

Phillip Lippe

PhD student in Artificial Intelligence

University of Amsterdam, QUVA


I’m a PhD student in the QUVA lab at the University of Amsterdam supervised by Efstratios Gavves and Taco Cohen. I am also part of the ELLIS PhD program in cooperation with Qualcomm, and a Google Developer Expert for JAX and Flax. My research mostly focuses on the intersection of causality and machine learning, particularly on causal representation learning and temporal data.

Previously, I was a student researcher at Google DeepMind and an intern at Microsoft Research. Before starting my PhD, I completed my Master degree in Artificial Intelligence at the University of Amsterdam, with my thesis on Categorical Normalizing Flows. My Bachelor was a cooperative study program between the DHBW Stuttgart and Daimler/Mercedes-Benz, where I conducted three research internships on deep learning and computer vision for autonomous driving.

  • Causality and Deep Learning
  • Causal Representation Learning
  • Discrete structure learning
  • MSc Artificial Intelligence, 2020

    University of Amsterdam

  • BSc Computer Science, 2018

    DHBW Stuttgart


[Jan 2024] Our paper “Towards the Reusability and Compositionality of Causal Representations” has been accepted as oral at CLeaR 2024! We discuss how to adapt and compose causal representations to new environments.

[Nov 2023] Two papers accepted at the NeurIPS 2023 workshop on Causal Representation Learning! Visit us during the poster session on “Hierarchical Causal Representation Learning” and “Towards the Reusability and Compositionality of Causal Representations”.

[Sep 2023] Two papers accepted at NeurIPS 2023! We will present “PDE-Refiner: Achieving Accurate Long Rollouts with Neural PDE Solvers” as spotlight and “Rotating Features for Object Discovery” as oral in New Orleans.

[Aug 2023] I joined Google DeepMind as a Student Research Intern for the next three months, working on large-scale multimodal pretraining with Mostafa Dehghani and more!

[Jul 2023] Our paper “Modeling Accurate Long Rollouts with Temporal Neural PDE Solvers” has been accepted to the Frontiers4LCD workshop at ICML 2023 as an Oral! We show how Neural Operators can stay longer accurate with a fast, diffusion-like refinement process.

[May 2023] Our recent paper “BISCUIT: Causal Representation Learning from Binary Interactions” has been accepted to UAI 2023 as a Spotlight! We explore causal representation learning from low-level action information in environments like embodied AI!

[Mar 2023] I started as a Research Intern in the AI4Science team at Microsoft Research, Amsterdam! I will be working on neural surrogates for PDE solving.

[Feb 2023] I was invited to the “Rising Stars in AI Symposium” at KAUST and will give a talk on Learning Causal Variables from Temporal Observations.

[Jan 2023] We have 3 papers accepted at ICLR 2023! Check out the papers on Causal Representation Learning, Object-Centric Representation Learning, and Subset Sampling.

[Dec 2022] I gave a talk on “Machine Learning with JAX and Flax” at the Google Developer Student Club, University of Augsburg.

[Nov 2022] Our work on “Complex-Valued Autoencoders for Object Discovery” has been accepted to TLMR! Check it out here.

[Sep 2022] Our work on “Weakly-Supervised Causal Representation Learning” has been accepted to NeurIPS 2022! Check it out here.

[Aug 2022] I gave an invited talk on “Learning Causal Variables from Temporal Sequences with Interventions” at the First Workshop on Causal Representation Learning at UAI 2022! Slides available here.

[Aug 2022] I joined the ML Google Developer Expert program for JAX+Flax! In this program, I will continue creating new tutorials on various topics in Deep Learning, both in JAX+Flax and PyTorch (tutorial website).

[Jun 2022] Check out our new work “iCITRIS: Causal Representation Learning for Instantaneous Temporal Effects”, in which we discuss causal representation learning in the case of instantaneous effects (link)!

[Jun 2022] Interested in learning JAX with Flax? Check out our new tutorials on our RTD website, translated from PyTorch to JAX+Flax!

[May 2022] Our recent work, “CITRIS: Causal Identifiability from Temporal Intervened Sequences”, has been accepted to ICML 2022 as spotlight! We explore causal representation learning for multidimensional causal factors. Check it out here!

[Jan 2022] Our work on “Efficient Neural Causal Discovery without Acyclicity Constraints” has been accepted to ICLR 2022! The paper can be found here.

[Jul 2021] I presented our work “Efficient Neural Causal Discovery without Acyclicity Constraints” at the 8th Causal Inference Workshop at UAI as on oral!

[Jun 2021] I have been accepted to the OxML and the CIFAR DLRL summer schools 2021. Looking forward to meeting lots of new people!

[Jan 2021] Our work on “Categorical Normalizing Flows via Continuous Transformations” has been accepted to ICLR 2021.

[Dec 2020] We have reached 4th place in the NeurIPS 2020 challenge “Hateful Meme Detection”, hosted by Facebook AI. We summarized our approach in this paper, and presented it at NeurIPS 2020.

[Oct 2020] For the Deep Learning course 2020 and 2021 at UvA, we created a series of Jupyter notebooks to teach important concepts from an implementation perspective. The notebooks can be found here, and are also published in the PyTorch Lightning documentation.

[Sep 2020] I have started my PhD at the University of Amsterdam under the supervision of Efstratios Gavves and Taco Cohen.

Recent Publications

Recent Talks