Lars C.P.M. Quaedvlieg

Lausanne, Switzerland ⛰️
Paris, France 🎨

Welcome to my personal website! I’m currently doing a research internship at InstaDeep, and I am pursuing an MSc in Data Science at École Polytechnique Fédérale de Lausanne, where I also do research at the Laboratory for Information and Inference Systems. My current research revolves around transformers and reinforcement learning, with a focus on foundation models for decision-making problems. Prior to my master’s degree, I completed a BSc in Data Science and Artificial Intelligence 🎓 at Maastricht University.

I am also currently involved with Rounded.News as the Chief AI Strategist, a start-up attempting to break filter bubbles using AI-powered news 📰.

During my bachelor’s studies, I had the opportunity to delve into applied research in various areas. As an AI research intern at Aucos AG, I worked on projects involving multi-camera multi-object tracking, and multi-agent systems.

My primary research interests lie within the exciting fields of machine learning and artificial intelligence, with a particular emphasis on reinforcement learning and multi-task learning. For more in-depth information about my work, feel free to explore the publications and projects pages!

Additionally, I am committed to leveraging my skills and knowledge to create positive impact and make a difference in people’s lives. I believe in the power of collaboration, innovation, and entrepreneurship to drive meaningful change. Through my involvement in various associations, hackathons, and initiatives, I have gained experience in addressing real-world challenges and contributing to sustainability efforts ♻. I am excited about opportunities to further expand my horizons and contribute to projects that align with this ideology.

When I’m not doing research or working on my projects, you can find me skiing down slopes ⛷, swimming in mountain lakes 🏖️ , exploring scenic hiking trails 🥾🏞️, or playing field hockey 🏑.

Thank you for visiting my website! I welcome collaboration and invite you to reach out (see the bottom of this page), whether you have ideas for impactful projects, opportunities for partnership, or simply want to connect!

News

Jul 13, 2023	I’m very excited to announce that I will be joining InstaDeep in Paris for the upcoming six months as a Research Intern. Together with the team, I will be working on the development of a general agent for multiple combinatorial optimization problems!
May 24, 2023	I’m thrilled to share that I’ve just submitted my first paper to NeurIPS! It has been an amazing and action-packed journey, filled with both hard work and enjoyment. Now, I eagerly await the outcome and hope that my paper gets accepted. Stay tuned for updates!

Selected publications

Maximum Independent Set: Self-Training through Dynamic Programming

L.C.P.M. Quaedvlieg*, L. Brusca*, S. Skoulakis, G. Chrysos, and V. Cevher

Advances in Neural Information Processing Systems (NeurIPS), Jul 2023

Abs

This work presents a novel graph neural network (GNN) framework for solving the maximum independent set (MIS) inspired by dynamic programming (DP). Specifically, given a graph, we propose a DP-like recursive algorithm based on GNNs that firstly constructs two smaller sub-graphs, predicts the one with the larger MIS, and then uses it in the next recursive call. To train our algorithm, we require annotated comparisons of different graphs concerning their MIS size. Annotating the comparisons with the output of our algorithm leads to a self-training process that results in more accurate self-annotation of the comparisons and vice versa. We provide numerical evidence showing the superiority of our method vs prior methods in multiple synthetic and real-world datasets.
Multi-Agent Reinforcement Learning with Graph Neural Networks for Online Multi-Hoist Scheduling

L.C.P.M. Quaedvlieg

Bachelor’s Thesis, Jul 2022

Abs PDF

This thesis explores an approach to solving the online multi-hoist scheduling problem by combining graph neural networks and multi-agent reinforcement learning. It approaches the problem by creating two sets of agents: source agents and hoists. When requested, a source agent selects a job from a queue of jobs in a source station to give to hoists. The hoists are responsible for picking up and dropping off jobs at stations, and coordinate with each other to avoid scenarios that would result in a deadlock. The devised algorithms are trained and benchmarked against other approaches, such as random and heuristic algorithms. Further insights into the methods are obtained from an analysis using dimensionality reduction on neural activations of the environment states. The results indicate that deadlocks are avoided in all experimental results. Furthermore, the approach in this thesis outperforms the other approaches in the benchmark by 7.50% to 10%. By analyzing the neural activations, it is shown that the hoist agents estimate that situations with many jobs being processed yield a higher job throughput than situations with less. Further research into the overall approach is recommended, but the results show potential to perform well against other approaches in existing literature.