Pascal Leroy

We released IMP-MARL: a Suite of Environments for Large-scale Infrastructure Management Planning via MARL. In IMP-MARL, agents decide the inspections and repairs needed to limit the risk of system failure, based on the probability distributions of damage. arxiv.org/abs/2306.11551

Interesting to link asymmetric learning and CTDE in MARL!

📝 "A Theoretical Justification for Asymmetric Actor-Critic Algorithms" was accepted at #ICML! Never heard of asymmetric actor-critic algorithms? Yet, many successful #RL applications use them. But these algorithms are not fully understood. Below, we provide some insights.

[Thanks for sharing] Dear all, my university is opening a call to give two-year positions for post-docs. If you are willing to join my group to work for example on energy systems, AI for defense, LLMs for the industry or reinforcement learning (all the things in which I am interested :) ), please do not hesitate to contact me.

[Thanks for sharing] We will have the immense pleasure to welcome the fantastic researcher @FlorianFelten1 for a research talk on Multi-Objective Reinforcement Learning in my research unit this Friday. Please do not hesitate to come! More information: montefiore.uliege.be/cms/c_12308857…

@FrancoisRozet Well done!

Our work was accepted at NeurIPS 2024 🥳 See you in Vancouver 🇨🇦

📢 Are you interested in diffusion models and inverse problems? Check out our new preprint "Learning Diffusion Priors from Observations by Expectation Maximization" with @g4ndry, François Lanusse and @glouppe! 🧵 arxiv.org/abs/2405.13712

📝 New paper "Parallelizing Autoregressive Generation with Variational State Space Models". It is an SSM-based sequence model that generates in //, while remaining recurrent. We will present it at the #ICML2024 workshop on the Next Generation of Sequence Modeling Architectures.

• Let us call him now Dr @PaLeroy, my 18th PhD student to graduate.  • He successfully defended his PhD thesis related to multi-agent reinforcement learning.  • His excellent manuscript can be downloaded here: hdl.handle.net/2268/317218 #reinforcementLearning #artificialIntelligence

The position is still open! 👇

Checkout this new workshop on coordination and cooperation in #MARL.

📢 Open PhD position in my group, on simulation-based inference and generative models for particle physics, on a joint project with Fabio Maltoni and Tilman Plehn inspirehep.net/jobs/2769416

[Thansk for sharing]. We are hiring new people for developing and using our tool GBOML for the planification of complex  energy systems. Here is the link to the job offer: forum.openmod.org/t/research-pos… #energyTransition  #optimisation #machineLearning

Our Multi-Agent RL framework EPyMARL, which extends the original PyMARL framework of @whi_rl, is widely used in MARL research (nearly 400 stars on Github!) and several MARL benchmarks have been built on it. We list three in this thread! Blog: agents.inf.ed.ac.uk/blog/epymarl/ (1/4)

In this episode of Mate with a Fantastic Researcher, @AdrienBolland from @DamienERNST1 takes us on a journey to explore reinforcement learning. He defines a set of criteria for when intrinsic and entropy-based RL works. He analyses them in a simple set of experiments to convey deep insights into the process of exploration in #RL. This is one of the most profound and well-executed analysis papers I have seen in a while! It is really awesome to see that some teams are still caring for fundamentals beyond the hype! Don't miss the episode: youtu.be/nqp1LUaFecQ Cheers to getting away from the philosophical hype :P #AI #MachineLearning

@EugeneVinitsky @SimonPrinceAI Agreed! I read the first twelve chapters to refresh my knowledge and I learned lots of new things. Can't wait to read the other chapters. I particularly liked the intermediate problems, which allow you to build better intuition along the way. Figurines are also really nice.

@SimonPrinceAI's "Understanding Deep Learning" is up to the hype. Unclear where else you'd find all these near-research-frontier topics so concisely explained

@prateekbhustali I would love to see a N vs N version!

Idea for MARL env

🎉 Exciting news! TorchRL paper has been accepted as a spotlight talk at @iclr_conf in Vienna this May! Huge thanks to the amazing OSS community, there would be no torchrl if it wasn't for our contributors. See you there! #ICLR2024 #Vienna arxiv.org/abs/2306.00577

@PabloGMorato Where are the selfies people?

After chatting with DALL-E about my impressions from #NeurIPS2023 poster sessions 😀

IMP-MARL in #NeurIPS2023, it’s now!