Gianluca Galletti

I'm a Master's student the final semester of computer science at TU Munich. Before that I got my Bachelor's degree at the University of Bologna, my birthplace and hometown.

From mid 2024 I'll be looking for a PhD position, ideally in the fields of Machine Learning for Physics and Dynamic Processes, Reinforcement Learning or 3D Computer Vision.

I have been practicing Traditional Archery for 5 years. I also enjoy Car Racing, Hiking, and sometimes Climbing.

Github / Google Scholar / CV / Email

Research

Currently working on Graph Neural Networks and Equivariance for particle fluid problems. Previously, I took part of various projects involving Reinforcement Learning, Process Mining, Computer Vision and Autonomous Driving.

Publications

JAX-SPH: A Differentiable Smoothed Particle Hydrodynamics Framework -- [code]
A. P. Toshev, H. Ramachandran*, J. A. Erbesdobler*, G. Galletti*, J. Brandstetter, N. A. Adams
2024, ICLR AI4Differential Workshop.

LagrangeBench: A Lagrangian Fluid Mechanics Benchmarking Suite -- [code]
A. P. Toshev*, G. Galletti*, F. Fritz, S. Adami, N. A. Adams
2023, NeurIPS Track on Datasets and Benchmarks.

Learning Lagrangian Fluid Mechanics with E(3)-Equivariant Graph Neural Networks -- [code]
A. P. Toshev, G. Galletti, J. Brandstetter, S. Adami, N. A. Adams
2023, Geometric Science of Information.

*Equal contribution

Projects

LagrangeBench

lagrangebench

LagrangeBench is a JAX-based machine learning benchmarking suite for Lagrangian problems. It provides a variety of models, datasets and utilities to train and evaluate particle-based approaches.

pyTORCS

pyTORCS-docker

Container-based TORCS environment/interface for reinforcement learning developed in Python, following the OpenAI Gym API style. TORCS is an open-source racing game. Part of my BSc thesis.

Equivariant JAX

segnn-jax painn-jax egnn-jax

Collection of some popular equivariant models implemented with the JAX library. Performance and runtimes are validated on the original configuration and against the results from the official papers.

Last update 10.12.2023

Template by Jon Barron.