Omer Rochman

Currently I'm doing a PhD at ULiège, under the supervision of Prof. Gilles Louppe. My main interest are neural emulation and solutions of XDEs, and I'm currently exploring the potential of diffusion models to solve forward and inverse PDE problems. Some of my recent work includes a fast and efficient neural emulator for particle-based simulations, a small study on the efficacy of different rollout stabilizations strategies, and a differentiable algorithm for model discovery. Previously, I worked as an AI researcher at appliedAI, a Munich based AI consultancy. I did my master's in Mathematics at the Technical University of Munich (TUM) and my bachelor's in Physics at the University of Leipzig.

Email / Twitter / BlueSky / Github

	A Neural Material Point Method for Particle-based Simulations Omer Rochman Sharabi, Sacha Lewin, Gilles Louppe (2024) project page / code / arXiv Under review. NeuralMPM can emulate point-cloud fluid systems that include multiple materials, each with their specific properties, at over 1000 FPS, vs 15 for the original simulator. It only needs trajectories of positions and velocities to learn, bypassing the tricky tuning of a simulator.
	Trick or treat? Evaluating stability strategies in graph network-based simulators Omer Rochman Sharabi, Gilles Louppe (2023) Paper We evaluated different stabilization strategies in the context of graph-based neural emulators (GNS).
	Differentiable composition for model discovery Omer Rochman Sharabi, Gilles Louppe (2022) Paper We propose DiffComp, a symbolic regressor that can learn arbitrary function compositions, including derivatives of various orders.
	Solving Schrödinger’s equation with Deep Learning Omer Rochman Sharabi Benjamin Schubert, Fabian Theis (2020) Master thesis We attempted to solve the Time Dependent Schrödinger Equation using PINNs, finding that they are unable to, and examine that failure.

Design stolen from Jon Barron's website. Source code