Title:	Learning based algorithm development for stochastic systems
DNr:	NAISS 2025/22-1116
Project Type:	NAISS Small Compute
Principal Investigator:	Arunava Naha <arunava.naha@liu.se>
Affiliation:	Linköpings universitet
Duration:	2025-08-22 – 2026-09-01
Classification:	20202
Keywords:

Abstract

I am currently working on two different problems, both of which require CPU-intensive computations, with additional GPU support being a valuable resource. The first problem involves controlling a closed-loop dynamical system with safety or risk constraints using actor-critic based reinforcement learning. The second problem focuses on developing a differentially private federated Gaussian process regression algorithm. To support our study, we need to conduct extensive simulation studies for both problems. While the computations are primarily CPU-intensive, we will utilize shallow neural networks in our models, which may benefit from additional GPU resources. Additionally, we will need to leverage multiple CPU cores in parallel.