Quantum Boltzmann Machines for Sample-Efficient Reinforcement Learning
Thore Gerlach · Michael Schenk · Verena Kain
Abstract
We propose Continuous Semi-Quantum Boltzmann Machines, extending previous free energy-based models to continuous-action reinforcement learning. We leverage exponential family priors and hybrid quantum-classical sampling to improve expressiveness and sample efficiency while reducing qubit requirements.
Chat is not available.
Successful Page Load