Freiburg 2019 – scientific programme
Sessions | Days | Selection | Search | Updates | Downloads | Help
FM: Fall Meeting
FM 23: Quantum & Information Science: Neural Networks, Machine Learning, and Artificial Intelligence I
FM 23.2: Invited Talk
Monday, September 23, 2019, 17:00–17:30, 3043
A framework for deep energy-based reinforcement learning with quantum speed-up — •Sofiene Jerbi1, Hendrik Poulsen Nautrup1, Lea Trenkwalder1, Hans Briegel1,2, and Vedran Dunjko3 — 1University of Innsbruck, Innsbruck, Austria — 2University of Konstanz, Konstanz, Germany — 3University of Leiden, Leiden, Netherlands
In the past decade, deep neural network approaches to machine learning have seen tremendous success in classification and generative modeling tasks. More recently, deep neural networks have emerged in the domain of reinforcement learning as a tool to solve decision-making problems at an unprecedented scale. However, this rapid evolution of deep reinforcement learning has its roots in mathematical optimization methods and disregards, to a large extent, the tantamount development of quantum algorithms. In contrast, projective simulation is a reinforcement learning algorithm inspired by the stochastic evolution of physical systems that enables a quantum speed-up in decision making. In this work, we develop a unifying framework based on generative energy-based models which benefits from the advantages of both deep reinforcement learning and projective simulation to solve complex and large-scale decision-making problems.