This paper introduces the generalized forward-backward splitting algorithm for minimizing convex functions of the form F + ∑i=1Gi, where F has a Lipschitzcontinuous gradient and the Gi’s are simple in the sense that their Moreau proximity operators are easy to compute. While the forward-backward algorithm cannot deal with more than n = 1 non-smooth function, our method generalizes it to the cas...

Goals for reinforcement learning problems are typically defined through handspecified rewards. To design such problems, developers of learning algorithms must inherently be aware of what the task goals are, yet we often require agents to discover them on their own without any supervision beyond these sparse rewards. While much of the power of reinforcement learning derives from the concept that...

Computing the time evolution of quantum systems composed of many coupled degrees of freedom presents serious challenges. Currently, simulations of timedependent properties in polyatomic molecules, clusters, or condensed phases employ a variety of approximations, which are largely built around variational, quantum-classical or semiclassical ideas. While approximate methods have proven extremely ...