Diverse Exploration via Conjugate Policies for Policy Gradient Methods

State-Dependent Exploration for Policy Gradient Methods

Policy Gradient methods are model-free reinforcement learning algorithms which in recent years have been successfully applied to many real-world problems. Typically, Likelihood Ratio (LR) methods are used to estimate the gradient, but they suffer from high variance due to random exploration at every time step of each training episode. Our solution to this problem is to introduce a state-depende...

Multilevel Conjugate Gradient Methods

Application of frames in Chebyshev and conjugate gradient methods

‎Given a frame of a separable Hilbert space $H$‎, ‎we present some‎ ‎iterative methods for solving an operator equation $Lu=f$‎, ‎where $L$ is a bounded‎, ‎invertible and symmetric‎ ‎operator on $H$‎. ‎We present some algorithms‎ ‎based on the knowledge of frame bounds‎, ‎Chebyshev method and conjugate gradient method‎, ‎in order to give some‎ ‎approximated solutions to the problem‎. ‎Then we i...

Conjugate Gradient Methods for Toeplitz Systems

A list of technical reports, including some abstracts and copies of some full reports may be found at: Object test coverage using finite state machines. September 1995. On balancing workload in a highly mobile environment. August 1995. Error analysis of a partial pivoting method for structured matrices. June 1995. Abstract In this expository paper, we survey some of the latest developments on u...

Accurate conjugate gradient methods for shifted systems

We present an efficient and accurate variant of the conjugate gradient method for solving families of shifted systems. In particular we are interested in shifted systems that occur in Tikhonov regularization for inverse problems since these problems can be sensitive to roundoff errors. The success of our method in achieving accurate approximations is supported by theoretical arguments as well a...