Similarity-based transfer learning of decision policies

Abstract
A problem of learning decision policy from past experience is considered. Using the Fully Probabilistic Design (FPD) formalism, we propose a new general approach for finding a stochastic policy from the past data.
View on arXivComments on this paper