Outcome Weighted Learning in Dynamic Treatment Regimes

This chapter discusses applications of information geometry in a paradigm of reinforcement learningReinforcement learning with emphasis on dynamic treatment regimesDynamic treatment regimes which have progressed recently in the learning algorithm with outcome weighed learning. The probabilistic framework for a triple expressing state, action, and reward is formulated in multiple stages, in which a decision functionDecision function defined by a state and action is estimated to make an optimal policy for a given dataset. Decision consistency for a decision functionDecision function is introduced by the state-value function in the space of all the decision functionsDecision function. We introduce the $$\varPsi $$ Ψ -divergence on the decision functionDecision function space with a generator function $$\varPsi $$ Ψ , and investigate statistical properties for the $$\varPsi $$ Ψ -loss function $$\varPsi $$ Ψ -loss function conducted by $$\varPsi $$ Ψ -divergence. An outcome-weighed learning algorithm for the decision functionDecision function is considered in a boosting approach in association with the prediction in supervised learning.