ANon-asymptotic Analysisof Non-parametric Temporal-Difference Learning

Theorem 1.Let n 9. Underassumption(A2) with 1 < 1, thereexistapositivereal number independentofnsuchthat, for 0 , (a) Using = 0n Also, simplecomputationsshowthatV is anaffinetransformofr: V (x)= ar(x)+ b, witha =( 1 (1 ")) 1 andb = a Wealsoacknowledgesupport fromthe European Research Council (gran...