Abstract
In this talk, I will revisit the problem of universal portfolio selection first proposed by Cover in 1991. This is a basic problem in information theory with several operational interpretations and connections to, among others, data compression and universal prediction. I will talk about recent work on a new portfolio selection strategy that adapts to a continuous side-information sequence, with a universal wealth guarantee against a class of state-constant rebalanced portfolios with respect to an unknown state function that maps each side-information symbol to a finite set of states.