Sunday Seminar: Abraham Neyman - Robust Optimization, Correlated Equilibrium, and Equilibrium

Mon, 29/11/202116:00-17:55
Elath Hall, 2nd floor, Feldman Building, Edmond Safra Campus
Abraham Neyman, HUJI



 It is well documented that economic agents' behavior is often incompatible with exact optimization. We believe that economic theory has overemphasized exact optimization, especially its implications for economic agents' behavior. The reason is that in most real-life applications, there is some imprecision in the speci_cation of the model, in particular in the speci_cation of the preference over the possible outcomes. Therefore, a rational economic agent may, or even should, forgo exact optimization in a single economic model with well-de_ned parameters, and prefer behavior that is approximately optimal in a whole class of economic models whose di_erences reect imprecision in specifying the parameters of the model. Moreover, it is desirable that the economic agent's behavior exhibits gradual change as the model changes. We call this type of behavior robust optimization. 

It is proved that: (A) a _nite Markov decision process (1) admits a robust optimization when the parameters range over all valuation { linear utilities (over streams of payo_s) that satis_es the time value of money principle, and

(2) does not admit robust optimization if the time value of money is replaced by monotonicity and the weak positive time preference; (B) strategic one shot game admit robust correlated equilibrium; and (C) n-person stochastic games and two-person repeated games with incomplete information on one side admit robust correlated equilibrium, when the payo_s are valuations.


 Refreshments available at 15:45 p.m.