The Setting
In the Parisi formula at zero temperature for the Sherrington-Kirkpatrick model, with covariance , the ground-state energy is the infimum over order parameters of the Parisi functional . Here is the value of a stochastic control problem in which a Brownian path started from at time is steered by a bounded drift and rewarded by its distance from the origin at time .
The terminal payoff has a corner at the origin. The question is how regular is as a function of the position before the terminal time, and how much moves when is perturbed in .
Order Parameters and the Control Problem
The time interval is the half-open interval .✦ An order parameter is a function from to that is nonnegative, nondecreasing, right-continuous and integrable.✦ The extension by zero of is the function on equal to on and to elsewhere; integrals of over intervals of are integrals of this extension.✦ The distance of two order parameters and is (1)✦
A Brownian system on a measurable space is a probability measure on together with a standard Brownian motion under ; we write for the expectation under .✦ For a time , the increment filtration of after is the filtration in which is the -algebra generated by the increments with .✦ A bounded control from time is a real process on , indexed by the time elapsed since , that is progressively measurable for and satisfies for all and .
Fix a Brownian system and an order parameter . For a time , a position and a bounded control from time , the control payoff is the random variable (2) The value function of is the function on given by (3)✦ the supremum running over all bounded controls from time . At a negative time , (2) and (3) are read with replaced by and with controls from time .
The Claim
Theorem 1. Let be a Brownian system, and let and be order parameters with value functions and . For every time , the function is even: (4)✦ For every time , the function is convex and -Lipschitz on : (5)✦ (6)✦ For every time and every , (7)✦ For every with and every order , there is a constant such that the -th derivative of in satisfies (8)✦ For every time and every , the value functions of and satisfy (9)✦
In the paper, the result reads: is even, convex and -Lipschitz in , with ; for every and , is bounded and jointly continuous on , uniformly over of bounded norm; and if in , then uniformly on for every and .
What Is Checked
The claim is stated over the value function (3), the supremum of the expected control payoff (2) over bounded controls progressively measurable for the increment filtration of a Brownian system, for order parameters on the time interval extended by zero, and over the distance (1).