English

General limit value in Dynamic Programming

Optimization and Control 2013-01-04 v1

Abstract

We consider a dynamic programming problem with arbitrary state space and bounded rewards. Is it possible to define in an unique way a limit value for the problem, where the "patience" of the decision-maker tends to infinity ? We consider, for each evaluation θ\theta (a probability distribution over positive integers) the value function vθv_{\theta} of the problem where the weight of any stage tt is given by θt\theta_t, and we investigate the uniform convergence of a sequence (vθk)k(v_{\theta^k})_k when the "impatience" of the evaluations vanishes, in the sense that tθtkθt+1kk0\sum_{t} |\theta^k_{t}-\theta^k_{t+1}| \rightarrow_{k \to \infty} 0. We prove that this uniform convergence happens if and only if the metric space vθk,k1{v_{\theta^k}, k\geq 1} is totally bounded. Moreover there exists a particular function vv^*, independent of the particular chosen sequence (θk)k({\theta^k})_k, such that any limit point of such sequence of value functions is precisely vv^*. Consequently, while speaking of uniform convergence of the value functions, vv^* may be considered as the unique possible limit when the patience of the decision-maker tends to infinity. The result applies in particular to discounted payoffs when the discount factor vanishes, as well as to average payoffs where the number of stages goes to infinity, and also to models with stochastic transitions. We present tractable corollaries, and we discuss counterexamples and a conjecture.

Keywords

Cite

@article{arxiv.1301.0451,
  title  = {General limit value in Dynamic Programming},
  author = {Jérôme Renault},
  journal= {arXiv preprint arXiv:1301.0451},
  year   = {2013}
}
R2 v1 2026-06-21T23:03:23.453Z