Smooth Information Gathering in Two-Player Noncooperative Games

In this paper we define a mathematical framework for modeling two-player noncooperative games in which one player is uncertain of the other player’s costs but can preemptively allocate information-gathering resources to reduce this uncertainty. We also provide a gradient-based algorithm to solve a two-stage game, and identify conditions under which the gradient of the first stage’s cost wrt the information-gathering resources, is well-defined. Links: paper, code. ¹

Framework overview.

Proposed algorithm

Conditions for existence of derivative of game's solution wrt. decision variables

Links: paper, code.

Footnotes §

1. Arxiv paper is outdated but I’ll upload the most recent version soon! ↩