Decentralized Learning in Stochastic Games with Local Information
Bibliographic record
Abstract
In the context of multi-agent systems with decentralized information structures, we study rigorously justified convergence results and associated learning algorithms that converge to equilibria. With this objective in mind, we first review classical equilibrium results, focusing on finite-player games with pure or mixed strategy sets. Results such as Kakutani’s fixed-point theorem and Sion’s minimax theorem establish existence under relatively broad conditions. Building on this background, we then study learning dynamics, including best and better response processes, in which players periodically revise and update strategies to optimize payoffs relative to their previous actions via a policy revision process. This induces a graph on the set of policies which facilitate our mathematical approach which combines graph theory, game theory, stochastic control, and Markov processes. While learning using best/better response dynamics converges under certain conditions reported in Arslan et.al, a new approach to policy revision, termed as satisficing (which may be viewed as a win-stay, lose-shift algorithm), introduced by Yongacoglu et.al provides a strictly richer graph network structure and is applicable to a much broader class of games. In particular, these generalize weakly acyclic games. The question we studied is to precisely characterize the set of games for which such a satisficing process ensures convergence to equilibrium. In particular, we addressed an open question raised by Yongacoglu et al. on necessary and sufficient conditions for convergence to equilibria from any initial policy profile. On sufficiency, we presented a generalization, relaxing requirements to allow multiple pure Nash equilibria, provided at least one is strict and subgame-unique. Our research also presented a nontrivial example of a game that admits a strict pure Nash equilibrium in each induced subgame that fails to converge via satisficing paths, showing that such conditions are insufficient, thus also leading to a necessity condition.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.014 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.003 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".