We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Multi-agent Performative Prediction: From Global Stability and Optimality to Chaos

Abstract: The recent framework of performative prediction is aimed at capturing settings where predictions influence the target/outcome they want to predict. In this paper, we introduce a natural multi-agent version of this framework, where multiple decision makers try to predict the same outcome. We showcase that such competition can result in interesting phenomena by proving the possibility of phase transitions from stability to instability and eventually chaos. Specifically, we present settings of multi-agent performative prediction where under sufficient conditions their dynamics lead to global stability and optimality. In the opposite direction, when the agents are not sufficiently cautious in their learning/updates rates, we show that instability and in fact formal chaos is possible. We complement our theoretical predictions with simulations showcasing the predictive power of our results.
Subjects: Machine Learning (cs.LG); Computer Science and Game Theory (cs.GT); Systems and Control (eess.SY)
Cite as: arXiv:2201.10483 [cs.LG]
  (or arXiv:2201.10483v1 [cs.LG] for this version)

Submission history

From: Fang-Yi Yu [view email]
[v1] Tue, 25 Jan 2022 17:26:12 GMT (933kb,D)

Link back to: arXiv, form interface, contact.