Dynamics of Competitive Learning: The Role of Updates and Memory

Abstract

We analyze how parallel versus sequential updating and synaptic memory affect phases of consensus and polarization in a competitive learning model. The phase diagram exhibits ordered and disordered regimes, with critical exponents matching the voter model universality class.

Methodology

We study a two-strategy spin-like model on a lattice, applying all combinations of parallel and sequential updates. Success rates over a memory window govern strategy updates. Monte Carlo simulations and finite-size scaling determine phase boundaries and critical behavior.

title: “Dynamics of Competitive Learning: The Role of Updates and Memory” pub_id: “2012-1-dynamics-competitive-learning-updates-memory” image: “https://images.pexels.com/photos/17808485/pexels-photo-17808485.jpeg” authors: “Bhat A. A. & Mehta A.” journal: “Physical Review E” year: 2012 link: “https://doi.org/10.1103/PhysRevE.85.011134” video: “” tags: [“plasticity”,”memory”,”criticality”,”competition”,”learning”] abstract: “We analyze how parallel versus sequential updating and synaptic memory affect phases of consensus and polarization in a competitive learning model. The phase diagram exhibits ordered and disordered regimes, with critical exponents matching the voter model universality class.” methodology: “We study a two-strategy spin-like model on a lattice, applying all combinations of parallel and sequential updates. Success rates over a memory window govern strategy updates. Monte Carlo simulations and finite-size scaling determine phase boundaries and critical behavior.” date: 2012-01-01 layout: publication —

We analyze how parallel versus sequential updating and synaptic memory affect phases of consensus and polarization in a competitive learning model. The phase diagram exhibits ordered and disordered regimes, with critical exponents matching the voter model universality class.