AlphaZero

Résumé

AlphaZero est une version généraliste d’AlphaGo Zero, un logiciel de go (jeu de stratégie abstrait chinois) qui a été adapté pour jouer aux échecs et au shogi (échecs japonais). AlphaZero a été créé par Demis Hassabis de DeepMind, une entreprise appartenant au groupe Google. Historique Le , DeepMind poste sur la plateforme de prépublication arXiv un article concernant AlphaZero, un programme utilisant l’approche généralisée d'AlphaGo Zero. Le style de jeu d'AlphaZero s'écarte des programmes de jeu habituels tout en requérant moins de calculs par mouvement en regard de ses concurrents. Selon DeepMind, AlphaZero a atteint en 24 heures un niveau de jeu supérieur aux humains au jeu d'échecs, au shogi et au go en battant les programmes champions du monde Stockfish (échecs), (shogi) et la version d’AlphaGo Zero ayant eu trois jours d'apprentissage. Le même mois, AlphaZero bat la version AlphaGo Zero (ayant 3 jours d’apprentissage) 60 parties à 40. Avec 8 heures de pratique et

Source officielle

À propos de ce résultat

Cette page est générée automatiquement et peut contenir des informations qui ne sont pas correctes, complètes, à jour ou pertinentes par rapport à votre recherche. Il en va de même pour toutes les autres pages de ce site. Veillez à vérifier les informations auprès des sources officielles de l'EPFL.

Publications associées

Chargement

Personnes associées

Chargement

Unités associées

Chargement

Concepts associés

Chargement

Cours associés

Chargement

Séances de cours associées

Chargement

, ,

We explore the relationship between polar and RM codes and we describe a coding scheme which improves upon the performance of the standard polar code at practical block lengths. Our starting point is the experimental observation that RM codes have a smaller error probability than polar codes under MAP decoding. This motivates us to introduce a family of codes that "interpolates" between RM and polar codes, call this family C-inter = {C-alpha : alpha is an element of [0, 1]}, where C-alpha vertical bar(alpha=1) is the original polar code, and C-alpha vertical bar(alpha=0) is an RM code. Based on numerical observations, we remark that the error probability under MAP decoding is an increasing function of a. MAP decoding has in general exponential complexity, but empirically the performance of polar codes at finite block lengths is boosted by moving along the family C-inter even under low-complexity decoding schemes such as, for instance, belief propagation or successive cancellation list decoder. We demonstrate the performance gain via numerical simulations for transmission over the erasure channel as well as the Gaussian channel.

Ieee-Inst Electrical Electronics Engineers Inc2014

Chargement

From Polar to Reed-Muller Codes: a Technique to Improve the Finite-Length Performance

Seyed Hamed Hassani, Marco Mondelli, Rüdiger Urbanke

We explore the relationship between polar and RM codes and we describe a coding scheme which improves upon the performance of the standard polar code at practical block lengths. Our starting point is the experimental observation that RM codes have a smaller error probability than polar codes under MAP decoding. This motivates us to introduce a family of codes that "interpolates" between RM and polar codes, call this family C-inter = {C-alpha:alpha is an element of[0;1]}, where C alpha vertical bar(alpha=1) is the original polar code, and C alpha vertical bar(alpha=0) is an RM code. Based on numerical observations, we remark that the error probability under MAP decoding is an increasing function of alpha. MAP decoding has in general exponential complexity, but empirically the performance of polar codes at finite block lengths is boosted by moving along the family C-inter even under low-complexity decoding schemes such as, for instance, belief propagation or successive cancellation list decoder. We demonstrate the performance gain via numerical simulations for transmission over the erasure channel as well as the Gaussian channel.

Ieee2014

Chargement