Hadamax Encoding: Elevating Performance in Model-Free Atari

Jacob E. Kooi, Zhao Yang, Vincent François-Lavet · 2025

Read Hadamax Encoding: Elevating Performance in Model-Free Atari by Jacob E. Kooi, Zhao Yang, Vincent François-Lavet (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-15345.txt.

Description

Jacob Kooi, Zhao Yang, and Vincent Francois-Lavet present a new encoder for game-playing AI, in a 2025 paper. In reinforcement learning, network design has stayed simple, since changes often gain little. Their encoder works on raw pixels from classic Atari games. It needs no model of the game world.

The encoder pools the largest values from products of two parallel layers. Built on a recent learning method, it set new best scores on the full 57-game Atari set. It beat the base method by 80 percent and topped a well-known rival by a wide margin. No learning settings were changed, so the gain comes from the design alone. The team shares the full code online.

Catalog record

CREATORJacob E. Kooi, Zhao Yang, Vincent François-Lavet
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE41 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-15345

Browse the full library · About the Tower of Babel