English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

AlphaZero is a self-play reinforcement learning algorithm that achievessuperhuman play in chess, shogi, and Go via policy iteration. To be aneffective policy improvement operator, AlphaZero's search requires accuratevalue estimates for the states appearing in its search tree. AlphaZero trainsupon self-play matches beginning from the initial state of a game and onlysamples actions over the first few moves, limiting its exploration of statesdeeper in the game tree. We introduce Go-Exploit, a novel search controlstrategy for AlphaZero. Go-Exploit samples the start state of its self-playtrajectories from an archive of states of interest. Beginning self-playtrajectories from varied starting states enables Go-Exploit to more effectivelyexplore the game tree and to learn a value function that generalizes better.Producing shorter self-play trajectories allows Go-Exploit to train upon moreindependent value targets, improving value training. Finally, the explorationinherent in Go-Exploit reduces its need for exploratory actions, enabling it totrain under more exploitative policies. In the games of Connect Four and 9x9Go, we show that Go-Exploit learns with a greater sample efficiency thanstandard AlphaZero, resulting in stronger performance against referenceopponents and in head-to-head play. We also compare Go-Exploit to KataGo, amore sample efficient reimplementation of AlphaZero, and demonstrate thatGo-Exploit has a more effective search control strategy. Furthermore,Go-Exploit's sample efficiency improves when KataGo's other innovations areincorporated.

Targeted Search Control in AlphaZero for Effective Policy Improvement