Mahjax: New JAX-based Mahjong AI Simulator Enables Large-Scale RL Research
Researchers have introduced Mahjax, a GPU-accelerated Mahjong simulator built in JAX, designed to drive large-scale reinforcement learning (RL) from scratch in complex games with imperfect information.
What happened?
A new research paper presents Mahjax, a fully vectorised Riichi Mahjong environment implemented in JAX. The objective is to enable large-scale parallelisation of rollouts on graphics cards (GPUs). The simulator is intended for reinforcement learning research and also includes a visualisation tool for debugging and interacting with trained AI agents.
Key facts
| Simulatornamn | Mahjax |
|---|---|
| Ramverk | JAX |
| Spel | Riichi Mahjong |
| Maximal genomströmning | 2 miljoner rollouts/s |
| Klassificering | cs.AI (Artificiell Intelligens) |
”Riichi Mahjong is a multi-player, imperfect-information game characterized by stochasticity and high-dimensional state spaces. These attributes present a unique combination of challenges that mirror complex real-world decision-making problems in reinforcement learning.”
”To facilitate such research, we introduce Mahjax, a fully vectorized Riichi Mahjong environment implemented in JAX to enable large-scale rollout parallelization on Graphics Processing Units (GPUs).”
”Experimental results demonstrate that Mahjax achieves throughputs of up to 2 million”
Why it matters
Mahjong is a multi-player game featuring incomplete information, stochasticity, and high-dimensional state spaces. These characteristics represent a challenging combination similar to complex decision-making problems in the real world. By enabling training from scratch, similar to AlphaZero, Mahjax can contribute to more generally applicable AI development beyond mere supervised learning from human game logs.
Who is affected?
Researchers and developers within artificial intelligence, particularly those working with reinforcement learning and complex game environments with imperfect information, are directly affected. Those interested in JAX-based applications and high-performance GPU computing for AI research are also concerned.
What else you should know
Mahjax achieves a throughput of up to 2 million rollouts per second, underlining its capability to handle large-scale training efficiently.
Quick answers about this story
Vad har hänt?
När hände det?
Varför spelar det roll?
Vilka tekniker används?
The link opens in a new window and leads to the publisher's own site.
Källan har spårats automatiskt från utgivaren via Aheadlines signalkedja.
AI-verktyg i artikeln
Topics
Get similar news straight to your inbox
The reader's room
Send in a question or an addition. The newsroom reads everything before it's published and replies when relevant. No AI-generated text – just people.
Sign in to submit a comment or question.
Read the article through your role
- Decide whether this affects strategy over 6–12 months or is just noise.
- Discuss with leadership: do we own the right question or does ownership need to move?
- Ask: what risk are we taking by NOT acting on this this quarter?
Generated angle — not editorial analysis of "Mahjax: New JAX-based Mahjong AI Simulator Enables Large-Sca"