Papers & Research

OpenAI trains agent from a single human demonstration to score 74500 on Montezuma's Revenge

Original title:Learning Montezuma’s Revenge from a single demonstration

AI summary

An agent trained by OpenAI starting from only a single human demonstration achieved a high score of 74500 on Montezuma's Revenge, surpassing previously published best results. The algorithm starts continuous play from carefully selected states in the demonstration and uses PPO to optimize game score for learning; PPO is the same reinforcement learning algorithm behind OpenAI Five.

Key Takeaways

  • An agent trained by OpenAI starting from only a single human demonstration achieved a high score…
  • The algorithm starts continuous play from carefully selected states in the demonstration and uses…
  • PPO is the same reinforcement learning algorithm behind OpenAI Five.

Quick Facts

Published
Updated
Source
OpenAI:官网动态(RSS · 排除企业/客户案例)
Category
Papers & Research
AI score
66
Read the original Read the full article at the original source
Why it matters

原文说明从单条人类演示中选取起始状态再用 PPO 优化得分的算法思路,读者可以理解这一强化学习方法的可迁移做法。

Related entities

Source & copyright

This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.

Original article · AIHOT entry

How to Cite

When citing this article, keep the title, date, and permalink, and credit Orange News as the source.

Orange News: OpenAI trains agent from a single human demonstration to score 74500 on Montezuma's Revenge. July 4, 2018. https://new.nat6.net/en/news/h6mlstx26g1p8egfntspy49mp/

View all →