Go-Explore: Solving Hard Exploration Problems
Go-Explore uses a two-step approach: Go (return to previous saved state), then Explore (explore from that state).
It manages to solve the hardest RL problems in the Atari domain, mainly Montezuma's Revenge, and reaches SOTA score for all other Atari games with the right abstraction.
Links to paper:
Go-Explore: https://arxiv.org/pdf/1901.10995.pdf
First return, then explore (follow-up on Go-Explore with goal-conditioned rewards): https://arxiv.org/pdf/2004.12919.pdf
0:00 Introduction
3:37 Montezuma’s Revenge Video
6:34 Intrinsic Motivation approaches
13:41 Detachment
15:52 Derailment
19:50 Torchlight Analogy
24:00 Go-Explore Overview
33:28 Updating the Memory Archive
35:34 Example
41:36 Robustify Phase
44:25 Discussion
Online AI blog: https://delvingintotech.wordpress.com/.
LinkedIn: https://www.linkedin.com/in/chong-min-tan-94652288/
Twitch: https://www.twitch.tv/johncm99
Twitter: https://twitter.com/johntanchongmin
Try out my games here: https://simmer.io/@chongmin
Other Videos By John Tan Chong Min
Other Statistics
Torchlight Statistics For John Tan Chong Min
There are 56 views in 1 video for Torchlight. About an hours worth of Torchlight videos were uploaded to his channel, less than 0.37% of the total video content that John Tan Chong Min has uploaded to YouTube.