Une IA programmée avec un Agent "curiosité" - Curiosity-driven Exploration by Self-supervised Prediction

Par Deepak Pathak, Pulkit Agrawal, Alexei A. Efros, Trevor Darrell @ University of California, Berkeley

Programmer un genre de "sentiment" de curiosité (appelé Intrinsic Curiosity Module (ICM)) à un programme d'IA pour qu'il essaye de se débrouiller dans Mario (et VizDoom).
Je n'irai pas aussi loin que Presse-citron, ce programme n'a pas appris à jouer, juste à essayer d'aller plus loin.  Mais si "bloqué" ou "mort" plusieurs fois dans le jeu au même endroit, il laisse actuellement tomber =)

Extrait du "paper":

pathak22.github.io_noreward-rl-txt.jpg
 

Source: Presse-citron "Une intelligence artificielle a appris à jouer à “Mario Bros” seule, grâce à sa curiosité"

Curiosity-driven Exploration by Self-supervised Prediction

pathak22.github.io_noreward-rl.jpg
pathak22.github.io/noreward-rl

La vidéo:

Crédits:

15 mai 2017

This video summarizes our ICML (International Conference on Machine Learning) 2017 paper on curiosity. More details about the project can be found at: http://pathak22.github.io/noreward-rl/

Project Abstract: In many real-world scenarios, rewards extrinsic to the agent are extremely sparse, or absent altogether. In such cases, curiosity can serve as an intrinsic reward signal to enable the agent to explore its environment and learn skills that might be useful later in its life. We formulate curiosity as the error in an agent's ability to predict the consequence of its own actions in a visual feature space learned by a self-supervised inverse dynamics model. Our formulation scales to high-dimensional continuous state spaces like images, bypasses the difficulties of directly predicting pixels, and, critically, ignores the aspects of the environment that cannot affect the agent. The proposed approach is evaluated in two environments: {\em VizDoom} and {\em Super Mario Bros}. Three broad settings are investigated: 1) sparse extrinsic reward, where curiosity allows for far fewer interactions with the environment to reach the goal; 2) exploration with no extrinsic reward, where curiosity pushes the agent to explore more efficiently; and 3) generalization to unseen scenarios (e.g. new levels of the same game) where the knowledge gained from earlier experience helps the agent explore new places much faster than starting from scratch.

Audio Credits:
(a) https://www.youtube.com/watch?v=1WAaq...
(b) https://www.youtube.com/watch?v=R4TPT...
(c) http://www.bensound.com

Haut de page