Cooperative Inverse Reinforcement Learning

Add to your list(s) Download to your calendar using vCal

• Robert Pinsler (University of Cambridge)
• Wednesday 25 October 2017, 17:00-18:30
• Cambridge University Engineering Department, CBL Seminar room BE4-38. For directions see http://learning.eng.cam.ac.uk/Public/Directions.

If you have a question about this talk, please contact Adrià Garriga Alonso.

The value alignment problem consists in ensuring the values of an AI system align with the values of its operator. A potential solution to this problem is formalised as the Inverse Reinforcement Learning (IRL) setting. In IRL , the goal is to infer the reward function of an agent (a human), just from observing its behaviour in the environment. In Cooperative IRL , the agents are allowed to interact. From this, more effective teaching strategies than passive observation emerge. We will talk about formalising this problem, and an algorithm to approximate good teaching strategies.

Talk slides: https://valuealignment.ml/talks/2017-10-25-cooperative-inverse-reinforcement-learning.pdf

This talk is part of the Engineering Safe AI series.

This talk is included in these lists:

• Cambridge University Engineering Department, CBL Seminar room BE4-38. For directions see http://learning.eng.cam.ac.uk/Public/Directions
• Cambridge talks
• Chris Davis' list
• Engineering Safe AI
• Trust & Technology Initiative - interesting events

Note that ex-directory lists are not shown.