22nd August 2019 in London at CodeNode

This SkillsCast was filmed at London Reinforcement Learning August

This talk with focus on Tangled Program Graphs playing Minecraft. TPGs are a form of genetic programming that has had recent success in visual RL tasks such as ALE, Dota2 and VizDoom. TPGs offer an number of advantages over deep learning techniques:

1) They can learn to play multiple games at the same time and efficiently reuse elements of their graph structure across all games

2) They require a fraction of the computational resource - No GPUs required!

Come to the talk to learn about TPGs, genetic programming and see an implementation of TPG learning to play Minecraft,

YOU MAY ALSO LIKE:

Thanks to our sponsors

Genetic Programming for RL on Minecraft

Matt Lee

Matt is currently finishing an MSc in Artificial Intelligence from QMUL. He has previous experience working in DevOps and Big Data. He is interested in Computer Vision and Reinforcement Learning.

SkillsCast

This talk with focus on Tangled Program Graphs playing Minecraft. TPGs are a form of genetic programming that has had recent success in visual RL tasks such as ALE, Dota2 and VizDoom. TPGs offer an number of advantages over deep learning techniques:

1) They can learn to play multiple games at the same time and efficiently reuse elements of their graph structure across all games

2) They require a fraction of the computational resource - No GPUs required!

Come to the talk to learn about TPGs, genetic programming and see an implementation of TPG learning to play Minecraft,

YOU MAY ALSO LIKE:

Thanks to our sponsors

About the Speaker

Genetic Programming for RL on Minecraft

Matt Lee

Matt is currently finishing an MSc in Artificial Intelligence from QMUL. He has previous experience working in DevOps and Big Data. He is interested in Computer Vision and Reinforcement Learning.