Monday, Nov 2nd: Computational Necromancy with Reinforcement Learning
Who: Dr. Bill Smart, Associate Professor of Computer Science, WUSTL
When: Monday, Nov 2nd at 3pm
Where: EB 1012
What: Animating the Dead: Computational Necromancy with Reinforcement Learning
Anthropologists are often interested in questions such as “What do
chimpanzees optimize for when they walk, and when they climb?”, “What
would Lucy (Australopithecus Afarensis) look like when it walked?”, and
“How did the modern human musculoskeletal system evolve from its
evolutionary ancestors?”. All of these problems can be cast as learning
a controller for a complex biomechanical system that optimizes some
criteria. Reinforcement learning offers a set of techniques to address
this problem formulation, if we can overcome the problems of learning in
a high-dimensional, non-linear system, continuous in states, actions,
and time.
In this talk, I will outline the types of biomechanics problems that
anthropologists are interested in, and show how reinforcement learning
can be used to address them. In particular, I will describe how
Differential Dynamic Programming can be used to learn controllers for
complex musculoskeletal simulations of humans, chimpanzees, and extinct
hominins, and how these learned controllers can be used to answer
fundamental questions in physical anthropology.
If you enjoyed this post, please consider to leave a comment or subscribe to the feed and get future articles delivered to your feed reader.