千葉大学国際医工学セミナー

International Seminar Series on Biomedical Engineering

第12回
日時：2011年11月18日（金）16:10 - 19:00
場所：工学部17号棟212教室 （西千葉キャンパス）

---

タイトル

New Methods for Real-time Machine Learning and their Application to Assistive Medical Robotics

演者

Patrick M. Pilarski, PhD

(Postdoctoral Fellow, Department of Computing Science
Reinforcement Learning and Artificial Intelligence Laboratory
Alberta Innovates Centre for Machine Learning)

概要

Machine learning has a major role to play within the domain of rehabilitation and assistive robotics. Perhaps the most significant challenge for future work in this area is the development of natural, flexible control and communication methods to connect users and their assistive devices---for example, methods to control a multi-function artificial limb using data-dense muscle and nerve signals from the human body. This seminar will examine the principal challenges for adaptive, intelligent artificial limbs within the context of real-time machine learning. Specifically, we will discuss recent ideas about the use of General Value Functions (GVFs) as a method for acquiring and maintaining predictive, temporally extended knowledge about the world. These predictions are formed and updated through ongoing, real-time sensorimotor interactions between a learning system and its environment. As a practical example, we will examine the use of GVFs to anticipate online signals in a clinical task involving the human control of a myoelectric training prosthesis.

講師略歴

Patrick M. Pilarski is a Postdoctoral Fellow with the Reinforcement Learning & Artificial Intelligence Laboratory and the Alberta Innovates Centre for Machine Learning, Dept. of Computing Science, at the University of Alberta, Edmonton, Canada. He received the Ph.D. in Electrical and Computer Engineering from the University of Alberta in 2009, and the B.ASc. in Electrical Engineering from the University of British Columbia in 2004. Patrick’s research interests include real-time machine learning, artificial intelligence, human-machine interfaces, robotics, and biomedical pattern analysis. These interests drive Patrick's applied research into assistive rehabilitation robots and lab-on-a-chip systems. His recent work focuses on developing new methods for adaptable, amputee-accessible prosthesis control. Patrick is the author of seventeen articles, and co-inventor of two international patent filings involving machine intelligence and pattern analysis for lab-on-a-chip medical devices. He is the recipient of an NSERC Postdoctoral Fellowship in Artificial Intelligence.