Offline vs. Online Learning for Deep Driving from Demonstrations

ABOUT THE PROJECT

At a glance

We are studying alternative learning methods for Driving from Demonstrations.  The focus on offline learning dating back almost 30 years has been largely supplanted by online learning methods such as DAgger, which focus on difficult-to-learn areas of state space and require tedious and error-prone corrective feedback from human supervisors. Our initial results suggest that offline learning may be preferable for highly-expressive policy classes such as Deep Learning. 
 

PRINCIPAL INVESTIGATORS

RESEARCHERS

THEMES

Ken Goldberg

 

Deep Learning

 

BDD Update Call

The primary agenda topic is a vehicle update from Ching-Yao Chan and Karl Zipser.

Sponsors and faculty are welcome to join the call.

If you have not already done so, please sign up for a WebEx account, rather than joining the calls as a guest.  It makes it easier for you to receive updates and access the recording afterwards, and easier for us to tell who is actually on the calls.  When you have a request to hear a call after it has happened, I can "invite" you to the call that has passed, only when you have a WebEx account (and tell me the email address to send the invitation to.)

Access Information
Where: WebEx Online
Meeting number: 198 951 241
Password: This meeting does not require a password.
Audio Connection: +1-415-655-0001 US TOLL
Access code: 198 951 241