Search |  Contact |  SRI Home Do not follow this link, or your host will be blocked from this site. This is a spider trap. Do not follow this link, or your host will be blocked from this site. This is a spider trap. Do not follow this link, or your host will be blocked from this site. This is a spider trap.A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A ASRI International.  333 Ravenswood Avenue.  Menlo Park, CA 94025-3493. SRI International is a nonprofit corporation.

AIC Seminar Series

Recent Directions in Machine Learning and AI

Sridhar MahadevanArtificial Intelligence Center, SRI International[Home Page]

Notice:  Introductory talk by the new Lab Director

Date:  Thursday, February 23rd 2017 at 4:00pm

Location:  EK255 (SRI E building)  (Directions)

Webex: 

WebEx recording (SRI WebEx access required):

https://sri-meetings.webex.com/sri-meetings/ldr.php?RCID=7f6b34151df82060191fc72ae9e99820

Slides (SRI Wiki access required):

https://wiki.sri.com/download/attachments/4919820/SRI%20AIC%20Feb%2024%202017%20Sridhar.pdf?version=1&modificationDate=1489786494167&api=v2
   Abstract

In this talk, I will briefly summarize recent trends in the field of machine learning and AI, drawing on examples from my recent research.

1) A longstanding approach to machine learning has been to view it as a geometric problem in Euclidean vector spaces. I will discuss how non-Euclidean manifold representations have had considerable success in machine learning, and illustrate manifold learning methods with two applications: analogical reasoning with continuous word representations, and transfer learning on Mars using data from Curiosity.

2) Gradient techniques have been used in machine learning for over five decades. I will describe a powerful new class of gradient methods, where weight updates are carried out not in the original space, but rather in a carefully chosen dual space. I will relate these new methods to concepts in optimization, such as proximal mappings and Bregman divergences. I will show how the use of dual space gradient methods led to the solution of a three-decade open problem in reinforcement learning on designing stable temporal difference methods.

3) Finally, since the 21st century increasingly involves cloud computing, I will briefly sketch out a vision for how many problems in AI and machine learning can be reformulated using a powerful equilibration framework, which generalizes the classical optimization based approach. I will begin with some new work that my students and I have just published on a deep learning model called GMAN (generalized multi-adversarial network), and then go over some examples of the use of this framework for next-generation Internet modeling, supply chain manufacturing, and health science applications.

   Bio for Sridhar Mahadevan

Professor Mahadevan (Ph.D., Computer Science, Rutgers University 1990) has been a professor at the College of Information and Computer Sciences at the University of Massachusetts, Amherst since 2001. His research interests span several subfields of artificial intelligence and computer science, including machine learning, multi-agent systems, planning, perception, and robotics.

   Note for Visitors to SRI

Please arrive at least 10 minutes early as you will need to sign in by following instructions by the lobby phone at Building E (or call Wilma Lenz at 650 859 4904, or Vicenta at Lopez at 650 859 5750). SRI is located at 333 Ravenswood Avenue in Menlo Park. Visitors may park in the parking lots off Fourth Street. Detailed directions to SRI, as well as maps, are available from the Visiting AIC web page. There are two entrances to SRI International located on Ravenswood Ave. Please check the Building E entrance signage.

SRI International
©2017 SRI International 333 Ravenswood Avenue, Menlo Park, CA 94025-3493
SRI International is an independent, nonprofit corporation. Privacy policy