Connectionists: CFP - NIPS 2005 Workshop - Inductive Transfer : 10 Years Later

[Daniel L. Silver]

NIPS 2005 Workshop - Inductive Transfer : 10 Years Later 

---------------------------------------------------------

Friday, Dec 9, Westin Resort and Spa in Whistler, British Columbia, Canada

 

Overview:

---------

 

Inductive transfer refers to the problem of applying the knowledge learned
in one or more 

tasks to learning for a new task. While all learning involves generalization
across problem 

instances, transfer learning emphasizes the transfer of knowledge across
domains, tasks, 

and distributions that are related, but not the same.  For example, learning
to recognize 

chairs might help to recognize tables; or learning to play checkers might
improve learning 

of chess.  While people are adept at inductive transfer, even across widely
disparate domains, 

currently we have little learning theory to explain this phenomena and few
systems exhibit

knowledge transfer.

 

At NIPS95 two of the current co-chairs lead a successful two-day workshop on
"Learning to Learn" 

that focused on lifelong machine learning methods that retain and reuse
learned knowledge.  

(The co-organizers of the NIPS95 workshop were Rich Caruana, Danny Silver,
Jon Baxter, 

Tom Mitchell, Lorien Pratt, and Sebastian Thrun.)  The fundamental
motivation for that meeting 

was the belief that machine learning systems would benefit from re-using
knowledge learned 

from related and/or prior experience and that this would enable them to move
beyond 

task-specific tabula rasa systems.  The workshop resulted in a series of
articles published in a 

special issue of Connection Science [CS 1996], Machine Learning [vol. 28,
1997] and a book 

entitled  "Learning to Learn" [Pratt and Thrun 1998].

 

Research in inductive transfer has continued since 1995 under a variety of
names: 

learning to learn, life-long learning, knowledge transfer, transfer
learning, multitask learning,

knowledge consolidation, context-sensitive learning, knowledge-based
inductive bias, 

meta-learning, and incremental/cumulative learning.  The recent burst of
activity in this 

area is illustrated by the research in multi-task learning within the kernel
and Bayesian 

contexts that has established new frameworks for capturing task relatedness
to improve learning

[Ando and Zhang 04, Bakker and Heskes 03, Jebara 04, Evgeniou, and Pontil
04, 

Evgeniou, Micchelli and Pontil 05, Chapelle and Harchaoui 05].  This NIPS
2005 workshop will 

examine the progress that has been made in ten years, the questions and
challenges that 

remain, and the opportunities for new applications of inductive transfer
systems.

 

In particular, the workshop organizers have identified three major goals:

(1) To summarize the work thus far in inductive transfer to develop a
taxonomy of research and 

highlight open questions,

(2) To share new theories, approaches, and algorithms regarding the
accumulation and re-use of 

learned knowledge to make learning more effective and more efficient,

(3) To discuss the formation of an inductive transfer special interest group
that might offer 

a website, benchmarking data, shared software, and links to various research
programs and 

related web resources.

 

Call for Papers:

----------------

We invite submission of workshop papers that discuss ongoing or completed
work dealing with 

Inductive Transfer (see below for a list of appropriate topics).  Papers
should be no more 

than four pages in the standard NIPS format.  Authorship should not be
blind.  

Please submit a paper by emailing it in Postscript or PDF format to
danny.silver at acadiau.ca with 

the subject line "ITWS Submission". We anticipate accepting as many as 8
papers for 15 minute 

presentation slots and up to 20 poster papers.  Please only submit an
article if at least one 

of the authors will attend the workshop to present the work.

 

The successful papers will be made available on the Web.  A special journal
issue or an edited 

book of selected papers also is being planned.

 

The 1995 workshop identified the most important areas for future research to
be:

* The relationship between computational learning theory and selective
inductive bias;

* The tradeoffs between storing or transferring knowledge in
representational and functional form;

* Methods of turning concurrent parallel learning into sequential lifelong
learning methods;

* Measuring relatedness between learning tasks for the purpose of knowledge
transfer;

* Long-term memory methods and cumulative learning; and

* The practical applications of inductive transfer and lifelong learning
systems. 

The workshop is interested in the progress that has been made in these areas
over the last ten 

years.  These remain key topics for discussion at the proposed workshop. 

 

More forward looking and important questions include:

* Under what conditions is inductive transfer difficult? When is it easy?

* What are the fundamental requirements for continual learning and transfer?

* What new mathematical models/frameworks capture/demonstrate transfer
learning? 

* What are some of latest and most advanced demonstrations of transfer
learning in machines 

(Bayesian, kernel methods, reinforcement)? 

* What can be learned from transfer learning in humans and animals? 

* What are the latest psychological/neurological/computational theories of
knowledge transfer 

in learning?

 

Important Dates:

----------------

19 Sep 05 - Call for participation

21 Oct 05 - Paper submission deadline

04 Nov 05 - Notification of paper acceptance

09 Dec 05 - Workshop in Whistler

 

Organizers:

--------------

Danny Silver, Jodrey School of Computer Science, Acadia University, Canada 

Rich Caruana, Department of Computer Science, Cornell University, USA 

Stuart Russell, Computer Science Division, University of California,
Berkeley, USA 

Prasad Tadepalli, School of Electrical Eng. and Computer Science, Oregon
State University, USA 

Goekhan Bakir, Max Planck Institute for Biological Cybernetics, Germany 

Kristin Bennett, Department of Mathematical Sciences, Rensselaer Polytechnic
Institute, USA 

Massimiliano Pontil, Dept. of Computer Science, University College London,
UK

 

For further Information:

------------------------

Please see the workshop webpage at http://iitrl.acadiau.ca/itws05/ 

Email danny.silver at acadiau.ca

 

===============================================

Daniel L. Silver, Ph.D.        <mailto:danny.silver at acadiau.ca>
danny.silver at acadiau.ca

Associate Professor          p:902-585-1105 f:902-585-1067

Intelligent Information Technology Research Laboratory

Jodrey School of Computer Science, Office 315

Acadia University, Wolfville, NS  B4P 2R6