Details of Grant 

EPSRC Reference:	GR/R12251/01
Title:	Tractable Stochastic Computation
Principal Investigator:	Williams, Professor C
Other Investigators:
Researcher Co-Investigators:
Project Partners:
Department:	Sch of Informatics
Organisation:	University of Edinburgh
Scheme:	Standard Research (Pre-FEC)
Starts:	26 November 2001	Ends:	25 February 2005	Value (£):	147,896
EPSRC Research Topic Classifications:	New & Emerging Comp. Paradigms
EPSRC Industrial Sector Classifications:	Information Technologies
Related Grants:
Panel History:
Summary on Grant Application Form
The proposal deals with the issue of intractability in large graphical models. Success in this is crucial to the wider application, development and understanding of processing based on networks of stochastic, non-linear units. Models based on this paradigm are widespread appearing, for example, as Hidden Markov Models in speech recognition, belief networks in medical diagnosis and sparsely connected graphs in cryptography. The methodologies proposed are inspired by statistical mechanics. (i) The development and deeper understanding of consistency techniques such as belief propagation, with potential applications in cryptography and other large, sparsely connected systems, (ii) The use of renormalisation type techniques in graphical models. This is also a type of consistency technique, but is, in principal, more powerful than belief propagation. (iii) Most graphical models are currently trained using maximum likelihood type techniques. We argue that this criterion may itself be the cause of intractability, since the likelihood contains a summation over all hidden nodes. We propose to study alternative methods for training large graphical models, such as average reconstruction error.Ultimately, the project aims to push artificial intelligence into the next phase - it's application to large scale systems. In nature, the occurrence of a great many interacting, non-linear units, is widespread, from neurons in the brain to ants in a colony. Understanding and developing tools for large scale artificial systems remains a great and promising open challenge.
Key Findings
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Potential use in non-academic contexts
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Impacts
Description This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk Summary Date Materialised
Sectors submitted by the Researcher
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
Project URL:
Further Information:
Organisation Website:	http://www.ed.ac.uk