Details of Grant 

EPSRC Reference:	GR/H80217/01
Title:	INCREMENTAL MACHINE LEARNING AND ADAPTIVE NEURAL NETWORK ARCHITECTURES
Principal Investigator:	Niranjan, Professor M
Other Investigators:	Fallside, Professor F
Researcher Co-Investigators:
Project Partners:
Department:	Engineering
Organisation:	University of Cambridge
Scheme:	Standard Research (Pre-FEC)
Starts:	01 December 1992	Ends:	31 March 1996	Value (£):	106,341
EPSRC Research Topic Classifications:
EPSRC Industrial Sector Classifications:
Related Grants:
Panel History:
Summary on Grant Application Form
The aim of this project is to study neural computing algorithms that are suitable for on-line or sequential problems that arise in applications such as time series analysis and real time control. We address, using the Radial Basis Function class of models,(a) evaluation of incremental training algorithms on a number of benchmark real world tasks,(b) the development of neural computing architectures that adapt to suit the complexity of the problem,(c) study the parallel between this class of algorithms and models of learning in statistical mechanics and(d) investigate parallel computing architectures suitable for implementing this class of algorithms.Progress:The project has made substantial progress and has met several of its objectives. Experimental evaluation of incremental learning algorithms have been completed on benchmark problems in time series analysis and pattern classification. An approach to prune large networks, based on our previous work on F-projections (or sequential projections in a function space) has been developed, implemented and evaluated. Work on the relationship of this approach to learning models in the statistical physics literature has also made substantial progress.Work on this project has also influenced the Investigators research on two areas. We now have a method of incrementally training multi layer perceptron (MLP) classifiers. In a related study, we have also looked at the Bayesian interpretation of this project. This leads to the ability to use multiple neural computing models online and recursively evaluate their relative performances. We were able to show that this allows model selection and/or combining multiple models to achieve better performances.
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.cam.ac.uk