Others involved:
Start date: 1985
Whilst neural networks are very well known as a method of machine learning, there are a variety of other methods that are not so well known and in many respects are superior. There is a lot more to learning than the task of classification done by neural networks, and the Contextual Memory System (CMS) addresses the important question of how one can model the human learning of a new subject with no prior knowledge, often known as the creation of new terms. Nearly all other machine learning models use some form of built-in features, whereas the CMS uses built-in feature creation methods to enable the information to be learned to be represented in a completely open and dynamic manner. The CMS is able to model the declarative learning of information and ensures that subsequent retrieval is more efficient by reorganising the knowledge representation through the experience of recall.
Most of the research to date has been using the CMS to represent mathematical knowledge in the Mathematics Understander (MU), but the system is designed to be used for other types of knowledge including linguistic and pictorial data. There are a number of further possible research projects in this area including experimental validation of the cognitive model of learning, application of the CMS to new domains, using the CMS as a student model in intelligent tutoring systems and computational learning experiments.
This work was awarded the prize for the best presented paper by the British AI society, AISB, in 1993.
Page by Edmund Furse
Last updated 15/April/96