Introduction to cognitive neural networks.
Neural networks will be placed into a historical perspective related to symbolic approaches and in the context of the artificial intelligence hypothesis. Students will familiarise themselves with the Leabra environment.
The individual neuron.
The idea of the components of a neuron as a 'detector' will be developed. Neural networks will be explained in terms of the biology of the brain at a cellular electro-transmission level. The neurobiology will be abstracted into an initial neural network framework, i.e. a set of mathematical equations. Single neuron simulations.
Networks of Neurons.
A general framework will be provided for neural network architectures both at an abstract level and in terms of networks in the cortex. Unidirectional (feedforward) and bi-directional (recurrent) interactions will be explained together with inhibitory mechanisms.
Model Learning.
A simple Hebbian model of learning will be outlined, pertaining to neurobiology and neural networks. Other models of unsupervised learning will be introduced.
Task Learning.
Error-driven task learning will be outlined; the delta rule and back propagation will be presented. A discussion of the biological implausibility of back propagation networks will follow. Motivated by this implausibility, the generalised recirculation algorithm will be introduced and its mathematical formulation and properties discussed.

70% Exam, 30% Coursework

• See http://readinglists.kent.ac.uk/

See the library reading list for this module

• a) Describe what is meant by neural networks, list a number of types of network and give a brief description of each together with some examples of their (actual or potential) applications.
• b) Select the appropriate neural network paradigm for a particular problem and be able to justify this choice based on knowledge of the properties and potential of this paradigm. To be able to compare the general capabilities of a number of such paradigms and give an overview of their comparative strengths and weaknesses
• c) Explain the mathematical equations that underlie neural networks, both the equations that define activation transfer and those that define learning
• d) Analyse cognitive and neurobiological phenomena from the point of view of their being computational systems. To be able to take these phenomena and identify the features which are important for computational problem solving.
• e) Simulate and understand neural networks using state of the art simulation technology and apply these networks to the solution of problems. In particular, to select from the canon of learning algorithms which is appropriate for a particular problem domain
• f) Discuss examples of computation applied to neurobiology and cognitive psychology, both in the instrumental sense of the application of computers in modelling and in the sense of using computational concepts as a way of understanding how biological and cognitive systems function.
• g) To have a detailed knowledge of an advanced specialised topic in cognitive neural networks. Furthermore, the student should be able to explain the key details of one or more of these specialised topics.
• h) To have the capacity to engage with the research literature in Computational Neuroscience.

None