In 2009 I was presented with two awards for a two year project carried out between RDA and The University of Reading. The two awards were: The Best Application of Management or Social Science and Best Partnership award for North West England. The project centred around the modernisation of a charity and bringing around a social change in volunteers. Within the two years a new website was introduced which allowed for better communication and a new database was introduced to better manage all aspects of the charity.
This report will demonstrate the workings of a Multi-layer-perceptron (MLP) Neural Network that is capable of implementing momentum. The network is capable of learning trends over a series of inputs so that once the network has been fully trained a prediction of future results can be made. In this report the MLP will be used to learn the XOR problem as well as learn trends over a set of sample statistics for the growth in traffic over 5 years. The network can be used to predict traffic flow in future years for up to 5 different road types.
Published as part of the 2005 Symposium for Cybernetics Annual Research Projects (SCARP). Haptic technology allows the simulation of real objects and real world physics within a virtual world. Whilst this, in many fields, is a solution to many practical issues it also draws attention away from the possibility of simulating objects and environments that simply can not exist in the real world. This can extend from simulating behaviours such as inverse gravity right through to impossible objects such as a tardis. Haptics can also be used to investigate how the human mind creates links between object properties. i.e. A small object is often seen to be light where as a large object is heavy. Haptics allows these basic, natural human assumptions to be tested by creating objects with properties that go against instinct.
The field of evolutionary computation exists to develop methods and techniques which can be implemented in to any given problem in order to find the most optimum solution through the mutation and breeding of a set of possible solutions. This breeding and mutation is often based on Darwin’s theories and are designed to mimic real world evolution They can prove to be most useful when the designer of the system does not actually know the solution. This paper will focus mainly on the branch of evolutionary computation known as Genetic Algorithms and their uses will be demonstrated through two function optimisation tasks. The techniques will finally be adapted to produce a solution for a Connect-4 playing system.