The Mathworks Inc, South Natick, Massachusetts developer of the MatLab interactive application for providing high-performance numeric computation and graphics visualisation in an intuitive environment, has announced the Neural Network Toolbox version 2.0 for use with MatLab. According to the company, Version 2.0’s speed enhancement together with the inclusion of new architectures will improve the ability of technical professionals to design, simulate, customise and apply neural networks. Enhanced features include 10-fold speed improvements for training back-propagation networks using the Levenberg-Marquardt algorithm, radial base functions, Elman recurrent networks, adaptive linear networks, learning vector quantisation, and integration with The MathWorks Simulink, a non-linear dynamic simulation environment, which simplifies evaluation of neural network behaviour in the context of complete system designs. Karin McEwen, international marketing manager for Mathworks, believes that the Version 2 enhancements will make the Neural Network Toolbox ideal for applying neural network technology to solve real world problems that require large amounts of data, such as noise cancellation, financial forecasts, medical image interpretation and robotics. Single user pricing for the Neural Network Toolbox in the UK begins at ?862 on iAPX-86-based personal computers. MatLab is currently being used by the Rover Group and the University of Warwick in a research project to build controllers for a futuristic environmentally friendly vehicle. The superior computation environment and visual capabilities of MatLab has aided high level research using fuzzy logic techniques to build controllers that will manage energy levels within hybrid vehicles, the researchers claim.
