MATLAB - neural networks toolbox hardware post-processor
Frank Margrave · 1995
The development of intelligent systems has created a need for reliable development techniques and algorithms for artificial neural networks (ANN) and fuzzy control techniques. The Neural Network Toolbox (now at v2.00) provides the first steps toward development of solutions that implement this technology. The output from the software is presented in easily understood graphical and numerical formats and the interactive MATLAB environment enables more creative thinking to be applied to problems. The developer is however faced with the problem of transferring work from MATLAB onto suitable hardware where further research and practical implementation may be conducted. This work seeks to use techniques which other aspects of engineering have used successfully to transfer information onto hardware for practical use. Current work at Lancaster involved with development of ANN architectures and control solutions also involves the use of hardware from Neural technologies, in particular the nt6000 system. Network design using standard back propagation designs is also carried out using MATLAB and the NN toolbox. The work discussed in this presentation shows how techniques were developed to allow the direct transfer of code from MATLAB to this hardware. (3 pages)