Learning in Multilayer Models

摘要: Most neural network programs for personal computers simply control a set of fixed, canned network-layer algorithms with pulldown menus. This new tutorial offers hands-on experiments different approach. A simple matrix language lets users create their own networks and combine networks, this is the only currently available software permitting combined simulation together other dynamic systems such as robots or physiological models. The enclosed student version DESIRE/NEUNET differs from full system in size its data area includes screen editor, compiler, color graphics, help screens, ready-to-run examples. Users can also add screens interactive menus.The book provides an introduction to simulation, on software, many complete including several backpropagation schemes, creeping random search, competitive learning without adaptive-resonance function "conscience," counterpropagation, nonlinear Grossberg-type neurons, Hopfield-type bidirectional associative memories, predictors, learning, biological clocks, identification, more.In addition, introduces simple, integrated environment programming, displays, report preparation. Even differential equations are entered ordinary mathematical notation. need not learn C LISP program neuron To permit truly experiments, extra-fast compilation unnoticeable, simulations execute faster than PC FORTRAN.The nearly 90 illustrations include block diagrams, computer programs, simulation-output graphs.Granino A. Kom has been Professor Electrical Engineering at University Arizona worked aerospace industry decade. He author ten engineering texts handbooks.