A high performance neural net simulation environment
U.A. Muller · 1994
The challenge of neural net simulators is to provide an easy to use environment on one hand and to cover the large amount of flexibility required for experimental research on the other hand. This paper describes an efficient approach where the tasks of determining the experimental setup and programming the actual algorithms are independent. A simple interactive command language is used for the former one and standard programming languages are used for the latter one. A practical implementation on the MUSIC parallel supercomputer (multiprocessor system with intelligent communication) confirms the efficiency of the concept.>