Distributed programming for neural networks
Nikola B. Serbedzija, Gerd Kock, Stefan Jähnichen · 2002
Presents a high-level approach for parallel and distributed programming of connectionist models. A generic description of an abstract connectionist model is given, providing means for necessary modifications and extensions. A concurrency model supports asynchronous communication among massively interconnected units, and distributed implementation provides a truly parallel and robust execution environment. This presentation covers the design rationales, programming model and implementation details, and is illustrated with concrete examples.>