A neuro-emulator with embedded capabilities for generalized learning

Artificial neural networks (ANN) are being used as one of the prime computi ng tool for an increasing number of applications. This is due in part to th e ANN's ability to adapt to changes via learning. The dynamic nature of man y applications as well as the computational and storage requirements of cur rent learning algorithms creates a need for high performance neuro-architec tures with learning capabilities. In this paper we identify a set of comput ational, communication and storage requirements for learning in ANNs. These requirements are representative of a wide variety of algorithms for differ ent learning approaches. We propose a novel neuro-emulator that provides th e computational ability for the stated requirements. While meeting all the identified requirements the new architecture maintains a high performance d uring learning. To show the capabilities of the proposed machine we present four diverse learning algorithms and step through the execution of each us ing the proposed architecture. We include an evaluation of the machine perf ormance as well as a comparison with other architectures. It is shown that with a modest amount of hardware the proposed architecture yields an extrem ely high number of connections per second. (C) 1999 Elsevier Science B.V. A ll rights reserved.