This work explains network dynamics, learning paradigms, and computational capabilities of feedforward, self-organization, and feedback neural network models-addressing specific problems such as data fusion and data modeling. It goes on to describe a neural network simulation software package - USTCNET and gives some segments of the program.

Neural Networks have been the theater of a dramatic increase of activities in the last five years. The interest of mixing results from fields as different as neurobiology, physics (spin glass theory), mathematics (linear algebra, statistics ... ), computer science (software engineering, hardware architectures ... ) or psychology has attracted a large number of researchers to the field. The perspective of dramatic improvements in many applications has lead important companies to launch new neural network programs and start-ups have mushroomed to address this new market. Throughout the world large programs are being set-up: in Japan the government has committed more than $18 million per year to its 20 year Human Frontier Science program; the DARPA and the US Navy have alloted more than $10 million per year each and other US government agencies are contributing to important but less ambitious programs. Neural networks are also a major research are in the supercomputing initiative. Europe has from the beginning taken an active part in funding major projects in the new field with BRAIN, BRA, ANNIE and PYGMALION (Esprit). Approximately $20 million has been invested to date since 1988 and new programs of nearly $30 million are being funded for the next 3 years. National projects in certain countries may globally double these amounts. Neural network conferences are attracting larger audiences than ever before. Prior to 1987 attendance never surpassed 300. The June 1989 IJCNN conference in Washington had over 2200 participants.