Assigning ADT modules with random neural networks
Lonnie R. Welch, Alexander D. Stoyenko, S. Chen · 2002
The authors apply random neural networks to the problem of assigning abstract data type modules (ADTs) to the processing elements of parallel computers. Although assignment of tasks has been discussed extensively in the literature, the automatic assignment of ADTs is a relatively new problem, and therefore they describe the problem in detail. This is followed by an introduction to the random neural network model and a presentation of the neural network solution to the assignment problem. Experimental results are presented comparing the solution to those obtained with random assignment and with a greedy heuristic. The random neural network is found to give significantly better results than the other two approaches in virtually every case.>