Title :
A method for data path synthesis using neural networks
Author :
Harmanani, Haidar
Author_Institution :
Dept. of Comput. Eng. & Sci., Lebanese American Univ., Beirut, Lebanon
Abstract :
Presents a deterministic parallel algorithm to solve the data path allocation problem in high-level synthesis. The method is based on the modified Hopfield neural network model of computation and the McCulloch-Pitts binary neuron model. The proposed algorithm has a running time complexity of O(1) for a neural network with n vertices and c cliques. A sequential simulator was implemented for the proposed algorithm on a Linux Pentium PC under X Windows. Several circuits hare been attempted, all yielding sub-optimal solutions.
Keywords :
Hopfield neural nets; computational complexity; high level synthesis; logic simulation; microcomputer applications; parallel algorithms; Linux Pentium PC; McCulloch-Pitts binary neuron model; X Windows; cliques; computational model; data path allocation problem; data path synthesis; deterministic parallel algorithm; high-level synthesis; modified Hopfield neural network; sequential simulator; sub-optimal solutions; time complexity; Circuit simulation; Computational modeling; Computer networks; High level synthesis; Hopfield neural networks; Linux; Network synthesis; Neural networks; Neurons; Parallel algorithms;
Conference_Titel :
Electrical and Computer Engineering, 1999 IEEE Canadian Conference on
Conference_Location :
Edmonton, Alberta, Canada
Print_ISBN :
0-7803-5579-2
DOI :
10.1109/CCECE.1999.807241
