Title :
Efficiently approximating max-clique in a Hopfield-style network
Author_Institution :
Dept. of Comput. Sci., State Univ. of New York, Buffalo, NY, USA
Abstract :
The author approximates max-clique in a special case of the Hopfield network whose stable states are maximal cliques. Several energy-descent optimizing dynamics, both discrete and continuous, are presented. One of these emulates, as a special case, two well known greedy algorithms for approximating MAX-CLIQUE. Detailed empirical comparisons of random graphs are reported. Mean-field annealing, an efficient approximation to simulated annealing, is judged to be the most effective
Keywords :
Hopfield neural nets; simulated annealing; Hopfield-style network; approximating max-clique; energy-descent optimizing dynamics; greedy algorithms; mean-field annealing; simulated annealing; Artificial intelligence; Computer science; Convergence; Greedy algorithms; Information retrieval; Intelligent networks; Nonlinear equations; Object recognition; Polynomials; Simulated annealing;
Conference_Titel :
Neural Networks, 1992. IJCNN., International Joint Conference on
Conference_Location :
Baltimore, MD
Print_ISBN :
0-7803-0559-0
DOI :
10.1109/IJCNN.1992.227000
