Mean-Field Theory of Spin Glasses and Neural Networks with Finite Coordination Number
The mean-field theory of dilute spin glasses and neural networks is studied in the limit where the average coordination number is finite (i.e., the average number of neighbors connected to each site). The zero-temperature phase diagram is calculated. Comparison between the properties of dilute neural networks and fully connected nets is presented. The relationship between the different phases and the percolation transition is discussed.
KeywordsSpin Glass Monte Carlo Step Average Coordination Number Percolation Transition Retrieval State
Unable to display preview. Download preview PDF.
- 3.I. Kanter and H. Sompolinsky, preprint.Google Scholar
- 5.I. Kanter and H. Sompolinsky, to be published.Google Scholar
- 8.M. Mezard and G. Parisi, preprint.Google Scholar
- 9.C. De Dominicis and P. Mottishaw, preprint.Google Scholar
- 10.P. Erdos and A. Reyni, in The Art of Counting, edited by J. Spencer ( MIT Press, Cambridge, MA, 1973 ).Google Scholar
- 13.I am grateful to D.S. Fisher for drawing my attention to this point.Google Scholar
- 14.B. Derrida, E. Gardner and A. Zippelius, preprint.Google Scholar