Phase transitions in dilute, locally connected neural networks

Journal Article · Wed Apr 15 00:00:00 EDT 1992 · Physical Review A. General Physics; (United States)

DOI:https://doi.org/10.1103/PhysRevA.45.6135· OSTI ID:7302754

Strandburg, K J ^[1]; Peshkin, M A ^[2]; Boyd, D F; Chambers, C; O'Keefe, B ^[3]

---

1. Argonne National Laboratory, Argonne, Illinois 60439 (United States) Northwestern University, Evanston, Illinois 60208 (United States)
2. Northwestern University, Evanston, Illinois 60208 (United States)
3. Argonne National Laboratory, Argonne, Illinois 60439 (United States)

We report numerical studies of the memory-loss'' phase transition in Hopfield-like symmetric neural networks in which the neurons are connected to all other neurons within a local neighborhood (dense, short-range connectivity). The number of connections per neuron {ital K} scales as the number of neurons {ital N} raised to a power less than 1 (i.e., {ital K}{similar to}{ital N}{sup {eta}}, {eta}{lt}1). We use the recently developed Lee-Kosterlitz finite-size scaling technique to determine the critical value of {eta} below which the first-order phase transition disappears.

• DOE Contract Number: W-31109-ENG-38
• OSTI ID: 7302754
• Journal Information: Physical Review A. General Physics; (United States), Journal Name: Physical Review A. General Physics; (United States) Vol. 45:8; ISSN 1050-2947; ISSN PLRAA
• Country of Publication: United States
• Language: English

Related Subjects

560000* -- Biomedical Sciences
Applied Studies
63 RADIATION, THERMAL, AND OTHER ENVIRON. POLLUTANT EFFECTS ON LIVING ORGS. AND BIOL. MAT.
99 GENERAL AND MISCELLANEOUS
990200 -- Mathematics & Computers
FINITE-RANGE INTERACTIONS
INTERACTIONS
NETWORK ANALYSIS
NEURAL NETWORKS
PHASE TRANSFORMATIONS
SIMULATION