AN APPROACH TO OPTIMIZATION OF SPIN CONFIGURATION IN SPIN-GLASS SYSTEMS BY CHAOTIC NEURAL-NETWORK METHOD

A chaotic neural network model with globally coupled map (GCM) is empl oyed to optimize spin configuration in Ising spin-glass systems. It ha s been supposed in ising spin glass systems that a vast number of meta stable states exist, which make it difficult to find the ground state and low energy metastable states. Through our calculations for Ising s pin systems, many such metastable states al zero temperature are obtai ned. From the results of the Sherrington-Kirkpatrick (SK) Ising model, we show the existence of a non-trivial tree structure among these low energy metastable states, which supports the existence of ultrametric al structure predicted by Mezard, Parisi, Sourlas, Thoulouse, and Vira soro.