Temporal association in neural networks at finite temperatures
M. Y. Choi, Jihyun Choi, Kibeom Park · Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics · 1998
Temporal association in neural networks, which retrieves time sequences of stored patterns, is made possible by introducing asymmetry in the synaptic coupling. Such temporal association in the asymmetric Hopfield model is first considered, with particular attention to the finite-temperature effects on the retrieval capability. We then turn to the dynamic model, which is the main topic of this paper, and investigate its temporal association properties both analytically and numerically. The phase diagram is obtained in the three-dimensional parameter space, and its structure is discussed according to the storage and other parameter values.