Remarks on the stability of asymmetric dynamical neural networks

The BSB (brain-state-in-a-box) neural network is restructured as a multivariable linear system with a nonlinear feedback. This reconstruction allows some well-developed techniques in the field of circuit and control theory to be used for resolving issues such as stability. The authors specifically consider the stability of the asymmetric BSB nets in terms of the nonexpansivity of the whole system and derive simple conditions to be satisfied by the weight matrix in order to guarantee the stability. It is seen that this framework can reproduce and generalize some of the previous results on the subject, like those previously developed by the authors (1989)