Zhaoping Li, Peter Dayan
Symmetrically connected recurrent networks have recently been used as models of a host of neural computations. However, be(cid:173) cause of the separation between excitation and inhibition, biolog(cid:173) ical neural networks are asymmetrical. We study characteristic differences between asymmetrical networks and their symmetri(cid:173) cal counterparts, showing that they have dramatically different dynamical behavior and also how the differences can be exploited for computational ends. We illustrate our results in the case of a network that is a selective amplifier.