Gamma Oscillations Facilitate Effective Learning in Excitatory-Inhibitory Balanced Neural Circuits

Figure 7

Special circuit structures formed by plasticity. (a, b) The relation between the sums of incoming excitatory and inhibitory synaptic weight of different neurons in the plastic circuit after learning, sorted according to the ascending order of inhibitory weight sum. Parameters are set as . (a) Synchronous with . (b) Asynchronous with. (c–f) Illustration of shuffling schemes. Each figure represents a connectivity matrix. In (c–e), we shuffled the elements of the connectivity sub-matrix of the chosen excitatory neurons (to encode the memory). (c) The input elements of each neuron (each row) are shuffled (total input is preserved). (d) The output elements of each neuron (each column) are shuffled (total input is changed). (e) The elements of each row are first shuffled and then the row is randomly shuffled (total excitatory input is preserved, but total inhibitory input is changed). The shuffling scheme of (f) the elements of each row of the submatrix of excitatory neurons is first shuffled, and then, the row of the entire matrix is randomly shuffled (total excitatory and inhibitory input are preserved). If the shuffled matrix results in synchronous/asynchronous dynamics, then it is marked in a red/black box.

(a)

(b)

(c)

(d)

(e)

(f)