Attractor neural networks with double well synapses
Fig 5
Dependence of Storage Capacity on Noise Strength r3 for different values of Potential Width C.
When C is small (e.g., C = 0, 1, 2, 3), the storage capacity p monotonically decreases with the noise strength r3. However, for larger values of C (e.g., C = 4, 5), the storage capacity first increases with noise intensity, reaches a peak and then decreases. Thus, noise facilitates the learning of new memories in such cases. Other parameters are given as N = 10, 000, r1 = 0.1, r2 = 1, θ = 0, f = 0.5, c = 0.05.