Skip to main content
Fig. 2 | The Journal of Mathematical Neuroscience

Fig. 2

From: Robust Exponential Memory in Hopfield Networks

Fig. 2

Learning critical networks with exponential memory by minimizing probability flow on few training patterns. For numbers of vertices \(v = 80, 64, 48\) (\(k = 40, 32, 24\)) with 50 trials each, the average percent of 1000 randomly drawn cliques that are memories vs. the fraction of training samples to total number of k-cliques. Inset displays enlarged version of the region demarcated by black square; filled regions indicate standard deviation errors over these 50 trials. Dotted lines are average percentage of correct bits after converging dynamics

Back to article page