{"id":"arxiv/2610.38437","url":"https://arxiv.org/abs/2610.38437","tags":["stat.ML","cond-mat.dis-nn","cs.IT","cs.LG","math.PR"],"score":0,"title":"Capacity of the Hopfield-Hebbian dynamics with correlated patterns","domain":"stat.ML","source":"arxiv","authors":["Mihailo Stojnic"],"reviews":0,"subject":"stat.ML","summary":"Allowing for small fraction of errors in retrieved stored binary patterns, Hopfield-Hebbian neural network (NN) associative memory dynamics [14] achieves linearly scaled capacity. The famous proportionality constant $α_c\\approx 0.14$ is obtained in a large dimensional scenario as the ratio of the number of reliably stored patterns $m$ and the patterns' length $n$ under the assumption that uniformly randomly chosen binary patterns are \\emph{independent} among themselves. While such an idealized scenario is now well understood, more practical scenarios with correlated/noisy datasets remaining…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-10-28","reactions":{}}