Memory melting in neural networks
Patricio A. Pérez · Physical Review A · 1989
A simple analysis and a numerical calculation show that the ability to retrieve information from the Hopfield network is enhanced when thermal noise is introduced. It is also shown that the best efficiency of the neural network is achieved in the range of temperatures ${T}_{1}$${T}_{2}$, where ${T}_{1}$ corresponds to the noise necessary to eliminate spurious memories and ${T}_{2}$ to the temperature at which the memorized states are no longer stable.