Dreaming neural networks: Forgetting spurious memories and reinforcing pure ones. (April 2019)

• Record Type: Journal Article
• Authors: Fachechi, Alberto; Agliari, Elena; Barra, Adriano
• Is Part Of: Neural networks. Volume 112(2019)
• Page Start: 24
• Page End: 40
• Publication Date: 2019-04
• Subjects: Unlearning -- Reinforcement learning -- Statistical mechanics -- Sleep&dream; Neural computers -- Periodicals; Neural networks (Computer science) -- Periodicals; Neural networks (Neurobiology) -- Periodicals; Nervous System -- Periodicals; Ordinateurs neuronaux -- Périodiques; Réseaux neuronaux (Informatique) -- Périodiques; Réseaux neuronaux (Neurobiologie) -- Périodiques; Neural computers; Neural networks (Computer science); Neural networks (Neurobiology); Periodicals; 006.32
• DOI: 10.1016/j.neunet.2019.01.006
• Languages: English
• ISSNs: 0893-6080

Abstract

Abstract: The standard Hopfield model for associative neural networks accounts for biological Hebbian learning and acts as the harmonic oscillator for pattern recognition, however its maximal storage capacity is α ∼ 0 . 14, far from the theoretical bound for symmetric networks, i.e. α = 1 . Inspired by sleeping and dreaming mechanisms in mammal brains, we propose an extension of this model displaying the standard on-line (awake) learning mechanism (that allows the storage of external information in terms of patterns) and an off-line (sleep) unlearning&consolidating mechanism (that allows spurious-pattern removal and pure-pattern reinforcement): this obtained daily prescription is able to saturate the theoretical bound α = 1, remaining also extremely robust against thermal noise. The emergent neural and synaptic features are analyzed both analytically and numerically. In particular, beyond obtaining a phase diagram for neural dynamics, we focus on synaptic plasticity and we give explicit prescriptions on the temporal evolution of the synaptic matrix. We analytically prove that our algorithm makes the Hebbian kernel converge with high probability to the projection matrix built over the pure stored patterns. Furthermore, we obtain a sharp and explicit estimate for the "sleep rate" in order to ensure such a convergence. Finally, we run extensive numerical simulations (mainly Monte Carlo sampling) to check the approximations underlying the analytical investigations (e.g., we developed the whole theory at the so called replica-symmetric level, as standard in the Amit–Gutfreund–Sompolinsky reference framework) and possible finite-size effects, finding overall full agreement with the theory. Highlights: The Hopfield model can sleep once learning is over. In this way the storage capacity reaches 1 (rather than 0.138). The network remains extremely robust to thermal noise. Hebbian coupling is forced to the projection matrix by sleeping. Extensive simulations are in full agreement with the theory.