How to escape atypical regions in the symmetric binary perceptron: a journey through connected-solutions states

Abstract

We study the binary symmetric perceptron model, and in particular its atypical solutions. While the solution-space of this problem is dominated by isolated configurations, it is also solvable for a certain range of constraint density $\alpha$ and threshold $\kappa$. We provide in this paper a statistical measure probing sequences of solutions, where two consecutive elements shares a strong overlap. After simplifications, we test its predictions by comparing it to Monte-Carlo simulations. We obtain good agreement and show that connected states with a Markovian correlation profile can fully decorrelate from their initialization only for $\kappa >{\kappa }_{\mathrm{no}-mem.\mathrm{state}}$ (${\kappa }_{\mathrm{no}-mem.\mathrm{state}}\sim \sqrt{0.91\log (N)}$ for $\alpha =0.5$ and $N$ being the dimension of the problem). For $\kappa <{\kappa }_{\mathrm{no}-mem.\mathrm{state}}$, we show that decorrelated sequences still exist but have a non-trivial correlations profile. To study this regime we introduce an $Ansatz$ for the correlations that we label as the nested Markov chain.

References in corpus (6)

• Efficient supervised learning in networks with binary synapses
• Origin of the computational hardness for learning with binary synapses
• Locked constraint satisfaction problems
• Constraint satisfaction problems with isolated solutions are hard
• On the Atypical Solutions of the Symmetric Binary Perceptron
• Statistical mechanics of the maximum-average submatrix problem

Cited by in corpus (1)

• The maximum-average subtensor problem: equilibrium and out-of-equilibrium properties