Exact full-RSB SAT/UNSAT transition in infinitely wide two-layer neural networks

Abstract

We analyze the problem of storing random pattern-label associations using two classes of continuous non-convex weights models, namely the perceptron with negative margin and an infinite-width two-layer neural network with non-overlapping receptive fields and generic activation function. Using a full-RSB ansatz we compute the exact value of the SAT/UNSAT transition. Furthermore, in the case of the negative perceptron we show that the overlap distribution of typical states displays an overlap gap (a disconnected support) in certain regions of the phase diagram defined by the value of the margin and the density of patterns to be stored. This implies that some recent theorems that ensure convergence of Approximate Message Passing (AMP) based algorithms to capacity are not applicable. Finally, we show that Gradient Descent is not able to reach the maximal capacity, irrespectively of the presence of an overlap gap for typical states. This finding, similarly to what occurs in binary weight models, suggests that gradient-based algorithms are biased towards highly atypical states, whose inaccessibility determines the algorithmic threshold.

39 pages, 12 figures

References in corpus (14)

• Fractal free energy landscapes in structural glasses
• The Overlap Gap Property: a Geometric Barrier to Optimizing over Random Structures
• Origin of the computational hardness for learning with binary synapses
• Statistical Mechanics of Deep Linear Neural Networks: The Back-Propagating Kernel Renormalization
• Unveiling the structure of wide flat minima in neural networks
• Learning through atypical "phase transitions" in overparameterized neural networks
• Gaussian Universality of Perceptrons with Random Labels
• Surfing on minima of isostatic landscapes: avalanches and unjamming transition
• High dimensional optimization under non-convex excluded volume constraints
• Tractability from overparametrization: The example of the negative perceptron
• On neural network kernels and the storage capacity problem
• High-dimensional manifold of solutions in neural networks: insights from statistical physics
• Fixed width treelike neural networks capacity analysis -- generic activations
• Exact capacity of the \emph{wide} hidden layer treelike neural networks with generic activations