Neural Channel capacity conjecture for spare subnetworks

A Neural Channel is an overparameterized neural network that can approximately preserve its original behavior while storing information that can be received through an efficient reception procedure. A spare subnetwork is a subnetwork available for storing information without substantially changing the model's original behavior. The capacity of a Neural Channel is the maximum amount of information that can be stored and received with the prescribed loss tolerance.

Neural Channel capacity conjecture. The capacity of a Neural Channel increases with the size of the spare subnetwork.

This conjecture connects the storage capacity of trained neural networks with the amount of unused or spare architecture available for encoding information. The paper states that its conjectures will be investigated empirically; no resolution is supplied in the given text.