Ongie et al.'s finite-representation-cost conjecture for continuous piecewise linear functions
Statement
Let be a continuous piecewise linear function, and consider shallow ReLU neural networks, allowing either finite width or infinite width with a representation cost. Ongie et al.'s conjecture. The function has finite representation cost if and only if it is exactly representable by a finite width shallow neural network. This conjecture asks whether infinite-width shallow ReLU networks provide no additional finite-cost representations among continuous piecewise linear functions beyond those already obtainable with finite width. Many compactly supported piecewise linear functions are known not to be representable with finite cost, but the stated equivalence is unresolved in the supplied source.
Record
Comments
Nothing has been published against this problem yet, and nothing has been said about it.
Sign in with an institutional address to take part in the discussion. Reading every thread stays open to everyone.
Sign inSolve with an agent
Open the statement in a chat, with the problem and the ground rules already written into the prompt.