ProbXiv
Explorersign in
Problem archiveProblem record

Statement

Let d53d=(d0,d55fd53dots,dL)d53d=(d_0,d55fd53dots,d_L) be a fixed vector of widths, with di>1d_i>1 for i=1,d55fd53dots,L−1i=1,d55fd53dots,L-1, and let rr denote the activation degree. The neurovariety d4b1d53d,rd4b1_{d53d,r} is the image variety associated with this architecture, and its expected dimension is the dimension predicted by the parameter count and ambient dimension.

Asymptotic expected-dimension conjecture. There exists a threshold d53fdd53f(d53d)d53fdd53f(d53d) such that, whenever r>d53fr>d53f, the neurovariety d4b1d53d,rd4b1_{d53d,r} attains the expected dimension.

The claim concerns eventual non-defectivity as the activation degree grows. It was previously asserted as a theorem, but the proof contained a mistake; the statement has subsequently been proved in the cited work on activation functions.

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 in

Solve with an agent

Open the statement in a chat, with the problem and the ground rules already written into the prompt.

This opens a third-party site. Nothing is posted back to ProbXiv and nothing you write there is recorded here — what a model gives you is an attempt, which a person still has to check.