REVIEW 2 cited by
Learning Narrow One-Hidden-Layer ReLU Networks
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
abstract
We consider the well-studied problem of learning a linear combination of $k$ ReLU activations with respect to a Gaussian distribution on inputs in $d$ dimensions. We give the first polynomial-time algorithm that succeeds whenever $k$ is a constant. All prior polynomial-time learners require additional assumptions on the network, such as positive combining coefficients or the matrix of hidden weight vectors being well-conditioned. Our approach is based on analyzing random contractions of higher-order moment tensors. We use a multi-scale analysis to argue that sufficiently close neurons can be collapsed together, sidestepping the conditioning issues present in prior work. This allows us to design an iterative procedure to discover individual neurons.
Forward citations
Cited by 2 Pith papers
-
Implicit High-Order Moment Tensor Estimation and Learning Latent Variable Models
A unified implicit moment tensor estimation framework yields poly(d,k)-time learners for mixtures of linear regressions, spherical Gaussians, and positive sums of ReLU activations, with one unproven step in the regres...
-
Gradient dynamics for low-rank fine-tuning beyond kernels
In a student-teacher model with Gaussian inputs and a rank-1 teacher perturbation, online SGD converges to the teacher in d k^{O(1)} iterations, independent of the activation's Hermite information exponent.
Discussion (0). Continue with ORCID to comment.