AI Paper Atlas

Research map / Optimization / Implicit bias and generalization

On the Effect of Initialization: The Scaling Path of 2-Layer Neural Networks

Sebastian Neumayer, Lénaïc Chizat, Michael Unser

JMLR 2024 · Optimization › Implicit bias and generalization · approach: Shallow ReLU network convergence

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
match the gradient-descent initialization scale beta to the scaling-path scale alpha (ablation)gradient descent training · 2D interpolation, 10 samplesregularized W2 objective gap: beta=alpha gives minimum for large alpha, e.g. 5.66e2 at alpha=3.3e0 vs 1.79e3 at beta=3.3e0, alpha=1e1Improved
use gradient descent solution instead of the variational scaling-path solution2-layer ReLU interpolation · 2D interpolation, 10 samplesfunctional value: 7.50e1 -> 9.36e1 at alpha=1e-1Hurt
increase regularization scale alphascaling path · 2D interpolation, 10 samplessmoothness of solutionImproved
increase alpha toward infinityscaling path · 2D interpolation, 10 samplesconvergence to kernel optimum: 7.73e2 -> 2.79e2 vs 2.83e2 NTK valueImproved
decrease alpha and beta toward small valuesgradient descent vs scaling path · 2D interpolation, 10 samplessimilarity of solutionsHurt
restrict support of measure to the smaller grid P instead of Q (ablation)scaling path discretization · 2D interpolation, 10 samplessolution qualityNo clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy