AI Paper Atlas

Research map / Optimization / Implicit bias and generalization

On the training dynamics of deep networks with $L_2$ regularization

Aitor Lewkowycz, Guy Gur-Ari

NeurIPS 2020 · Optimization › Implicit bias and generalization · approach: Explicit and implicit regularization interplay · 27 citations (OpenAlex)

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
train for a number of steps proportional to 1/lambda instead of a fixed budget while decreasing L2overparameterized networks (WRN, FC, CNN) · CIFAR-10 / MNISTtest accuracyImproved
reduce L2 parameter beyond a certain value under a fixed training budgetoverparameterized networks · CIFAR-10 / MNISTtest accuracyHurt
use predicted L2 (from measured c) instead of tuned L2WRN28-10 with momentum, augmentation · CIFAR-10test accuracy: 0.967 -> 0.960No clear effect
use AutoL2 dynamical L2 schedule instead of tuned constant L2 (no LR schedule)WRN28-10 · CIFAR-10test accuracy and training speedImproved
use AutoL2 together with a learning rate scheduleWRN28-10 · CIFAR-10test accuracyHurt
use small nonzero L2 instead of noneoverparameterized networks trained to optimum · CIFAR-10 WRN28-10test accuracyImproved

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