AI Paper Atlas

Research map / Learning theory and algorithms / Sample complexity

Pareto Frontiers in Deep Feature Learning: Data, Compute, Width, and Luck

Benjamin L. Edelman, Surbhi Goel, Sham M. Kakade, eran malach, Cyril Zhang

NeurIPS 2023 · Learning theory and algorithms › Sample complexity · approach: Robust and surrogate-loss learning bounds · 0 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
increase MLP width up to 10^52-layer MLP · Offline sparse paritysuccess probabilityImproved
increase MLP width2-layer MLP · Offline sparse paritysample efficiencyImproved
increase MLP width2-layer MLP · Offline sparse parityiterations to convergenceImproved
sparse axis-aligned (2-hot) initialization instead of dense initialization (ablation)2-layer MLP · Offline sparse paritysuccess probability and sample efficiencyImproved
initialize each neuron with s=2 hot weights versus other sparsity levels (ablation)Sparse init MLP · Offline sparse paritysuccess probabilityImproved
reduce dataset size toward the edge of feasibility2-layer MLP · Offline sparse parityiterations to convergenceHurt
increase dataset size m2-layer MLP · Offline sparse paritysuccess probabilityMixed
increase MLP width on downsampled tabular dataMLP · OpenML tabular benchmark (16 tasks)test errorImproved
sparse axis-aligned initialization on tabular data (ablation)MLP · OpenML tabular benchmark (16 tasks)test errorMixed
wide sparse-init MLP versus tuned random forestMLP · OpenML tabular benchmarktest errorMixed

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