AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

SubTab: Subsetting Features of Tabular Data for Self-Supervised Representation Learning

Talip Ucar, Ehsan Hajiramezanali, Lindsay Edwards

NeurIPS 2021 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Invariance-based self-supervised learning · 9 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
Reconstruct the full feature space from feature subsets rather than reconstructing only the input subsetSubTab · Tabular representation learningdownstream classification accuracyImproved
Add swap-noise to the input (ablation)SubTab · MNISTaccuracy: 97.26 -> 97.47Improved
Add distance loss (ablation)SubTab with swap-noise · MNISTaccuracy: 97.47 -> 97.52Improved
Replace swap-noise with zero-out noise (ablation)SubTab · MNISTaccuracy: 97.47 -> 97.25No clear effect
Replace swap-noise with Gaussian noise (ablation)SubTab · MNISTaccuracy: 97.47 -> 97.25No clear effect
Add contrastive loss to reconstruction loss (ablation)SubTab · MNISTaccuracy: 97.13 -> 97.26Improved
Use contrastive loss alone instead of reconstruction loss (ablation)SubTab · MNISTaccuracy: 97.13 -> 97.11No clear effect
Shuffle feature order (ablation)SubTab with swap-noise and distance loss · MNISTaccuracy: 97.52 -> 97.2No clear effect
Increase latent dimension from 128 to 512 while aggregating subset embeddings (ablation)SubTab · MNISTaccuracy: 97.52 -> 97.86Improved
Concatenate subset embeddings instead of aggregating them (ablation)SubTab with latent dimension 512 · MNISTaccuracy: 97.86 -> 95.92Hurt
Use a shallow one-layer encoder and decoder instead of the deep architecture (ablation)SubTab · MNISTaccuracy: 97.86$\pm$0.07 -> 98.31$\pm$0.06Improved
Use a shallow one-layer encoder and decoder instead of the deep architecture (ablation)SubTab · Obesityaccuracy: 71.13$\pm$4.08 -> 66.88$\pm$5.35Hurt

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