Research map / Label-efficient and robust learning / Noisy labels
A Path to Simpler Models Starts With Noise
Lesia Semenova, Harry Chen, Ronald Parr, Cynthia Rudin
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 changed | Where | Measured | Effect |
|---|---|---|---|
| increase uniform label noise (up to 25 percent) | sparse decision trees (TreeFARMS, depth 4) · 12 tabular datasets | number of trees in the Rashomon set | Improved |
| increase uniform label noise | sparse decision trees (depth 4) · 12 tabular datasets | number of patterns in the Rashomon set | Improved |
| increase uniform label noise | sparse decision trees (depth 4) · 12 tabular datasets | pattern diversity | Mixed |
| increase label noise | CART · four real-world datasets | train-validation accuracy gap | Improved |
| increase label noise, with depth chosen by cross-validation | CART · four real-world datasets | optimal tree depth | Improved |
| increase label noise, with number of estimators chosen by cross-validation | gradient boosted trees · tabular datasets | best number of estimators | Improved |
| increase hypothesis space complexity (tree depth 1 to 7) | decision trees · 14 datasets | Rashomon ratio | Hurt |
| increase linear model complexity (sparsity 1 to 4) | sparse linear models · 6 datasets | pattern Rashomon ratio | Hurt |
| increase margin noise (Gaussians moved closer) | fixed linear model · synthetic Gaussians in 3, 5, 7, 10 dimensions | variance of 0-1 loss | Improved |
| add additive attribute noise (Gaussian) | fixed linear model · synthetic Gaussians | variance of 0-1 loss | Improved |
| add additive attribute noise to ridge regression | ridge regression · theory | Rashomon ratio | Improved |
| compare noisier datasets (Compas, Coffee House, FICO) with cleaner ones (Car Evaluation, Monks1, Monks3) | sparse decision trees · tabular datasets | models in Rashomon set | Improved |
Closest papers by meaning
- Position: Amazing Things Come From Having Many Good Models (ICML 2024 · similarity 0.90)
- Using Noise to Infer Aspects of Simplicity Without Learning (NeurIPS 2024 · similarity 0.89)
- Dropout-Based Rashomon Set Exploration for Efficient Predictive Multiplicity Estimation (ICLR 2024 · similarity 0.88)
- RashomonGB: Analyzing the Rashomon Effect and Mitigating Predictive Multiplicity in Gradient Boosting (NeurIPS 2024 · similarity 0.86)
- DataDecide: How to Predict Best Pretraining Data with Small Experiments (ICML 2025 · similarity 0.85)
- SORTeD Rashomon Sets of Sparse Decision Trees: Anytime Enumeration (NeurIPS 2025 · similarity 0.85)
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.
