Research map / Optimization / Nonconvex and smooth optimization
Coresets for Data-efficient Training of Machine Learning Models
Baharan Mirzasoleiman, Jeff Bilmes, Jure Leskovec
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 |
|---|---|---|---|
| Use a 10% CRAIG-selected weighted subset instead of the full training set | SGD, SVRG, and SAGA on Covtype · Covtype | speedup: 2.75x, 4.5x, and 2.5x | Improved |
| Use a 30% CRAIG-selected subset instead of the full dataset | SGD on Ijcnn1 · Ijcnn1 | speedup: 5.6x | Improved |
| Select a 10% CRAIG subset rather than a 10% random subset | SGD, SVRG, and SAGA · Covtype | training loss | Improved |
| Train on a 10% CRAIG-selected subset rather than the full dataset | SGD, SVRG, and SAGA · Covtype | training loss | No clear effect |
| Train on a 10% CRAIG-selected subset rather than the full dataset | IG methods · Covtype | test error rate | No clear effect |
| Train on a 50% CRAIG-selected subset each epoch rather than the full dataset | SGD · MNIST | speedup: 2x to 3x | Improved |
| Train on CRAIG-selected subsets rather than the full dataset | SGD · MNIST | generalization performance | Improved |
| Select subsets with CRAIG rather than randomly at the same subset size | SGD on ResNet-20 · CIFAR10 | test accuracy | Improved |
| Select subsets more frequently | SGD on ResNet-20 · CIFAR10 | test accuracy | Improved |
| Use larger subsets or update them more frequently | SGD on ResNet-20 · CIFAR10 | test accuracy | Improved |
| Use the greedy CRAIG ordering rather than random permutations of the same subset | Incremental gradient training · Not specified | convergence rate | Improved |
Closest papers by meaning
- Adaptive Second Order Coresets for Data-efficient Machine Learning (ICML 2022 · similarity 0.87)
- Stable Coresets via Posterior Sampling: Aligning Induced and Full Loss Landscapes (NeurIPS 2025 · similarity 0.86)
- Refined Coreset Selection: Towards Minimal Coreset Size under Model Performance Constraints (ICML 2024 · similarity 0.86)
- Gradient Regularized Newton Boosting Trees with Global Convergence (NeurIPS 2026 · similarity 0.85)
- On the Convergence of Loss and Uncertainty-based Active Learning Algorithms (NeurIPS 2024 · similarity 0.85)
- Incremental Gradient Descent with Small Epoch Counts is Surprisingly Slow on Ill-Conditioned Problems (ICML 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.
