Research map / Learning theory and algorithms / Prediction and decision losses
Effective Parallelisation for Machine Learning
Michael Kamp, Mario Boley, Olana Missura, Thomas Gärtner
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 |
|---|---|---|---|
| Parallelize base learners on data subsets and aggregate their hypotheses with Radon points | regularised risk minimisation algorithms · SUSY | runtime: 721 times faster | Improved |
| Parallelize WekaLogReg with the Radon machine | WekaLogReg · CASP9 | runtime: 15% faster | Improved |
| Parallelize WekaLogReg with the Radon machine | WekaLogReg · CASP9 | AUC: 2.6% better | Improved |
| Parallelize WekaLogReg with the Radon machine | WekaLogReg · 20_newsgroups | AUC: 2.2% worse | Hurt |
| Replace Radon-point aggregation with averaging-at-the-end (ablation) | Radon machine · 5 large datasets | AUC: at least 10% | Hurt |
| Replace Radon-point aggregation with averaging-at-the-end (ablation) | Radon machine · 5 large datasets | runtime | Improved |
| Increase aggregation height from h=1 to h=max (ablation) | Radon machine · datasets used in the base-learner comparison | runtime | Improved |
| Increase aggregation height from h=1 to h=max (ablation) | Radon machine with WekaSGD · codrna | AUC | Mixed |
| Increase aggregation height from h=1 to h=max (ablation) | Radon machine · SUSY and HIGGS | runtime: an order of magnitude | Improved |
| Use the Radon machine instead of regularised least squares regression | regularised least squares regression · YearPredictionMSD | RMSE: 12.57 -> 13.64 | Hurt |
| Use the Radon machine instead of regularised least squares regression | regularised least squares regression · YearPredictionMSD | runtime: 197-times faster | Improved |
| Use the Radon machine with h=max instead of Spark learners | WekaLogReg · SUSY, wikidata, and CASP9 | AUC | Improved |
Closest papers by meaning
- Pfeife: Automatic Pipeline Parallelism for PyTorch (ICML 2025 · similarity 0.86)
- ATA: Adaptive Task Allocation for Efficient Resource Management in Distributed Machine Learning (ICML 2025 · similarity 0.85)
- A Class of Parallel Doubly Stochastic Algorithms for Large-Scale Learning (JMLR 2020 · similarity 0.85)
- Optimal Parallelization of Boosting (NeurIPS 2024 · similarity 0.84)
- Parallelizing non-linear sequential models over the sequence length (ICLR 2024 · similarity 0.84)
- UniAP: Unifying Inter- and Intra-Layer Automatic Parallelism by Mixed Integer Quadratic Programming (CVPR 2025 · similarity 0.82)
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.
