Research map / Learning theory and algorithms / Sample complexity
Linear and Kernel Classification in the Streaming Model: Improved Bounds for Heavy Hitters
Arvind V. Mahankali, David Woodruff
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 |
|---|---|---|---|
| keep only the top K=400 weights instead of the full weight vector (ablation) | online logistic regression classifier · RCV1 test half | accuracy: 95.7% -> 93.9% | Hurt |
| Algorithm 2 (JL point query) instead of WM-Sketch, 16 KB, K=120 | l2 point query sketch · RCV1, lambda=1e-5 | relative error of top-weight recovery: roughly 15% improvement | Improved |
| WM-Sketch instead of Algorithm 1 at 16 KB, K=120 | l2 point query sketch · RCV1, lambda=1e-5 | relative error of top-weight recovery | Improved |
| Algorithm 1 instead of WM-Sketch at small memory (2 KB and 4 KB) | l2 point query sketch · RCV1, lambda=1e-5 | relative error of top-weight recovery | Improved |
| increase memory budget (JL/point query sketch vs WM-Sketch) | combined JL/point query sketch · RCV1 | weight estimation error | Improved |
Closest papers by meaning
- New Bounds for Kernel Sums via Fast Spherical Embeddings (ICML 2026 · similarity 0.85)
- Streaming Algorithms For $\ell_p$ Flows and $\ell_p$ Regression (ICLR 2025 · similarity 0.85)
- Space-Optimal Streaming Algorithms via Efficient Encodings (NeurIPS 2026 · similarity 0.85)
- High-Dimensional Geometric Streaming for Nearly Low Rank Data (ICML 2024 · similarity 0.83)
- Beyond 1/2-Approximation for Submodular Maximization on Massive Data Streams (ICML 2018 · similarity 0.83)
- Adversarially Robust Dense-Sparse Tradeoffs via Heavy-Hitters (NeurIPS 2024 · 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.
