Research map / Learning theory and algorithms / Prediction and decision losses
An online passive-aggressive algorithm for difference-of-squares classification
Lawrence K. Saul
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 hidden dimensionality d of DoS classifier from 1 to 16 (ablation) | DoS passive-aggressive classifier · INFIMNIST online training (100M examples) | cumulative and moving training error rate | Improved |
| use DoS model with d in {2,4,8,16} instead of linear passive-aggressive model | linear PA classifier · INFIMNIST online training (100M examples) | error rate | Improved |
| use rank-one DoS (d=1) instead of linear model | linear PA classifier · INFIMNIST online training | error rate | Hurt |
| average models across time with symmetry-aligned averaging (ablation) | DoS classifiers · INFIMNIST test set | test error rate | Improved |
| smaller hidden dimensionality d (ablation) | DoS classifier · INFIMNIST non-averaged test error | test error variability | Hurt |
| larger model (d=16) trained longer (ablation) | DoS classifier · INFIMNIST | error rate vs training examples | Improved |
| vary random seed | DoS classifier · first 10M INFIMNIST examples | error rate | No clear effect |
Closest papers by meaning
- Stabilizing Linear Passive-Aggressive Online Learning with Weighted Reservoir Sampling (NeurIPS 2024 · similarity 0.84)
- Online Linear Classification with Massart Noise (ICML 2025 · similarity 0.83)
- QPM: Discrete Optimization for Globally Interpretable Image Classification (ICLR 2025 · similarity 0.83)
- R2Det: Exploring Relaxed Rotation Equivariance in 2D Object Detection (ICLR 2025 · similarity 0.82)
- Diffusion-Classifier Synergy: Reward-Aligned Learning via Mutual Boosting Loop for FSCIL (NeurIPS 2025 · similarity 0.82)
- Decoupled Optimisation for Long-Tailed Visual Recognition (AAAI 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.
