Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Learning Non-Linear Dynamics of Decision Boundaries for Maintaining Classification Performance
Atsutoshi Kumagai, Tomoharu Iwata
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 |
|---|---|---|---|
| Model dynamics of each classifier parameter with a Gaussian process instead of linear autoregression (AAAI16) | future-classifier model · Linear | AUC: 0.940 -> 0.988 | Improved |
| Use GP dynamics instead of vector autoregression | future-classifier model · Non-linear | AUC: 0.952 -> 0.984 | Improved |
| Use GP dynamics instead of vector autoregression | future-classifier model · SPAM2 | AUC: 0.907 -> 0.956 | Improved |
| Use GP dynamics instead of vector autoregression | future-classifier model · ELEC2 | AUC: 0.926 -> 0.921 | Hurt |
| Use GP dynamics instead of vector autoregression | future-classifier model · ONP | AUC: 0.533 -> 0.705 | Improved |
| Use GP dynamics instead of vector autoregression | future-classifier model · BLOG | AUC: 0.827 -> 0.838 | Improved |
| Predict future classifier vs. batch logistic regression trained on all data | future-classifier model · Non-linear | AUC: 0.302 -> 0.984 | Improved |
| Predict future classifier vs. present-only classifier | future-classifier model · BLOG | AUC: 0.754 -> 0.838 | Improved |
| Predict future classifier vs. online logistic regression | future-classifier model · BLOG | AUC: 0.730 -> 0.838 | Improved |
| Vary the number of training data per time unit | future-classifier model · SPAM2 | AUC | No clear effect |
Closest papers by meaning
- Strategic Classification with Non-Linear Classifiers (NeurIPS 2025 · similarity 0.83)
- Leveraging Uncertainty Estimates To Improve Classifier Performance (ICLR 2024 · similarity 0.83)
- Provably Reliable Classifier Guidance via Cross-Entropy Control (NeurIPS 2026 · similarity 0.82)
- Learning the Optimal Stopping for Early Classification within Finite Horizons via Sequential Probability Ratio Test (ICLR 2025 · similarity 0.82)
- Probing the Decision Boundaries of In-context Learning in Large Language Models (NeurIPS 2024 · similarity 0.82)
- On the Training Convergence of Transformers for In-Context Classification of Gaussian Mixtures (ICML 2025 · similarity 0.81)
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.
