Research map / Optimization / Optimizers
Cogra: Concept-Drift-Aware Stochastic Gradient Descent for Time-Series Forecasting
Kohei Miyaguchi, Hiroshi Kajino
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 |
|---|---|---|---|
| replace vSGD moment estimator with SMT (Cogra-b vs vSGD-b) (ablation) | vSGD · real-world datasets (A(1)-A(3), EEG, GAS) | relative cumulative loss: 1.00 -> 0.328 on GAS; 1.00 -> 0.612 on EEG; 1.00 -> 0.797 on A(1) | Improved |
| replace vSGD moment estimator with SMT (ablation) | vSGD · AR(0) mean tracking | squared estimation error: 60% lower error | Improved |
| use one global learning-rate block instead of blocks (ablation) | Cogra · real-world datasets | relative cumulative loss: 0.612 -> 1.47e+9 (EEG) | Hurt |
| use local per-scalar learning rates instead of blocks (ablation) | Cogra · EEG and GAS | relative cumulative loss: 0.612 -> 4.32e+4 (EEG); 0.328 -> 7.87e+6 (GAS) | Hurt |
| use local per-scalar learning rates instead of blocks (ablation) | Cogra · A(1) activity data (low dimension) | relative cumulative loss: 0.797 -> 0.903 | Mixed |
Closest papers by meaning
- Tackling Time-Series Forecasting Generalization via Mitigating Concept Drift (ICLR 2026 · similarity 0.88)
- Probabilistic Forecasting with Stochastic Interpolants and Föllmer Processes (ICML 2024 · similarity 0.84)
- Neural MJD: Neural Non-Stationary Merton Jump Diffusion for Time Series Prediction (NeurIPS 2025 · similarity 0.83)
- Multivariate Probabilistic Time Series Forecasting with Correlated Errors (NeurIPS 2024 · similarity 0.83)
- LangTime: A Language-Guided Unified Model for Time Series Forecasting with Proximal Policy Optimization (ICML 2025 · similarity 0.83)
- Time-Series Forecasting for Out-of-Distribution Generalization Using Invariant Learning (ICML 2024 · similarity 0.83)
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.
