Research map / Optimization / Implicit bias and generalization
Multivariate Boosted Trees and Applications to Forecasting and Control
Lorenzo Nespoli, Vasco Medici
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 |
|---|---|---|---|
| use multivariate boosted trees with Fourier output basis instead of LightGBM MIMO | LightGBM MIMO · Grid load forecasting (31 series) | RMSE | Improved |
| use multivariate boosted trees with Fourier basis instead of LightGBM MISO | LightGBM MISO · Grid load forecasting (31 series) | RMSE and MAPE: 14 time series in RMSE, 11 in MAPE | Mixed |
| increase number of Fourier harmonics beyond the best value (ablation) | MBT forecaster · Grid load forecasting (31 series) | RMSE and MAPE | No clear effect |
| add Fourier smoothness regularization to MBT (ablation) | MBT · M4 hourly (414 series) | MAPE and RMSE | Improved |
| use MBT instead of LightGBM MIMO/MISO without exogenous features | MBT · M4 hourly (414 series) | MAPE and RMSE | Improved |
| add previous-step forecast errors as MBT inputs instead of minT reconciliation | MBT · Hierarchical load forecasting (31 series) | RMSE | Mixed |
| use MBT instead of least squares VSC | Least-squares VSC · Voltage sensitivity (simulated Swiss LV grid, 10-fold CV) | RMSE | Improved |
| use PCC power as MBT feature instead of meteorological variables only (ablation) | MBT · Voltage sensitivity (simulated Swiss LV grid) | RMSE | Improved |
| use multivariate boosted trees instead of separate LightGBM per quantile | LightGBM per quantile · Hourly aggregated load, 11 quantiles | quantile score Qs | Improved |
| use multivariate boosted trees instead of separate LightGBM per quantile | LightGBM per quantile · Hourly aggregated load, 11 quantiles | mean quantile crossing | Improved |
| add quantile refitting to linear quantile loss MBT (ablation) | MBT linear quantile loss · Hourly aggregated load, 11 quantiles | quantile loss | No clear effect |
Closest papers by meaning
- Boulevard: Regularized Stochastic Gradient Boosted Trees and Their Limiting Distribution (JMLR 2022 · similarity 0.88)
- Regression Trees Know Calculus (NeurIPS 2025 · similarity 0.87)
- Tree-Based Models for Correlated Data (JMLR 2022 · similarity 0.87)
- Pushing the Envelope of Gradient Boosting Forests via Globally-Optimized Oblique Trees (CVPR 2022 · similarity 0.86)
- Minimax Rates and Spectral Distillation for Tree Ensembles (NeurIPS 2026 · similarity 0.84)
- Statistical Inference for Gradient Boosting Regression (NeurIPS 2025 · similarity 0.84)
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.
