Gradient Boosting Model
An ensemble learning method that minimizes a loss function by progressively adding weak learners, usually decision trees; common implementations include XGBoost, LightGBM and CatBoost.
Timeline (last 90 days)
Timestral still outperforms the original gradient boosting model in the corrected backtest: MASE 0.66 versus 0.71, but the improvement is far smaller than the initially reported 0.49.
On the temporal split test set (878 comparisons from 655 trials), a gradient-boosted emulator trained on registry data eliminated about one-fifth of the error of a simple baseline.
The gradient-boosted model's intervention skill value is -0.12 (95% CI -0.21 to -0.05), indicating that the model performs better when intervention information is hidden.
The study tested alternative featurization, stratification, temporal cutting, and same-molecule retrieval settings, and none of them yielded a confidence interval for intervention skill above zero.
All Facts (4)
Timestral still outperforms the original gradient boosting model in the corrected backtest: MASE 0.66 versus 0.71, but the improvement is far smaller than the initially reported 0.49.
50%UnverifiedOn the temporal split test set (878 comparisons from 655 trials), a gradient-boosted emulator trained on registry data eliminated about one-fifth of the error of a simple baseline.
50%UnverifiedThe gradient-boosted model's intervention skill value is -0.12 (95% CI -0.21 to -0.05), indicating that the model performs better when intervention information is hidden.
50%UnverifiedThe study tested alternative featurization, stratification, temporal cutting, and same-molecule retrieval settings, and none of them yielded a confidence interval for intervention skill above zero.
50%