[P] PerpetualBooster v1.9.0 - GBM with no hyperparameter tuning, now with built-in causal ML, drift detection, and conformal prediction

2026-02-26 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, quick

Summary

PerpetualBooster v1.9.0 is a gradient boosting machine (GBM) implemented in Rust that eliminates traditional hyperparameter tuning, replacing it with a single `budget` parameter. This update significantly expands its capabilities, integrating a full suite of causal machine learning estimators, including Double Machine Learning, meta-learners (S/T/X), uplift (R-learner), instrumental variables, policy learning, and fairness-aware objectives. The new version also features built-in data and concept drift detection, which operates without requiring ground truth labels or retraining. For calibration, it offers conformalized quantile regression (CQR) for prediction intervals and isotonic calibration for classification. PerpetualBooster now supports 19 objectives for regression, classification, and ranking, along with multi-output capabilities via `MultiOutputBooster` and improved O(n) continual learning. Benchmarks show it matches Optuna + LightGBM accuracy with up to a 405x wall-time speedup and outperforms AutoGluon v1.2 on 18/20 OpenML tasks, inferring up to 5x faster and avoiding out-of-memory errors.

Key takeaway

For MLOps Engineers and AI Scientists seeking to streamline model deployment and enhance robustness, PerpetualBooster offers a compelling alternative to traditional GBMs. Its "budget" parameter simplifies tuning, while integrated causal ML, drift detection, and conformal prediction capabilities reduce the need for separate tools. You should consider evaluating PerpetualBooster for projects requiring fast, accurate, and interpretable models, especially where hyperparameter optimization is a bottleneck or causal inference is critical.

Key insights

PerpetualBooster simplifies GBM deployment by replacing hyperparameter tuning with a single budget parameter.

Principles

• Budget-based generalization replaces hyperparameter tuning.
• Causal estimators can leverage budget-based generalization.

Method

The PerpetualBooster algorithm uses a single `budget` parameter for generalization, integrating causal ML, drift detection, and conformal prediction directly into its Rust core.

In practice

• Use `budget` parameter instead of hyperparameter tuning.
• Apply CQR for robust prediction intervals.
• Detect drift using trained tree structure.

Topics

• Gradient Boosting Machines
• Causal Machine Learning
• Conformal Prediction
• Drift Detection
• Hyperparameter Optimization

Code references

• perpetual-ml/perpetual

Best for: MLOps Engineer, AI Scientist, Research Scientist, Machine Learning Engineer, Data Scientist, AI Engineer

Related on AIssential

• First-Principles Optimizer Matches Adam on CIFAR…No Tuning — A first-principles optimizer dynamically adapts learning rates, matching Adam's performance without hyper-parameter tuning.
• Where CatBoost beats XGBoost and LightGBM — and what the book is honest about — CatBoost leads GBDT rankings on diverse tabular datasets, excelling with categorical features due to its unique architectural decisions.
• Choosing the Right Model is Hard. Maintaining Accuracy is Harder. — Continuous monitoring and iterative refinement of open-source LLMs in production can enhance accuracy and efficiency.
• Time-Aware Latent Space Bayesian Optimization — TALBO adapts latent-space Bayesian optimization to dynamic objectives by integrating time into both the surrogate and generative representation.
• Radial Suppression Accelerates Algorithmic Generalization: A Geometric Analysis of Delayed Generalization — Radial inflation of hidden representations drives memorization-generalization delays in neural networks.
• Online Conformal Prediction: Enforcing monotonicity via Online Optimization — New online conformal prediction methods ensure nested prediction sets across multiple confidence levels.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.