Repository navigation
v0.1.0
🚀 JAXBoost v0.1.0
Initial Release - Differentiable gradient boosting in JAX
Highlights
JAXBoost is a gradient boosting implementation using soft (differentiable) oblivious trees. The entire model is trained end-to-end with gradient descent, rather than traditional greedy tree-building.
✨ Features
Soft Oblivious Trees - Differentiable tree structures with sigmoid routing
Hyperplane Splits - Linear combinations of features for richer decision boundaries
End-to-End Training - Gradient-based optimization via optax
GPU Acceleration - Fully vectorized JAX operations
This is an alpha release for learning and experimentation. Not intended for production use.
Full Changelog: https://github.com/jxucoder/jaxboost/commits/v0.1.0