Lattice
A TensorFlow library implementing constrained and interpretable lattice-based models with shape constraints like monotonicity and convexity.
What is Lattice?
TensorFlow Lattice is a library that implements constrained and interpretable lattice-based models in TensorFlow. It provides Keras layers and canned estimators that allow injection of domain knowledge through shape constraints like monotonicity, convexity, and pairwise trust, helping models extrapolate better when serving distributions differ from training data.
Machine learning practitioners and data scientists working with TensorFlow who need interpretable models with guaranteed constraints, particularly in domains where domain knowledge or policy-driven requirements must be enforced.
Developers choose TensorFlow Lattice for its ability to enforce common-sense shape constraints directly in the model architecture, ensuring predictable behavior and better extrapolation, which is especially valuable in high-stakes applications where model interpretability and reliability are critical.
Overview
Lattice methods in TensorFlow
Use Cases
Best For
- Enforcing monotonic relationships in model predictions (e.g., higher income should not decrease credit score)
Related Projects
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
