tensorsl 0.2.0

Tensor Separation Learning: interpretable regression via sums of rank-1 separable components.

tensorsl

Tensor Separation Learning (TSL) — a glass-box regression model. A fitted model is a sum of stages, where each stage is the ordered difference of two non-negative rank-1 products of univariate functions. This keeps the model inspectable while still capturing interactions between features.

tensorsl is the Python package (imported as tsl_py): a fit/predict estimator that slots into typical ML pipelines, backed by a Rust core.

Install

pip install tensorsl

Prebuilt wheels are published for Linux, macOS, and Windows, so no toolchain is needed. To build from source instead, you need a Rust toolchain (the core is pure Rust — no system math libraries required).

Usage

import numpy as np
from tsl_py import TSLRegressor

rng = np.random.default_rng(0)
X = rng.normal(size=(500, 5))
y = X[:, 0] * X[:, 1]

model = TSLRegressor().fit(X, y)
preds = model.predict(X)

The fitted model also exposes per-feature partial dependence and feature importances for interpretation.

Links

• Documentation: https://jyliuu.github.io/TSL/
• Source: https://github.com/jyliuu/TSL

License

MIT