lpspline 0.1.11

lpspline

LPSpline

LPSpline is a Python package for building and optimizing linear spline models using an intuitive additive API. It provides a flexible way to model non-linear relationships using various spline types like Piecewise Linear, B-Splines, Cyclic Splines, and Categorical Factors.

Documentation

Visit the documentation for a complete overview and some examples with plots.

Installation

Install lpspline via pip directly from the repository, or if published:

pip install lpspline

Main Features

• Additive model definition
• CVXPY backend for optimization
• Multiple spline types: Linear, Piecewise Linear, B-Splines, Cyclic Splines, Categorical Factors, Constant
• Penalties on the splines: Ridge, Lasso
• Constraints on the splines: Monotonic, Convex, Concave, Anchor, Bound
• Save and load models with a single line of code
• Polars DataFrame integration
• Nice plots using matplotlib and pimpmyplot

Sandbox

Visit the marimo playground for a live demo

Quick Start

Here a small code example:

import numpy as np
import polars as pl
from lpspline import l, pwl, bs, cs
from lpspline.viz import plot_diagnostic
from lpspline.datasets import load_demo_dataset


# ---------------------------------------- Data Generation
X, y = load_demo_dataset(samples = 1000)

# ---------------------------------------- Model Definition
model = (
    +l(term='xl', by='xfactor')
    + pwl(term='xpwl', knots=3)
    + bs(term="xbs", knots=10, degree=2)
    + cs(term="xcyc", order=3)
    + f(term="xfactor")
)

# ---------------------------------------- Model Fitting
model.fit(X, y)

# ---------------------------------------- Model Prediction
predictions = model.predict(X)

# ---------------------------------------- Model Visualization
plot_diagnostic(model=model, X=X, y=y, ncols=3)

Expected output

Once the model is fitted, you will see a detailed summary to the console and a diagnostic plot showing the fitted splines.

========================================================================================================================
✨ Model Summary ✨
========================================================================================================================
Problem Status: ✅ optimal
------------------------------------------------------------------------------------------------------------------------
Spline Type          | Term         | Tag             | Constraints          | Penalties            | Params
------------------------------------------------------------------------------------------------------------------------
🟢 Linear            | xl           | linear          | None                 | None                 | 6       
🟢 PiecewiseLinear   | xpwl         | pwl             | None                 | None                 | 5       
🟢 BSpline           | xbs          | bspline         | None                 | None                 | 11      
🟢 CyclicSpline      | xcyc         | cyclicspline    | None                 | None                 | 7       
🟢 Factor            | xfactor      | factor          | None                 | None                 | 3       
------------------------------------------------------------------------------------------------------------------------
📊 Total Parameters                                                                                 | 32
========================================================================================================================

Model fitted successfully.