diPLSlib documentation

Introduction

diPLSlib is a Python library designed for domain adaptation in multivariate calibration, with a focus on privacy-preserving regression and calibration model maintenance. It provides a scikit-learn compatible API and implements advanced methods for aligning data distributions across different domains, enabling robust and transferable regression models.

The library features several state-of-the-art algorithms, including:

• Domain-Invariant Partial Least Squares (di-PLS/mdi-PLS): Aligns feature distributions between source and target domains to improve model generalization.

• Graph-based Calibration Transfer (GCT-PLS): Minimizes discrepancies between paired samples from different domains in the latent variable space.

• Kernel Domain Adaptive PLS (KDAPLS): Projects data into a reproducing kernel Hilbert space for non-parametric domain adaptation.

• Differentially Private PLS (EDPLS): Ensures privacy guarantees for sensitive data using the \((\epsilon, \delta)\)-differential privacy framework.

diPLSlib is suitable for chemometrics, analytical chemistry, and other fields where robust calibration transfer and privacy-preserving modeling are required. For more details, usage examples, and API documentation, please refer to the sections below.

Contents:

• diPLSlib package
  • diPLSlib.functions module
  • diPLSlib.models module
  • diPLSlib.utils subpackage
  • Module contents