polysetlib 1.0.0

PolySet – Python library for generating distribution-aware polymer embeddings

PolySet – Python Library for Polymer Embeddings

A representation framework for polymers seen as ensembles rather than idealized single objects.

Why PolySet?

Real polymer materials are never single molecules. They are populations of chains — each with its own length, each contributing differently to the properties of the bulk material.

Current machine-learning approaches, however, typically reduce a polymer to a single symbolic structure, ignoring its internal statistical nature.

PolySet restores this missing dimension by representing a polymer as a finite ensemble of weighted chains consistent with its macroscopic descriptors.

PolySet changes the input — how the polymer is perceived by the model, not how the model learns.

What PolySet Does?

Given basic information such as the number-average molar mass (Mn), molar mass of the repeat unit (M0) and dispersity (Đ), PolySet constructs a representative ensemble of chains that respects these values and can be used as input for any downstream ML encoder.

PolySet optionally includes four sampling modes (basic, short, middle, long) that allow users to explore different plausible tail behaviors when only Mn and Đ are available. These are implementation-level options and do not modify the core philosophy of PolySet

Installation

pip install polysetlib

Documentation & Examples

More information and conceptual background are available here: https://khalidferji.academicwebsite.com/pages/polyset.

Citation

If PolySet contributes to your work, please cite:

ferji, khalid . (2025). PolySet: Statistical Ensemble Polymer Embeddings Dataset for Machine Learning [Data set]. Zenodo. https://doi.org/10.5281/zenodo.17861022.