Polyatomic Complexes: A topologically-informed learning representation for atomistic systems

TL;DR

This paper introduces a new topologically-informed representation for atomistic systems that satisfies key structural constraints, offers a general encoding algorithm, and achieves performance comparable to state-of-the-art methods.

Contribution

It presents a novel, theoretically grounded representation for chemical structures that is generalizable and efficient, with open-source code and datasets.

Findings

Satisfies all structural, geometric, efficiency, and generalizability constraints

Provides a universal algorithm for encoding atomistic systems

Achieves performance comparable to current leading methods

Abstract

Developing robust representations of chemical structures that enable models to learn topological inductive biases is challenging. In this manuscript, we present a representation of atomistic systems. We begin by proving that our representation satisfies all structural, geometric, efficiency, and generalizability constraints. Afterward, we provide a general algorithm to encode any atomistic system. Finally, we report performance comparable to state-of-the-art methods on numerous tasks. We open-source all code and datasets. The code and data are available at https://github.com/rahulkhorana/PolyatomicComplexes.