Inferring colloidal interaction from scattering by machine learning

A machine learning solution for the potential inversion problem in elastic scattering is outlined. The inversion scheme consists of two major components, a generative network featuring a variational autoencoder which ex-tracts the targeted static two-point correlation functions from experimentally measured scattering cross sections, and a Gaussian process framework which probabilistically infers the relevant structural parameters from the in-verted correlation functions. Via a case study of charged colloidal suspensions, the feasibility of this approach for quantitative study of molecular interaction is critically benchmarked and its merit over existing deterministic approaches, in terms of numerical accuracy and computationally efficiency, is demonstrated.

Automated summary

A machine learning solution is proposed to solve the potential inversion problem in elastic scattering. The proposed solution consists of a generative network using a variational autoencoder to extract the targeted static two-point correlation functions from experimentally measured scattering cross sections, and a Gaussian process framework for probabilistically inferring the relevant structural parameters. A case study on charged colloidal suspensions is used to critically evaluate the feasibility of this approach for quantitative study of molecular interaction and demonstrate its advantages over existing deterministic approaches in terms of numerical accuracy and computational efficiency.