Modeling the Optical Properties of Biological Structures using Symbolic Regression

TL;DR

This paper introduces a symbolic regression-based machine learning method to derive interpretable models of biological optical properties from spectral data, successfully retrieving physically meaningful dispersion models.

Contribution

The study presents a novel application of symbolic regression to model biological optical properties, producing interpretable and physically consistent dispersion equations.

Findings

Successfully retrieved dispersion models with physical meaning

Expressions are similar to known dielectric models

Method effective across multiple biological structures

Abstract

We present a Machine Learning approach based on Symbolic Regression to derive, from either numerically generated or experimentally measured spectral data, closed-form expressions that model the optical properties of biological materials. To evaluate the performance of our approach, we consider three case studies with the aim of retrieving the refractive index of the materials that constitute the biological structures considered. The results obtained show that, in addition to retrieving readable and dimensionally homogeneous dispersion models, the expressions found have a physical meaning and their algebraic form is similar to that of the models used to characterize the dispersive behavior of transparent dielectrics in the visible region.