Modeling the optical properties of biological structures using symbolic regression

Area of Science:

• Biophysics
• Materials Science
• Computational Science

Background:

• Accurate modeling of biological material optical properties is crucial for various applications.
• Existing models may lack physical interpretability or struggle with diverse spectral data.
• Developing methods to derive closed-form expressions for optical properties is an ongoing challenge.

Purpose of the Study:

• To present a machine learning approach based on symbolic regression for deriving closed-form expressions of biological material optical properties.
• To evaluate the performance of this approach using spectral data (numerical and experimental).
• To retrieve the refractive index of constituent materials in biological structures.

Main Methods:

• Utilized symbolic regression, a machine learning technique, to analyze spectral data.
• Applied the method to both numerically generated and experimentally measured spectral data.
• Conducted three case studies focused on retrieving the refractive index.

Main Results:

• Successfully derived readable and dimensionally homogeneous dispersion models.
• The obtained expressions possess physical meaning.
• The algebraic form of the derived models aligns with established models for transparent dielectrics in the visible spectrum.

Conclusions:

• The symbolic regression approach effectively models optical properties of biological materials.
• The derived models are physically interpretable and comparable to conventional methods.
• This technique offers a powerful tool for characterizing the dispersive behavior of biological materials.