Domain-Specific Knowledge Integration Through Curriculum Learning Method in Machine Learning for Spectral Analysis

Abstract

We introduce a novel enhancement to machine learning models for predicting UV spectra from molecular structures represented by SMILES strings. By integrating domain-specific knowledge through a curriculum learning method, the predictive performance of the models is improved. The curriculum learning approach allows the model to progressively learn from simpler to more complex spectral patterns, enabling better performance. This advancement demonstrates the potential of domain knowledge implemented machine learning to enhance spectral analysis and its applications in drug discovery and materials science.