Modeling Variance: A Variance-Motivated Approach to Molecular Prediction

In this variance-motivated study, the variance of a multi-thousand molecular dataset of Coulomb matrices is analyzed. This paper presents novel statistical methods and models that can aide in molecular prediction and analysis. A blended statistical/ML model is introduced for classifying data as Normal as well as a model for visualizing variance. Linear regression is also used to show a potential simple and 1 dimensional molecular descriptor, for some molecules. Paper includes literature review.