Bayesian Gaussian networks for multidimensional classification of morphologically characterized neurons in the NeuroMorpho repository

Abstract

A class-bridge decomposable multidimensional Gaussian net- work is presented as an interpretable and powerful model, to account for the morphological di erences that exist between di erent neurons when varying the species, gender, brain region, cell types and developmental stage of the animal of origin. Also this work includes a learning algorithm that makes use of the CB-decomposablility property to alleviate the inference complexity and use it to learn complex network structures that take into account relationships between classes. The model is trained with data from NeuroMorpho (v5.7) and the nal model is used to test the predictive power of the learning algorithm for Bayesian networks and, given its interpretability, to extract knowledge at a neuroscience lev