Eliminating subjectivity, quantifying uncertainty, and using machine learning for phylogenetic inference

Abstract

In this talk I will outline our attempts to quantify the uncertainty in phylogenetic data analysis pipelines and how we predict the degree of difficulty of a phylogenetic analysis prior to conducting actual likelihood-based inferences. I will also show how this predicted difficulty can be deployed for accelerating phylogenetic Maximum Likelihood search algorithms. Time permitting, I will also outline our preliminary experiments to predict the difficulty of the Multiple Sequence Alignment task and how we can eliminate subjectivity in the assembly process of natural language datasets with the aim to reconstruct phylogenies of natural languages. I will conclude with an overview of other research activities in our group.