EvoNN & EVE

Learning from phylogenetic trees

Phylogenetic trees contain information about the processes that produced biological diversity. My work combines neural inference with stochastic diversification models to study those processes.

For EvoNN, I designed ensemble neural networks that learn from tree structure, branching times and summary statistics. Alongside this inference work, I developed models and fast simulators for diversification influenced by evolutionary relatedness.

Models, inference and software

The work connects biological questions with simulation, model fitting and evaluation. EvoNN provides neural inference from phylogenies; evesim supports simulation of evolutionary relatedness dependent birth–death processes. The modelling and software work connects diversification assumptions to simulations used for inference.

The associated publications and software links are listed below. They include published work on parameter estimation and on the effects of evolutionary relatedness, together with a separate preprint on identifying relatedness effects from phylogenies.