Workshop 2: Decoding the Past: Deep Learning for Macroevolutionary Analyses

Date: July 27, 2025

Location: Univeristy of Zurich, specific room TBD

Mode: This workshop will be held in person.

Length: Full day

Participant cap:  ~30, aimed at students and early career researchers

Workshop leaders:

• Daniele Silvestro, University of Fribourg silvestro.daniele@gmail.com  
• Torsten Hauffe, University of Fribourg torsten.hauffe@gmail.com
• Juan L. Cantalapiedra, Museo Nacional de Ciencias Naturales, Madrid jlopezcant@gmail.com
• Fernando Blanco, University of Gothenburgfblancosegovia@gmail.com
• Rebecca Cooper, University of Fribourg rebecca.cooper@unifr.ch

Description:

Understanding how biodiversity has changed through geologic time and which mechanisms mediate the uneven extinction and diversification of species are central challenges in paleobiology, particularly due to the incomplete nature of the fossil record. Deep learning is a powerful tool that can assist paleontologists in estimating past biodiversity and exploring the different factors that could drive dynamics. In this workshop, we present two new methods based on deep learning: (1) Deep Dive for estimating past biodiversity from fossil occurrence data incorporating spatial, taxonomic, and temporal biases, and (2) birth-death neural networks (BDNN) in PyRate for exploring the differential effects of abiotic and biotic variables on the diversification rates of extinct species without the necessity to include phylogenies in the model. At the end of the workshop, participants will understand the theoretical background of the approaches and be able to run fossil biodiversity analysis, diversification (extinction and speciation) rate estimations, and variable testing over diversification rates using both software packages and will have the necessary tools to customize analyses.