Demographic history describes how populations evolved over time.

It includes events such as:

• population size changes
• splits
• migrations
• admixture
• selection

Understand population history

Conservation biology studies

There are some known examples of non-identifiable models:

However, generally classes are unknown

Generative AI models learn the underlying probability distribution of the training data, enabling them to generate new samples with similar characteristics.

• Variational autoencoder for drug design [Gangwal et al. 2024]:
• Discrete diffusion-like model for Bayesian sampling of phylogenetic trees [Zhou et al. 2024]

Train generative AI model to explore classes of non-identifiable demographic histories.

Once trained, a generative AI model enables a fast sampling procedure.

It allows to address various problems:

• Unravel complex admixture history of domesticated species (cattle)
• Investigate the regions that act as barriers to gene flow in recently diverged species (butterfly)