Research reproduction · PyTorch
PhyloVAE
A PyTorch implementation and experiment around variational autoencoders for phylogenetic tree-topology data, based on the PhyloVAE paper presented as an ICLR 2025 poster.
Approach
Work with tree-topology representations
The repository documents construction of tree-topology encodings and node embeddings, training a variational autoencoder, and producing representations for downstream analysis.
Reproduce and measure against a baseline
The CV reports about an 18% reduction in reconstruction loss versus a baseline autoencoder. The public repository README credits the original PhyloVAE paper authors; the result is presented here as an implementation and experiment based on that research, not original authorship of the paper.
Outcome
An ML implementation project with a CV-reported reconstruction-loss improvement, linked to the original ICLR 2025 paper and its authors.
- Python
- PyTorch
- Variational autoencoders
- Phylogenetics