Tree Variational Autoencoders
About
We propose Tree Variational Autoencoder (TreeVAE), a new generative hierarchical clustering model that learns a flexible tree-based posterior distribution over latent variables. TreeVAE hierarchically divides samples according to their intrinsic characteristics, shedding light on hidden structures in the data. It adapts its architecture to discover the optimal tree for encoding dependencies between latent variables. The proposed tree-based generative architecture enables lightweight conditional inference and improves generative performance by utilizing specialized leaf decoders. We show that TreeVAE uncovers underlying clusters in the data and finds meaningful hierarchical relations between the different groups on a variety of datasets, including real-world imaging data. We present empirically that TreeVAE provides a more competitive log-likelihood lower bound than the sequential counterparts. Finally, due to its generative nature, TreeVAE is able to generate new samples from the discovered clusters via conditional sampling.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Cellular Lineage Inference | Limb (cell lineage) | DP34.7 | 14 | |
| Cellular Lineage Inference | Weinreb (cell lineage) | DP60.4 | 12 | |
| Cellular Lineage Inference | LHCO cell lineage | DP38.3 | 12 | |
| Hierarchical Generative Modeling | Fashion MNIST | DP53.4 | 7 | |
| Hierarchical Generative Modeling | MNIST | DP87.9 | 7 | |
| Hierarchical Generative Modeling | 20 Newsgroups text | DP17.5 | 7 | |
| Hierarchical Generative Modeling | Cifar10 32x32 (50k samples) | DP35.3 | 6 | |
| Lineage Inference | LineageVAE Day 4 | Ratio of Observed Time Points30.4 | 5 | |
| Lineage Inference | LineageVAE Day 2 | Ratio of Observed Time Points12.1 | 5 | |
| Lineage Inference | LineageVAE Day 6 | Ratio of Observed Time Points56.4 | 5 |