Self-Supervised Noise Embeddings (Self-SNE) for dimensionality reduction and clustering

This is an alpha release currently undergoing development. Examples and documentation will be added upon release of the accompanying publication. Not all features have been validated and may change without notice. Use at your own risk.

Self-SNE is a probabilistic family of self-supervised deep learning models for compressing high-dimensional data to a low-dimensional embedding. It is a general-purpose algorithm that works with multiple types of data including images, sequences, and tabular data. It uses self-supervised objectives to preserve structure in the compressed latent space. Self-SNE can also (optionally) simultaneously learn a cluster distribution (a prior over the latent embedding) during optimization.

References

If you use Self-SNE for your research please cite version 1 of our preprint (an updated version is forthcoming):

@article{graving2020vae,
	title={VAE-SNE: a deep generative model for simultaneous dimensionality reduction and clustering},
	author={Graving, Jacob M and Couzin, Iain D},
	journal={BioRxiv},
	year={2020},
	publisher={Cold Spring Harbor Laboratory}
}

License

Released under a Apache 2.0 License. See LICENSE for details.