paper.md

title	tags	authors	affiliations	date	bibliography
ivis: dimensionality reduction in very large datasets using Siamese Networks	dimensionality reduction unsupervised learning neural network	name affiliation Benjamin Szubert 1 name affiliation orcid Ignat Drozdov 1 0000-0001-6727-4688	name index Bering Limited 1	18 July 2019	paper.bib

Summary

ivis is a dimensionality reduction technique that implements a Siamese Neural Network architecture trained using a novel triplet loss function. Results on simulated and real datasets demonstrate that ivis preserves global data structures in a low-dimensional space and adds new data points to existing embeddings using a parametric mapping function.

ivis is easily integrated into standard machine learning pipelines through a scikit-learn compatible API and scales well to out-of-memory datasets. Both supervised and unsupervised dimensionality reduction modes are supported.

Further information on the algorithm and its application to single cell datasets can be found in [@ivis_scirep]. Implementation of the ivis algorithm is available on GitHub.

Acknowledgements

This work was supported by funding from the European Commission’s Seventh Framework Programme [FP7-2007-2013] under grant agreement n°HEALTH-F2-2013-602114 (Athero-B-Cell).

References