Discovery of Single Independent Latent Variable

Abstract

Latent variable discovery is a central problem in data analysis with a broad range of applications in applied science. In this work, we consider data given as an invertible mixture of two statistically independent components, and assume that one of the components is observed while the other is hidden. Our goal is to recover the hidden component. For this purpose, we propose an autoencoder equipped with a discriminator. Unlike the standard nonlinear ICA problem, which was shown to be non-identifiable, in the special case of ICA we consider here, we show that our approach can recover the component of interest up to entropy-preserving transformation. We demonstrate the performance of the proposed approach in several tasks, including image synthesis, voice cloning, and fetal ECG extraction.

Links

Audio Samples

We present a set of the conversions generated by our method when converting seen speakers i.e. found in our method training data. For each conversion we display the source sample from the seen speaker, a sample recorded by the target speaker and finally the generated sample by applying the Voice Conversion.

Citation

@inproceedings{shaham2022,
  author = {Uri Shaham and Jonathan Svirsky and Ori Katz and Ronen Talmon},
  title = {Discovery of Single Independent Latent Variable},
  booktitle = {Proc. NeurIPS},
  year = {2022}
}

License

MIT License. Feel free to use any of the material in your own work, as long as you give us appropriate credit by mentioning the title and author list of our paper.

Acknowledgments

We thank Neurips 2022 reviewers assigned to our paper for their helpful comments and discussions.