Self-supervised similarity search for large scientific datasets
arXiv:2110.13151
Abstract
We present the use of self-supervised learning to explore and exploit large unlabeled datasets. Focusing on 42 million galaxy images from the latest data release of the Dark Energy Spectroscopic Instrument (DESI) Legacy Imaging Surveys, we first train a self-supervised model to distill low-dimensional representations that are robust to symmetries, uncertainties, and noise in each image. We then use the representations to construct and publicly release an interactive semantic similarity search tool. We demonstrate how our tool can be used to rapidly discover rare objects given only a single example, increase the speed of crowd-sourcing campaigns, and construct and improve training sets for supervised applications. While we focus on images from sky surveys, the technique is straightforward to apply to any scientific dataset of any dimensionality. The similarity search web app can be found at https://github.com/georgestein/galaxy_search
arXiv comment5 pages, 2 figures. The similarity search web app can be found at https://github.com/georgestein/galaxy_search. Accepted to the Fourth Workshop on Machine Learning and the Physical Sciences (NeurIPS 2021). ArXiv admin note: text overlap with arXiv:2110.00023
References in corpus (10)
- Maps of Dust IR Emission for Use in Estimation of Reddening and CMBR Foregrounds
- Auto-Encoding Variational Bayes
- UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
- A Simple Framework for Contrastive Learning of Visual Representations
- Bootstrap your own latent: A new approach to self-supervised Learning
- Improved Baselines with Momentum Contrastive Learning
- The Strong Gravitational Lens Finding Challenge
- Self-supervised Pretraining of Visual Features in the Wild
- Mining for Strong Gravitational Lenses with Self-supervised Learning
- The serendipitous observation of a gravitationally lensed galaxy at z = 0.9057 from the Blanco Cosmology Survey: The Elliot Arc