Image Details

Choose export citation format:

Dimensionality Reduction of SDSS Spectra with Variational Autoencoders

  • Authors: Stephen K. N. Portillo, John K. Parejko, Jorge R. Vergara, and Andrew J. Connolly

2020 The Astronomical Journal 160 45.

  • Provider: AAS Journals

Caption: Figure 3.

Mean-squared reconstruction error for principal component analysis, non-negative matrix factorization, non-variational autoencoder, and variational autoencoder (VAE). The VAE outperforms the other methods, with the greatest advantage being at small numbers of parameters. The VAE has the greatest relative advantage for broad-line active galaxies, which also have the highest absolute reconstruction error.

Other Images in This Article

Show More

Copyright and Terms & Conditions

Additional terms of reuse