Corpus ID: 88517649

From Variational to Deterministic Autoencoders

@article{Ghosh2020FromVT,
  title={From Variational to Deterministic Autoencoders},
  author={Partha Ghosh and Mehdi S. M. Sajjadi and Antonio Vergari and Michael J. Black and Bernhard Sch{\"o}lkopf},
  journal={ArXiv},
  year={2020},
  volume={abs/1903.12436}
}
  • Partha Ghosh, Mehdi S. M. Sajjadi, +2 authors Bernhard Schölkopf
  • Published 2020
  • Mathematics, Computer Science
  • ArXiv
  • Variational Autoencoders (VAEs) provide a theoretically-backed framework for deep generative models. However, they often produce "blurry" images, which is linked to their training objective. Sampling in the most popular implementation, the Gaussian VAE, can be interpreted as simply injecting noise to the input of a deterministic decoder. In practice, this simply enforces a smooth latent space structure. We challenge the adoption of the full VAE framework on this specific point in favor of a… CONTINUE READING

    Citations

    Publications citing this paper.
    SHOWING 1-10 OF 35 CITATIONS

    Variance Constrained Autoencoding

    VIEW 3 EXCERPTS
    CITES METHODS

    Regularized Autoencoders via Relaxed Injective Probability Flow

    VIEW 18 EXCERPTS
    CITES METHODS, BACKGROUND & RESULTS
    HIGHLY INFLUENCED

    Deterministic Decoding for Discrete Data in Variational Autoencoders

    VIEW 2 EXCERPTS
    CITES BACKGROUND

    Variance Loss in Variational Autoencoders

    VIEW 5 EXCERPTS
    CITES METHODS & BACKGROUND
    HIGHLY INFLUENCED

    Variational Autoencoders Pursue PCA Directions (by Accident)

    VIEW 2 EXCERPTS
    CITES BACKGROUND

    Balancing reconstruction error and Kullback-Leibler divergence in Variational Autoencoders

    VIEW 7 EXCERPTS
    CITES BACKGROUND & METHODS
    HIGHLY INFLUENCED

    FILTER CITATIONS BY YEAR

    2018
    2020

    CITATION STATISTICS

    • 5 Highly Influenced Citations

    • Averaged 12 Citations per year from 2018 through 2020

    • 62% Increase in citations per year in 2020 over 2019

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 65 REFERENCES

    VAE with a VampPrior

    VIEW 2 EXCERPTS

    Importance Weighted Autoencoders

    VIEW 1 EXCERPT

    Variational Lossy Autoencoder

    VIEW 1 EXCERPT

    Diagnosing and Enhancing VAE Models

    VIEW 6 EXCERPTS
    HIGHLY INFLUENTIAL

    Taming VAEs

    VIEW 1 EXCERPT

    Resampled Priors for Variational Autoencoders

    VIEW 6 EXCERPTS
    HIGHLY INFLUENTIAL