2020

A Note on Data Biases in Generative Models

Esser, Patrick, Rombach, Robin, Ommer, Björn

Understand

It is tempting to think that machines are less prone to unfairness and prejudice.

  • However, machine learning approaches compute their outputs based on data.
  • While biases can enter at any stage of the development pipeline, models are particularly receptive to mirror biases of the datasets they are trained on and therefore do not necessarily reflect truths about the world but, primarily, truths about the data.
  • To raise awareness about the relationship between modern algorithms and the data that shape them, we use a conditional invertible neural network to disentangle the dataset-specific information from the information which is shared across different datasets.

Built on

  • J. H. Langlois and L. A. Roggman, “Attractive faces are only average,”

    1990

    Earlier work this paper cites.

  • A. Torralba and A. Efros, “Unbiased look at dataset bias,”

    2011

    Earlier work this paper cites.

  • Z. Liu, P. Luo, X. Wang, and X. Tang, “Deep learning face attributes in the wild,” in

    2015

    Earlier work this paper cites.

  • T. Karras, T. Aila, S. Laine, and J. Lehtinen, “Progressive Growing of GANs for Improved Quality, Stability, and Variation,” 2018

    2018

    Earlier work this paper cites.

  • T. Karras, S. Laine, and T. Aila, “A Style-Based Generator Architecture for Generative Adversarial Networks,” 2019

    2019

    Earlier work this paper cites.

  • I. Goodfellow, “"4.5 years of GAN progress", Twitter post, accessed October 08, 2020,” 2019. [Online]. Available:

    2019

    Earlier work this paper cites.

Similar

  • M.-Y. Liu, X. Huang, A. Mallya, T. Karras, T. Aila, J. Lehtinen, and J. Kautz., “Few-shot Unsueprvised Image-to-Image Translation,” in

    2019

    Cited alongside, same era.

  • M. Mitchell, S. Wu, A. Zaldivar, P. Barnes, L. Vasserman, B. Hutchinson, E. Spitzer, I. D. Raji, and T. Gebru, “Model cards for model reporting,”

    2019

    Cited alongside, same era.

  • Y. LeCun, “"ML systems are biased when data is biased.", Twitter post, accessed October 08, 2020,” 2020. [Online]. Available:

    2020

    Cited alongside, same era.

  • A. Gupta, C. Lanteigne, V. Heath, M. B. Ganapini, E. Galinkin, A. Cohen, T. D. Gasperis, M. Akif, and R. Butalid, “The State of AI Ethics Report (June 2020),” 2020

    2020

    Cited alongside, same era.

  • Y. Kilcher, “"[Trash] Automated Inference on Criminality using Face Images", YouTube video, accessed October 08, 2020,” 2020. [Online]. Available:

    2020

    Cited alongside, same era.

  • A. Kurenkov, “Lessons from the pulse model and discussion,”

    2020

    Cited alongside, same era.

Then

  • R. Rombach, P. Esser, and B. Ommer, “Network-to-Network Translation with Conditional Invertible Neural Networks,” in

    2020

    Closest in time.

  • Y. Choi, Y. Uh, J. Yoo, and J.-W. Ha, “StarGAN v2: Diverse Image Synthesis for Multiple Domains,” 2020

    2020

    Closest in time.

  • S. Menon, A. Damian, S. Hu, N. Ravi, and C. Rudin, “Pulse: Self-supervised photo upsampling via latent space exploration of generative models,” 2020

    2020

    Closest in time.

  • Y. Kilcher, “"[Drama] Yann LeCun against Twitter on Dataset Bias", YouTube video, accessed October 08, 2020,” 2020. [Online]. Available:

    2020

    Closest in time.

  • F. Offert, “"There Is No (Real World) Use Case for Face Super Resolution", blog post, accessed October 08, 2020,” 2020. [Online]. Available:

    2020

    Closest in time.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…