Corpus ID: 91175758

Towards GAN Benchmarks Which Require Generalization

  title={Towards GAN Benchmarks Which Require Generalization},
  author={Ishaan Gulrajani and Colin Raffel and Luke Metz},
  • Ishaan Gulrajani, Colin Raffel, Luke Metz
  • Published 2019
  • Mathematics, Computer Science
  • ArXiv
  • For many evaluation metrics commonly used as benchmarks for unconditional image generation, trivially memorizing the training set attains a better score than models which are considered state-of-the-art; we consider this problematic. We clarify a necessary condition for an evaluation metric not to behave this way: estimating the function must require a large sample from the model. In search of such a metric, we turn to neural network divergences (NNDs), which are defined in terms of a neural… CONTINUE READING
    Detecting Overfitting of Deep Generative Networks via Latent Recovery
    • 12
    • Highly Influenced
    • PDF
    Fair Generative Modeling via Weak Supervision
    • 5
    • PDF
    Stable Rank Normalization for Improved Generalization in Neural Networks and GANs
    • 6
    • PDF
    Top-K Training of GANs: Improving Generators by Making Critics Less Critical
    • 1
    A Non-Parametric Test to Detect Data-Copying in Generative Models
    • 1
    • PDF
    Generating Private Data Surrogates for Vision Related Tasks
    • 2019
    A Hamiltonian Monte Carlo Method for Probabilistic Adversarial Attack and Learning


    Publications referenced by this paper.
    Improved Techniques for Training GANs
    • 3,587
    • Highly Influential
    • PDF
    Generative Adversarial Nets
    • 17,799
    • PDF
    Rethinking the Inception Architecture for Computer Vision
    • 8,606
    • PDF
    Improved Training of Wasserstein GANs
    • 3,333
    • PDF
    Progressive Growing of GANs for Improved Quality, Stability, and Variation
    • 2,073
    • PDF
    Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification
    • 8,022
    • PDF
    A note on the evaluation of generative models
    • 608
    • Highly Influential
    • PDF
    GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
    • 1,750
    • Highly Influential
    • PDF
    Learning from Simulated and Unsupervised Images through Adversarial Training
    • 1,025
    • PDF
    Realistic Evaluation of Deep Semi-Supervised Learning Algorithms
    • 267
    • PDF