Corpus ID: 174797735

What do AI algorithms actually learn? - On false structures in deep learning

@article{Thesing2019WhatDA,
  title={What do AI algorithms actually learn? - On false structures in deep learning},
  author={L. Thesing and Vegard Antun and A. Hansen},
  journal={ArXiv},
  year={2019},
  volume={abs/1906.01478}
}
  • L. Thesing, Vegard Antun, A. Hansen
  • Published 2019
  • Mathematics, Computer Science
  • ArXiv
  • There are two big unsolved mathematical questions in artificial intelligence (AI): (1) Why is deep learning so successful in classification problems and (2) why are neural nets based on deep learning at the same time universally unstable, where the instabilities make the networks vulnerable to adversarial attacks. We present a solution to these questions that can be summed up in two words; false structures. Indeed, deep learning does not learn the original structures that humans use when… CONTINUE READING
    4 Citations

    References

    SHOWING 1-10 OF 45 REFERENCES
    Intriguing properties of neural networks
    • 6,092
    • PDF
    Towards Deep Learning Models Resistant to Adversarial Attacks
    • 2,940
    • PDF
    Deep neural networks are easily fooled: High confidence predictions for unrecognizable images
    • 1,840
    • PDF
    One Pixel Attack for Fooling Deep Neural Networks
    • 852
    • PDF
    A Direct Approach to Robust Deep Learning Using Adversarial Networks
    • 34
    • PDF
    On instabilities of deep learning in image reconstruction and the potential costs of AI
    • 74
    • PDF
    Explaining and Harnessing Adversarial Examples
    • 6,533
    • PDF
    DeepFool: A Simple and Accurate Method to Fool Deep Neural Networks
    • 2,059
    • PDF
    Distillation as a Defense to Adversarial Perturbations Against Deep Neural Networks
    • 1,574
    • PDF
    Universal Adversarial Perturbations
    • 1,105
    • PDF