Corpus ID: 231662111

Can stable and accurate neural networks be computed? - On the barriers of deep learning and Smale's 18th problem

@article{Antun2021CanSA,
  title={Can stable and accurate neural networks be computed? - On the barriers of deep learning and Smale's 18th problem},
  author={Vegard Antun and Matthew J. Colbrook and A. Hansen},
  journal={ArXiv},
  year={2021},
  volume={abs/2101.08286}
}
  • Vegard Antun, Matthew J. Colbrook, A. Hansen
  • Published 2021
  • Computer Science, Mathematics
  • ArXiv
  • Deep learning (DL) has had unprecedented success and is now entering scientific computing with full force. However, DL suffers from a universal phenomenon: instability, despite universal approximating properties that often guarantee the existence of stable neural networks (NNs). We show the following paradox. There are basic well-conditioned problems in scientific computing where one can prove the existence of NNs with great approximation qualities, however, there does not exist any algorithm… CONTINUE READING
    1 Citations

    Figures and Tables from this paper

    Computing Spectral Measures and Spectral Types
    • 2
    • PDF

    References

    SHOWING 1-10 OF 161 REFERENCES
    Optimal Approximation with Sparsely Connected Deep Neural Networks
    • 133
    • PDF
    Neural Network Approximation
    • 3
    • PDF
    Nonlinear Approximation and (Deep) ReLU Networks
    • 49
    • PDF
    Intriguing properties of neural networks
    • 6,083
    • PDF
    Deep Learning: An Introduction for Applied Mathematicians
    • 70
    • PDF
    The troublesome kernel: why deep learning for inverse problems is typically unstable
    • 17
    • PDF
    Neural Networks and Rational Functions
    • 38
    • PDF
    Ada-LISTA: Learned Solvers Adaptive to Varying Models
    • 4
    • PDF
    Theoretical Linear Convergence of Unfolded ISTA and its Practical Weights and Thresholds
    • 77
    • PDF