Certifying the True Error: Machine Learning in Coq with Verified Generalization Guarantees

@inproceedings{Bagnall2019CertifyingTT,
  title={Certifying the True Error: Machine Learning in Coq with Verified Generalization Guarantees},
  author={Alexander Bagnall and G. Stewart},
  booktitle={AAAI},
  year={2019}
}
  • Alexander Bagnall, G. Stewart
  • Published in AAAI 2019
  • Computer Science
  • We present MLCERT, a novel system for doing practical mechanized proof of the generalization of learning procedures, bounding expected error in terms of training or test error. MLCERT is mechanized in that we prove generalization bounds inside the theorem prover Coq; thus the bounds are machine checked by Coq’s proof checker. MLCERT is practical in that we extract learning procedures defined in Coq to executable code; thus procedures with proved generalization bounds can be trained and deployed… CONTINUE READING
    9 Citations
    Verification of ML Systems via Reparameterization
    • 1
    • PDF
    Neural Network Verification for the Masses (of AI graduates)
    • Highly Influenced
    • PDF
    Relative Robustness of Quantized Neural Networks Against Adversarial Attacks
    • PDF
    A Safety Framework for Critical Systems Utilising Deep Neural Networks
    • 4
    • PDF
    A Formal Proof of PAC Learnability for Decision Stumps
    • 2
    • PDF

    References

    SHOWING 1-10 OF 27 REFERENCES
    Developing Bug-Free Machine Learning Systems With Formal Mathematics
    • 38
    • Highly Influential
    • PDF
    Computing Nonvacuous Generalization Bounds for Deep (Stochastic) Neural Networks with Many More Parameters than Training Data
    • 286
    • PDF
    Provable defenses against adversarial examples via the convex outer adversarial polytope
    • 657
    • PDF
    Proving expected sensitivity of probabilistic programs
    • 19
    • PDF
    Reluplex: An Efficient SMT Solver for Verifying Deep Neural Networks
    • 724
    • Highly Influential
    • PDF
    TensorFlow: A system for large-scale machine learning
    • 8,540
    • PDF
    Robustness and generalization
    • 226
    • PDF
    Auto-Encoding Variational Bayes
    • 10,438
    • PDF
    Automatic differentiation in PyTorch
    • 7,348