Corpus ID: 2850605

Quasi-Newton Methods: A New Direction

@article{Hennig2012QuasiNewtonMA,
  title={Quasi-Newton Methods: A New Direction},
  author={Philipp Hennig and Martin Kiefel},
  journal={ArXiv},
  year={2012},
  volume={abs/1206.4602}
}
  • Philipp Hennig, Martin Kiefel
  • Published in ICML 2012
  • Computer Science, Mathematics
  • Four decades after their invention, quasi-Newton methods are still state of the art in unconstrained numerical optimization. Although not usually interpreted thus, these are learning algorithms that fit a local quadratic approximation to the objective function. We show that many, including the most popular, quasi-Newton methods can be interpreted as approximations of Bayesian linear regression under varying prior assumptions. This new notion elucidates some shortcomings of classical algorithms… CONTINUE READING

    Create an AI-powered research feed to stay up to date with new papers like this posted to ArXiv

    Citations

    Publications citing this paper.
    SHOWING 1-10 OF 54 CITATIONS

    Probabilistic Interpretation of Linear Solvers

    • Philipp Hennig
    • Mathematics, Computer Science
    • SIAM Journal on Optimization
    • 2015
    VIEW 6 EXCERPTS
    CITES METHODS

    Steady-state optimization for biodiesel production in a reactive distillation column

    VIEW 8 EXCERPTS
    CITES METHODS
    HIGHLY INFLUENCED

    Fast Probabilistic Optimization from Noisy Gradients

    VIEW 8 EXCERPTS
    CITES METHODS & BACKGROUND

    Conjugate Gradients for Kernel Machines

    VIEW 3 EXCERPTS
    CITES BACKGROUND & RESULTS

    On the construction of probabilistic Newton-type algorithms

    VIEW 4 EXCERPTS
    CITES BACKGROUND & METHODS
    HIGHLY INFLUENCED

    A Probabilistic Incremental Proximal Gradient Method

    VIEW 1 EXCERPT
    CITES METHODS

    A modern retrospective on probabilistic numerics

    VIEW 1 EXCERPT
    CITES METHODS

    Deep kernel learning for integral measurements

    VIEW 1 EXCERPT
    CITES METHODS

    FILTER CITATIONS BY YEAR

    2012
    2019

    CITATION STATISTICS

    • 2 Highly Influenced Citations

    • Averaged 9 Citations per year from 2017 through 2019

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 16 REFERENCES

    Numerical Optimization

    VIEW 2 EXCERPTS
    HIGHLY INFLUENTIAL

    The Handbook of Matrices

    VIEW 1 EXCERPT
    HIGHLY INFLUENTIAL

    Variable Metric Method for Minimization

    VIEW 4 EXCERPTS
    HIGHLY INFLUENTIAL

    Convex Optimization

    VIEW 1 EXCERPT

    A family of variable metric updates

    VIEW 3 EXCERPTS