Corpus ID: 1611815

Better Approximation and Faster Algorithm Using the Proximal Average

@inproceedings{Yu2013BetterAA,
  title={Better Approximation and Faster Algorithm Using the Proximal Average},
  author={Y. Yu},
  booktitle={NIPS},
  year={2013}
}
  • Y. Yu
  • Published in NIPS 2013
  • Computer Science, Mathematics
  • It is a common practice to approximate "complicated" functions with more friendly ones. In large-scale machine learning applications, nonsmooth losses/regularizers that entail great computational challenges are usually approximated by smooth functions. We re-examine this powerful methodology and point out a nonsmooth approximation which simply pretends the linearity of the proximal map. The new approximation is justified using a recent convex analysis tool— proximal average, and yields a novel… CONTINUE READING
    65 Citations
    Accelerated Stochastic Gradient Method for Composite Regularization
    • 27
    • Highly Influenced
    • PDF
    Gradient Descent with Proximal Average for Nonconvex and Composite Regularization
    • 22
    • Highly Influenced
    • PDF
    Efficient Learning with a Family of Nonconvex Regularizers by Redistributing Nonconvexity
    • 35
    • PDF
    VAGA: a novel viscosity-based accelerated gradient algorithm
    • 5
    Minimizing Nonconvex Non-Separable Functions
    • 21
    • PDF
    Selective Linearization For Multi-Block Convex Optimization
    • 1
    • PDF

    References

    SHOWING 1-10 OF 23 REFERENCES
    Smoothing proximal gradient method for general structured sparse regression
    • 196
    • Highly Influential
    • PDF
    Stochastic Alternating Direction Method of Multipliers
    • 213
    • PDF
    Proximal Splitting Methods in Signal Processing
    • P. L. Combettes, J. Pesquet
    • Computer Science, Mathematics
    • Fixed-Point Algorithms for Inverse Problems in Science and Engineering
    • 2011
    • 2,052
    • PDF
    Accelerated and Inexact Forward-Backward Algorithms
    • 126
    • PDF
    Smooth minimization of non-smooth functions
    • Y. Nesterov
    • Mathematics, Computer Science
    • Math. Program.
    • 2005
    • 2,185
    • Highly Influential
    • PDF
    Structured sparsity through convex optimization
    • 275
    • PDF
    A Fast Iterative Shrinkage-Thresholding Algorithm for Linear Inverse Problems
    • 8,060
    • Highly Influential
    • PDF
    The Proximal Average: Basic Theory
    • 87
    • Highly Influential
    • PDF
    Sparsity and smoothness via the fused lasso
    • 2,055
    • PDF
    Gradient methods for minimizing composite functions
    • Y. Nesterov
    • Mathematics, Computer Science
    • Math. Program.
    • 2013
    • 876
    • Highly Influential
    • PDF