Prediction-correction for Nonsmooth Time-varying Optimization via Forward-backward Envelopes

  title={Prediction-correction for Nonsmooth Time-varying Optimization via Forward-backward Envelopes},
  author={Nicola Bastianello and Andrea Simonetto and Ruggero Carli},
  journal={ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
We present an algorithm for minimizing the sum of a strongly convex time-varying function with a time-invariant, convex, and nonsmooth function. The proposed algorithm employs the prediction-correction scheme alongside the forward-backward envelope, and we are able to prove the convergence of the solutions to a neighborhood of the optimizer that depends on the sampling time. Numerical simulations for a time-varying regression problem with elastic net regularization highlight the effectiveness… 

Figures from this paper

Distributed Optimization in Time-Varying Environments
A distributed approach to solving optimization problems in a distributed manner is critical in many systems that consist of autonomous agents that are to co-operate with each other.


Prediction-Correction Interior-Point Method for Time-Varying Convex Optimization
An interior-point method for solving a class of convex optimization problems with time-varying objective and constraint functions and ensures that the solution to this dynamical system globally asymptotically converges to the optimal solution at an exponential rate.
A Class of Prediction-Correction Methods for Time-Varying Convex Optimization
This paper proposes algorithms with a discrete time-sampling scheme to find and track the solution trajectory based on prediction and correction steps, while sampling the problem data at a constant rate of 1/h, where h is the sampling period.
Dual Prediction–Correction Methods for Linearly Constrained Time-Varying Convex Programs
This paper shows how to tackle linearly constrained continuously-varying problem by prediction–correction in the dual space and proves similar asymptotical error bounds as their primal versions.
Prediction-Correction Algorithms for Time-Varying Constrained Optimization
The proposed methods are shown to improve the convergence speed of existing prediction-correction methods when applied to unconstrained problems and to design first-order prediction steps that rely on the Hessian of the cost function (and do not require the computation of its inverse).
Envelope Functions: Unifications and Further Properties
This paper provides properties of the general envelope function that sharpen corresponding known results for the special cases and presents a new interpretation of the underlying methods as being majorization–minimization algorithms applied to their respective envelope functions.
Forward–backward quasi-Newton methods for nonsmooth optimization problems
This work proposes an algorithmic scheme that enjoys the same global convergence properties of FBS when the problem is convex, or when the objective function possesses the Kurdyka–Łojasiewicz property at its critical points, and analysis of superlinear convergence is based on an extension of the Dennis and Moré theorem.
Gradient Methods for Nonstationary Unconstrained Optimization Problems
Estimates of intervals (variations) between exact and approximate solutions are found and an asymptotic behavior of these estimates is defined.
A Parametric Nonconvex Decomposition Algorithm for Real-Time and Distributed NMPC
A novel decomposition scheme to solve parametric nonconvex programs as they arise in Nonlinear Model Predictive Control (NMPC) is presented and it is shown that the performance of the optimality-tracking scheme can be enhanced via a continuation technique.
Proximal Newton methods for convex composite optimization
Two proximal Newton methods for convex nonsmooth optimization problems in composite form are proposed, which combines the global efficiency estimates of the corresponding first-order methods, while achieving fast asymptotic convergence rates.
Forward-Backward Envelope for the Sum of Two Nonconvex Functions: Further Properties and Nonmonotone Linesearch Algorithms
It is shown that the forward-backward envelope (FBE), an exact and strictly continuous penalty function for the original cost, still enjoys favorable first- and second-order properties which are key for the convergence results of ZeroFPR.