Efficient First-Order Algorithms for Adaptive Signal Denoising

Matlab reproducible experiments from the following paper:

Dmitrii Ostrovskii, Zaid Harchaoui. Efficient First-Order Algorithms for Adaptive Signal Denoising. ICML 2018.

We use AdaFilter codes for the efficient implementation of adaptive signal denoising, including them as a submodule.

We also use CVX software for disciplined convex programming.

Installation

1. Make sure that CVX is installed on your computer (commercial license is not needed). CVX installation instructions can be found here.
2. Download or clone the repository, and add the following path in MATLAB:

AlgoRec/AdaFilter/code

Running the experiments

The experiments, in the order of appearance in the paper, are launched via the following MATLAB commands:

exp_perf_MP_random(N,ifReproduce);
exp_perf_MP_coherent(N,ifReproduce);
exp_certificates(N,ifReproduce);
exp_complexity(N,ifReproduce);
exp_sigm(N,ifReproduce);

Plots will appear in folders plots-<...>, where <...> corresponds to the experiment.

Simulation data for the first four experiments will appear in folder sims-perf, and for the last experiment in folder sims-sigm.

• Running a script with ifReproduce = 1 will first launch simulations, and then produce plots for the obtained data. After that, ifReproduce = 0 can be used to produce the plots without running the simulations again.

• N is the number of Monte-Carlo trials. To reproduce the results from the paper, one must set N = 20 for exp_sigm and N = 10 in all other cases. Smaller values of N can be used to obtain faster (and less precise) results.