skscope-reproducibility

Here are some reproducibility code of sparse convex optimization algorithms in library skscope. The name of files in this repository is the corresponding algorithm, and *.py files contains the code of reproducibility experiments, *.csv files contains results of reproducibility experiments, *.ipynb files plot figures of results.

GraSP

Here we reproduce the result of Greedy sparsity-constrained optimization.

This is the figure of reproducing:

This is the figure of the paper:

HTP and IHT

Here we reproduce the result of Hard Thresholding Pursuit Algorithms: Number of Iterations.

This is the figure of reproducing:

This is the figure of the paper:

Foba and OMP

Here we reproduce the result of Forward-Backward Greedy Algorithms for General Convex Smooth Functions over A Cardinality Constraint. In the experiment of this paper, algorithms use information not obtainable in practice to select sparsity level. And we get different results here because we use cross validation to select sparsity level.

This is the figure of reproducing:

This is the figure of the paper:

Note that the 'forward-gdt' refers to the OMP algorithm.