High performance ℓ₁-minimization solvers for sparse sensing and signal recovery problems.

Python – The python binding is available as a package on pypi via pip install:

pip install sparsesolvers

Here is a toy example:

import sparsesolvers as ss
import numpy as np

N = 10

# Create an example sensing matrix
A = np.random.normal(loc=0.025, scale=0.025, size=(N, N)) + np.identity(N)

# An incoming signal
signal = np.zeros(N)
signal[2] = 1

# Use the homotopy solver to produce sparse solution, x.
x, info = ss.Homotopy(A).solve(signal, tolerance=0.1)

# Example output: error=0.064195, sparsity=0.9, argmax=2
print("error=%f, sparsity=%f, argmax=%i" % (
    info.solution_error, 1 - np.count_nonzero(x) / np.double(N),
    np.argmax(x)))

1. A. Y. Yang, Z. Zhou, A. Ganesh, S. S. Sastry, and Y. Ma – Fast ℓ₁-minimization Algorithms For Robust Face Recognition – IEEE Trans. Image Processing, vol. 22, pp. 3234–3246, Aug 2013.

2. R. Chartrand, W. Yin – Iteratively Reweighted Algorithms For Compressive Sensing – Acoustics Speech and Signal Processing 2008. ICASSP 2008. IEEE International Conference, pp. 3869-3872, March 2008.

3. D. O’Leary – Robust Regression Computation Using Iteratively Reweighted Least Squares – Society for Industrial and Applied Mathematics, 1990

Sparse solvers is also a c++14 library for your own projects. The python binding is a good example of how you can incorporate the solvers in to your own c++ projects with minimal effort.

At a minimum, you will need:

• CMake 3.2

• A reasonably compliant C++14 compiler, e.g.:

  Windows	Linux	Mac
  VS 2015	gcc 5.3 / clang 3.6	XCode 7.3

First, clone the repository and its submodules:

git clone --recurse-submodules https://github.com/rayglover-ibm/sparse-solvers
cd sparse-solvers

Configure and build using CMake in the typical way:

mkdir build && cd build
cmake ..
cmake --build . [--config Release]

Run the test suite and/or (if you've enabled them) benchmarks:

ctest -VV . [--config <config>]

There are a number of sparse solvers specific CMake options:

To build the python package (.whl) you will need the relevant Python development package, such as python-dev for Debian/Ubuntu. For Windows/Mac I recommend Conda. To build the wheel:

mkdir build-py && cd build-py
cmake -Dsparsesolvers_WITH_PYTHON=ON ..
cmake --build . --target bdist_wheel [--config Release]

Once the wheel has been created (usually in build-py/bindings/python/dist) you can install it with pip locally in the usual way:

pip install <path/to/sparsesolvers.whl>

Copyright 2017 International Business Machines Corporation

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.