Skip to main content
Help us improve PyPI by participating in user testing. All experience levels needed!

Python package for solving large scale L1 regularizedleast squares problems.

Project description

BuildStatus

This is a large scale L1 regularized Least Square (L1-LS) solver written in Python. The code is based on the MATLAB code made available on Stephen Boyd’s l1_ls page.

L1LSProblem

Installation

You can install the bleeding edge directly from the source:

pip install git+https://github.com/musically-ut/l1-ls.py.git@master#egg=l1ls

This package is also available on PyPi.

pip install l1ls

Usage

The module exposes two functions:

  • l1ls(A, y, lmbda, x0=None, At=None, m=None, n=None, tar_gap=1e-3, quiet=False, eta=1e-3, pcgmaxi=5000), and,
  • l1ls_nonneg(A, y, lmbda, x0=None, At=None, m=None, n=None, tar_gap=1e-3, quiet=False, eta=1e-3, pcgmaxi=5000)

They can be used as follows:

import l1ls as L
import numpy as np

A = np.array([[1, 0, 0, 0.5], [0, 1, 0.2, 0.3], [0, 0.1, 1, 0.2]])
x0 = np.array([1, 0, 1, 0], dtype='f8')  # Original signal
y = A.dot(x0)                            # noise free signal
lmbda = 0.01                             # regularization parameter
rel_tol = 0.01

[x, status, hist] = L.l1ls(A, y, lmbda, tar_gap=rel_tol)
# answer_x = np.array([0.993010, 0.00039478, 0.994096, 0.00403702])

If your matrix A is sparse, pass it in CSR format format for best performance.

Reference

  • S.-J. Kim, K. Koh, M. Lustig, S. Boyd, and D. Gorinevsky. An Interior-Point Method for Large-Scale l1-Regularized Least Squares, (2007), IEEE Journal on Selected Topics in Signal Processing, 1(4):606-617.

Project details


Release history Release notifications

This version
History Node

0.2.1

History Node

0.2.0

History Node

0.1.0

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging CloudAMQP CloudAMQP RabbitMQ AWS AWS Cloud computing Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page