pycasso 0.1.1

Picasso Python Package

PICASSO: Penalized Generalized Linear Model Solver - Unleash the Power of Non-convex Penalty

Unleash the power of nonconvex penalty

L1 penalized regression (LASSO) is great for feature selection. However when you use LASSO in very noisy setting, especially when some columns in your data have strong colinearity, LASSO tends to give biased estimator due to the penalty term. As demonstrated in the example below, the lowest estimation error among all the lambdas computed is as high as 16.41%.

Requirements

• Linux or MacOS

It may take lots of effort to build on Windows. One way to do it is using CMAKE and MSVC. Be careful of issues like the system bits.

Installation

In the following process, you may need to be root (sudo).

Install from source file (Github):

• Clone picasso.git via git clone https://github.com/jasonge27/picasso.git

• Make sure you have setuptools

  Using Makefile

• Run sudo make Pyinstall command.

  Using CMAKE

• Build the source file first via the cmake with CMakeLists.txt in the root directory. (You will see a .so or .lib file under (root)/lib/ )

• Run cd python-package; sudo python setup.py install command.

Install from PyPI:

• pip install pycasso

• Note: Owing to the setting on different OS, our distribution might not be working in your environment (especially in Windows). Thus please build from source.

You can test if the package has been successfully installed by:

import pycasso
pycasso.test()

Usage

import pycasso
x = [[1,2,3,4,5,0],[3,4,1,7,0,1],[5,6,2,1,4,0]]
y = [3.1,6.9,11.3]
s = pycasso.Solver(x,y)
s.train()
s.predict()

For Developer

Please follow the sphinx syntax style

To update the document: cd doc; make html

Copy Right

Author:

Jason Ge, Haoming Jiang

Maintainer:

Haoming Jiang <[e-mail]>