Skip to content

biotrump/rodrigob-totally_corrective_boosting

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

39 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

This project is based on the original 2009 code from Professors Vishwanathan and Manfred,
http://www.stat.purdue.edu/~vishy/erlpboost

there are significant changes. In particular:
      - Compilation using cmake 
      - Massive renamings and file moves, to improve readability (e.g. wl becomes weak_learner)



Old readme below

------------

Introduction
============

This software package provides a framework for implementing and testing
a variety of boosting algorithms. Currently implemented algorithms
include:

	1) LP Boost [1]
	2) ERLP Boost [2]
	3) Adaboost [3]
	4) Corrective algorithm [4] 

For ERLP Boost, which is the main focus of this package, a number of
optimizers have been implemented:

	   1) Spectral Projected Gradient [5]
	   2) Hager and Zhang algorithm [6]
	   3) Augmented Lagrangian Algorithm [7]
	   4) Coordinate Descent 

Some sample datasets and configuration files are also included with the
distribution. Currently, we can read and handle datasets such as
reuters, real-sim, news20, and astro-ph with relative ease. 

Compiling and Installing Instructions
=====================================

Compiling and installing is a simple matter of ./configure and make. If
you want to install support for LPBoost or Augemented Lagrangian with
the TAO solver then read on. 

In order to compile with TAO support (needed for one variant of the
Augmented Lagrangian Algorithm) you need to first install PETSc (release 
3.0.0-p4) from http://www.mcs.anl.gov/petsc/petsc-as/
and TAO (release 1.10, not yet available to the public) from http://www.mcs.anl.gov/research/projects/tao/

Then

./configure --with-tao

The configure script expects PETSC_DIR, TAO_DIR and PETSC_ARCH
environment variables to be set. 

In order to compile LPBoost support you need to install the COIN-OR LP
solver available from: https://projects.coin-or.org/Clp

Then

	./configure --with-clp

To specify the location of the CLP install use

   ./configure --with-clp --with-clpdir=PREFIX

where PREFIX points to the installation prefix of CLP (the include files
are installed in PREFIX/include/coin/ directory). 

To compile run

   make 

To run the code you 

   cd src 

and 

    ./erlpboost name_of_config_file

For sample configuration files look in the conf directory. 

You need to point your LD_LIBRARY_PATH to the location of the TAO, PETSc
and CLP libraries when you run the erlpboost binary.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published