Icsiboost - Open-source implementation of Boostexter (Adaboost based classifier)

  •        1078

Boosting is a meta-learning approach that aims at combining an ensemble of weak classifiers to form a strong classifier. Adaptive Boosting (Adaboost) is a greedy search for a linear combination of classifiers by overweighting the examples that are misclassified by each classifier. icsiboost implements Adaboost over stumps (one-level decision trees) on discrete and continuous attributes (words and real values). See http://en.wikipedia.org/wiki/AdaBoost and the papers by Y. Freund and R. Schapire for more details. This approach is one of the most efficient and simple to combine continuous and nominal values. Our implementation is aimed at allowing training from millions of examples by hundreds of features (or millions of sparse features) in a reasonable time/memory. It includes classification time code for c, python and java. Here is an excellent tutorial on Boosting: http://nips.cc/Conferences/2007/Program/event.php?ID=575 WARNING: we are planning to switch to git for revision control NEWS: 2012-05-21: Switched to dual license: GPL and BSD. Choose the license that best fits your project. Old revisions are GPL only. 2011-02-15: icsiboost is now in archlinux (aur package icsiboost-svn) 2010-12-23: QualityTesting tracks svn releases for time, memory and error rate on a sample dataset. 2010-12-10: Added script to convert example files from icsiboost to svm_light/mlcomp format including ngram/cutoff management (icsiboost_to_svm.py) 2010-10-20: Fixed a bug with continuous features (r159) and removed the need for the --display-maxclass option: it is now the default when examples have a single label (r160). The old way of computing the error rate is still used in the multilabel scenario. 2010-10-03: Added support for Solaris (r154), also added win32 downloads (requires cygwin1.dll) 2010-05-05: Maintenance release: better error handling of discrete features declared in names file (r130). 2010-01-24: Added Stanislas' patch to display error rates based on argmax instead of sign decisions. 2009-10-19: Added a rudimentary java implementation of the classifier. 2009-10-10: Released the optimal_threshold.pl script to get a better decision threshold on unbalanced data (for binary problems only). 2009-07-29: There is now a decoder in pure python. It's quite slow (and could be optimized), but is useful for small projects in python and educational purposes. 2009-04-08 WARNING: On multiclass problems, icsiboost does not compute the error rates the same way boostexter does. This does not result in lower performing models, and an option for getting compatible values will be implemented in the future. 2009-03-30 You can now specify the type of text expert and its length on a per-column basis in the names file (previously set globally with -N ngram -W 3...). Example: "words:text." becomes "words:text:expert_type=ngram expert_length=5 cutoff=3." which is equivalent to -N ngram -W 5 --cutoff 3, only for that the words column. You should use the svn version to get the latest fixes (change log). do not use r96: a bug made training fail (all users should upgrade to r102 which fixes major bugs) WARNING: if you trained a model with -N ngram -W length, you must pass the same options at test time, otherwise the related weak classifiers will be ignored (unless you specify it in the names file). Get and Compile (you need PCRE >= 01-December-2003): svn checkout http://icsiboost.googlecode.com/svn/trunk/ .cd icsiboostautoreconfautomake -a./configure CFLAGS=-O3makeProgram usage (revision r124): USAGE: ./icsiboost [options] -S --version print version info -S defines model/data/names stem -n number of boosting iterations (also limits test time classifiers, if model is not packed) -E set smoothing value (default=0.5) -V verbose mode -C classification mode -- reads examples from -o long output in classification mode -N choose a text expert between fgram, ngram and sgram (also ":text:expert_type=" in the .names) -W specify window length of text expert (also ":text:expert_length=" in .names) --dryrun only parse the names file and the data file to check for errors --cutoff ignore nominal features occuring unfrequently (also ":text:cutoff=" in .names) --drop drop text features that match a regular expression (also ":text:drop=" in .names) --no-unk-ngrams ignore ngrams that contain the "unk" token --jobs number of threaded weak learners --do-not-pack-model do not pack model (this is the default behavior) --pack-model pack model (for boostexter compatibility) --output-weights output training examples weights at each iteration --posteriors output posterior probabilities instead of boosting scores --model save/load the model to/from this file instead of .shyp --resume resume training from a previous model (can use another dataset for adaptation) --train bypass the .data filename to specify

http://code.google.com/p/icsiboost

Tags
Implementation
License
Platform

   




Related Projects

vim-olive - Vim Mode Line Verifier


Vim Mode Line Verifier

vim-mash - Motion Activated Search Highlighter for Vim


Motion Activated Search Highlighter for Vim

vim-grillz - Flash your wicked grillz!


Flash your wicked grillz!

vim-foist - Complete whole lines from any partial therein


Complete whole lines from any partial therein

vim-efmc - Vim Error Format Compiler


Vim Error Format Compiler


uzbl-utrs - mkng uzbl tlrbl


mkng uzbl tlrbl

toycsv - A toy CSV parser written in ruby + lexr + racc


A toy CSV parser written in ruby + lexr + racc

tiktok - TikTok provides a simple asynchronous timer object for VimL.


TikTok provides a simple asynchronous timer object for VimL.

tabby - Using Vim's Tabs the Right Way


Using Vim's Tabs the Right Way

SohiVila - GNU source-highlight Vim language


GNU source-highlight Vim language

SinTax - A DSL for generating Vim syntax highlighting files


A DSL for generating Vim syntax highlighting files

rkdots - Generate a graphviz visualisation of the given javascript statement.


Generate a graphviz visualisation of the given javascript statement.

RelNumBar - Show relativenumbers alongside normal numbers in Vim


Show relativenumbers alongside normal numbers in Vim

Punisher - Punisher hurts you where it hurts most - your time.


Punisher hurts you where it hurts most - your time.

noisy - Don't chat quietly (weechat channel noises)


Don't chat quietly (weechat channel noises)

newlisp-manual - Asciidoc version of the newLISP manual


Asciidoc version of the newLISP manual

Land-of-newLISP - Selected snippets from Land of Lisp rewritten in newLISP


Selected snippets from Land of Lisp rewritten in newLISP

Fossilise - Poor man's collaborative editing in Vim


Poor man's collaborative editing in Vim

firstly - Convert Between Numeric, Spelt, and Short & Long Ordinal Forms of Numbers


Convert Between Numeric, Spelt, and Short & Long Ordinal Forms of Numbers