Skip to content

The Avatar Tools codes are an implementation of ensembles specifically for decision trees.

License

Notifications You must be signed in to change notification settings

sandialabs/avatar

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Avatar Tools

Version 1.1

Supervised machine learning is the process of using past experience to predict the future. "Ensembles" are a machine-learning meta-method that can be applied to most machine learning algorithms. Ensembles generally greatly improve accuracy, reduce or remove most of the design issues presented by machine learning, and are admirably suited to parallel and distributed computation.

The Avatar Tools codes are an implementation of ensembles specifically for decision trees.

Some features that distinguish Avatar Tools from other "ensembles for decision trees" codes are:

  • Does the bookkeeping necessary for out of bag (OOB) validation.

  • Can use OOB validation to automatically determine optimal ensemble size.

  • Provides an MPI-based parallel implementation, for distributed operation.

  • Provides convenient tools for cross-validation, to assess the accuracy provided by a training set.

  • Handles both plain text and Exodus simulation data.

The Avatar Tools codes are intended for Unix machines, and are known to build and pass their tests on Linux, Mac OS X, and Solaris machines.

See INSTALL for installation instructions,

See the files in support/ for sample data and a brief tutorial.

Citations

Please cite one or more of the following papers, if you wish to cite the Avatar Tools:

@INPROCEEDINGS{Chawla2,
  author = {Nitesh Chawla and Thomas Moore and Kevin Bowyer and Lawrence Hall
        and Clayton Springer and Philip Kegelmeyer},
  title = {Bagging is a Small-Data-Set Phenomenon},
  booktitle = {International Conference on Computer Vision and Pattern Recognition
        (CVPR)},
  year = {2001}
}

@ARTICLE{CaBoHaKe02,
  author = {Nitesh V.~Chawla and Kevin W.~Bowyer and Lawrence O.~Hall and W.~Philip
        Kegelmeyer},
  title = {{SMOTE}: Synthetic Minority Over-sampling Technique},
  journal = {Journal of Artificial Intelligence Research},
  year = {2002},
  volume = {16},
  pages = {321-357},
  url = {http://adsabs.harvard.edu/abs/2011arXiv1106.1813B},
}

ARTICLE{ChHaBoKe04,
  author = {Nitesh V.~Chawla and Lawrence O.~Hall and Kevin W.~Bowyer and W.~Philip
        Kegelmeyer },
  title = {Learning ensembles from bites: A scalable and accurate approach},
  journal = {Journal of Machine Learning Research},
  year = {2004},
  volume = {5},
  pages = {421--451}
}

@INPROCEEDINGS{Chawla4,
  author = {Nitesh V.~Chawla and Lawrence O.~Hall and Kevin W.~Bowyer and Thomas
        E.~Moore and W.~Philip Kegelmeyer},
  title = {Distributed Pasting of Small Votes},
  booktitle = {International Workshop on Multiple Classifier Systems},
  year = {2002},
  address = {Sardegna, Italy},
  month = {June}
}

TECHREPORT{Hall2,
  author = {L.O.~Hall and K.W.~Bowyer and N.~Chawla and T.~Moore and W.~Philip
        Kegelmeyer},
  title = {{AVATAR} --- Adaptive Visualization Aid for Touring and Recovery},
  institution = {Sandia National Laboratories},
  year = {2000},
  type = {Sandia Report},
  number = {SAND2000-8203},
  month = {January}
}

@article {CiHoCh11,
   author = {Cieslak, David and Hoens, T. and Chawla, Nitesh and Kegelmeyer, W.},
   title = {Hellinger distance decision trees are robust and skew-insensitive},
   journal = {Data Mining and Knowledge Discovery},
   publisher = {Springer Netherlands},
   issn = {1384-5810},
   vol = {24}, 
   issue = 1,
   pages = {136--158},
   url = {http://dx.doi.org/10.1007/s10618-011-0222-1},
   note = {10.1007/s10618-011-0222-1},
   year = {2012}
}

Acknowledgments

The Avatar Tools come from a decade-long machine learning research program led by Philip Kegelmeyer at Sandia National Laboratories, with contributions from collaborators at the University of South Florida (notably, Professor Larry Hall and students Robert Banfield and Larry Shoemaker) and at Notre Dame (Computer Science Department Chair Kevin Bower and Professor Nitesh Chawla.)

Early research implementations largely came from Robert Banfield and Steven Eschrich, at USF. The current production implementation is by Ken Buch, a contractor with Limit Point Systems.

If you have questions, contact:

Philip Kegelmeyer, wpk@sandia.gov

About

The Avatar Tools codes are an implementation of ensembles specifically for decision trees.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published