# Statistical Machine Learning and Dissolved Gas Analysis


Companion code for the paper:
"Statistical Machine Learning and Dissolved Gas Analysis: A Review"
P Mirowski, Y LeCun
Power Delivery, IEEE Transactions on, 2012
http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6301810
http://engineering.nyu.edu/power/sites/engineering.nyu.edu.power/files/uploads/DGOA%20predictive.pdf

An appendix to the paper submission 
"Statistical Machine Learning and Dissolved Gas Analysis: A Review" 
that describes the machine learning algorithms with further details, 
is available at: https://piotrmirowski.files.wordpress.com/2020/01/mlreview4dgoa_appendix.pdf


## Requirements:
The following libraries need to be installed 
(and the Matlab paths configured accordingly):
LibSVM with Matlab interface, 
available at: http://www.csie.ntu.edu.tw/~cjlin/libsvm/
Low-Density Separation, available at: http://olivier.chapelle.cc/lds/
The Matlab Statistics Toolbox

A copy of the `bolasso` library is available in this repository. 
It was taken from repository:
[github.com/probml/pmtk1](https://github.com/probml/pmtk1/tree/master/pmtk/optim/Lasso) 
by Matthew Dunham and it is an implementation of Francis R. Bach's Bolasso algorithm
described in his [ICML 2008 paper](http://www.di.ens.fr/~fbach/icml_bolasso.pdf).


## Installation:
After download, unzip and configure the required paths.


## Tutorial:
In directory Code_Duval, execute under Matlab the following file:
Duval.m


## License:
Please refer to the GNU General Public License, 
available at: http://www.gnu.org/


## References for the data:
M. Duval and A. dePablo, "Interpretation of gas-in-oil analysis using new IEC 
publication 60599 and IEC TC 10 databases", 
IEEE Electrical Insulation Magazine, vol. 17, pp. 3141, 2001.