ausblenden:
Schlagwörter:
-
Zusammenfassung:
We have developed an R Interface for our Machine Learning Toolbox SHOGUN. It features algo- rithms to train hidden markov models and learn regression and 2-class classification problems. While the toolbox’s focus is on kernel methods such as Support Vector Machines, it also im- plements a number of linear methods like Linear Discriminant Analysis, Linear Programming Machines and Perceptrons. It provides a generic SVM object interfacing to seven different SVM implementations, among them the state of the art LibSVM[1] and SVMlight[2]. Each of these can be combined with a variety of kernels. The toolbox not only provides efficient implementations of the most common kernels, like the Linear, Polynomial, Gaussian and Sigmoid Kernel but also comes with a number of recent string kernels as e.g. the Spectrum or Weighted Degree Kernel (with shifts). For the latter the efficient linadd[4] optimizations are implemented. Also SHOGUN offers the freedom of working with custom pre-computed kernels. One of its key features is the “combined kernel” which can be constructed by a weighted linear combination of a number of sub-kernels, each of which not necessarily working on the same domain. An optimal sub-kernel weighting can be learned using Multiple Kernel Learning.[3] The input feature-objects can be dense, sparse or strings and of type int/short/double/char and can be converted into different feature types. Chains of “preprocessors” (e.g. substracting the mean) can be attached to each feature object allowing for on-the-fly pre-processing. SHOGUN also supports MatlabTM, Octave and Python-numarray. The Source Code is freely available for academic non commercial use under http://www.fml.mpg.de/raetsch/shogun.
