Implementation of incremental Growing Neural Gas Support Vector Machines (iGNGSVM) for MOA. See it in '/moa/classifiers'.
See accepted manuscript here.
Implementation of Online Incremental Learning Support Vector Machine (OI-SVM) for MOA. See it in '/moa/classifiers'.
OISVM and IGNGSVM have been compared previously in http://hdl.handle.net/10016/19258.
Both implementations use LibSVM for WEKA, which jar file is present in the root of this repo.
Support Vectors are retrieved from the SVM using the jar's object.
Noise reduction algorithm based in Nearest Neighbors. Proposed by Dennis L. Wilson in 1972. See it in '/weka/filters'.
Growing Neural Gas is implemented in MOA as topology generation method for IGNGSVM. Proposed by Fritzke in 1995.
EvaluateChunksTwoFiles, in '/moa/tasks', evaluates test and train in parallel to allow online evaluation for non-stationary environments in presence of concept drift. Albeit this is not the only Task that can be run in MOA for this purpose. For example, run_example.sh uses EvaluateInterleavedTestThenTrain as task.