Generalizing, Optimizing, and Decoding Support Vector Machine Classification
Mario Michael Krell, Sirko Straube, Hendrik Wöhrle, Frank Kirchner
series DFKI Documents, volume 14-07, Nov/2014. DFKI GmbH.
This poster summarizes the content of a paper, presented at the ECML PhD session. It focusses on the main
contributions of the main author for his PhD thesis.
The classification of complex data requires the good choice of a classifier and the composition of processing
steps. We provide three approaches to ease work of data scientists. First, we summarize the connections
between support vector machine (SVM) variants and introduce a generalized model which shows that these
variants are not to be taken separately but that they are highly connected. Second, we present a framework
to optimize the processing chain and the hyperparameters of the used algorithms, including the classifier. At
the final third step, we provide an approach to get back information from the optimized processing chain with
the help of the backtransformation. This approach enables a joint visualization of the complete processing
chain in the input data space and thereby allows for a joint interpretation of preprocessing and classification
to decode the decision process. With the help of the framework, these approaches can be directly used and