PMC:1892782 / 15985-16524 JSONTXT

{"target":"https://pubannotation.org/docs/sourcedb/PMC/sourceid/1892782","sourcedb":"PMC","sourceid":"1892782","source_url":"https://www.ncbi.nlm.nih.gov/pmc/1892782","text":"We use libsvm 2.8 [18] with SVM type C_SVC, a radial basis function (RBF) kernel, probability estimates and descriptor vectors scaled linearly to the interval [-1, +1]. The scaling avoids that descriptors which have a large variance dominate the classification. The values for the RBF kernel parameters γ and C were identified by a grid search in the parameter space applying grid.py of the libsvm 2.8 package with a 5-fold cross-validation on the training data. Maximal prediction accuracy is achieved with parameters C = 128 and γ = 0.5.","tracks":[]}