As discussed in #23, textmodel_svm() fails with a large number of training documents. However, it works when using the more powerful textmodel_svmlin() instead. Yet, currently, textmodel_svmlin() does not rescale the support vectors into probabilities for multiple classes. Adding this feature to textmodel_svmlin() would solve problems similar to #23.
As discussed in #23,
textmodel_svm()fails with a large number of training documents. However, it works when using the more powerfultextmodel_svmlin()instead. Yet, currently,textmodel_svmlin()does not rescale the support vectors into probabilities for multiple classes. Adding this feature totextmodel_svmlin()would solve problems similar to #23.