hintsvm implements the active learning algorithm described in "Active Learning with Hinted Support Vector Machine." by Li, Chun-Liang, Chun-Sung Ferng, and Hsuan-Tien Lin.
hintsvm is based on libsvm, modified from libsvm3.1.
There are two parameters added to the original libsvm:
- '-s 5' for hintsvm.
- '-W weight file'.
Usage:
- Use '-W weight_file' to assign weights for each instance which is used to adjust the weight between labeled and unlabeled instances.
- Please make sure all weights are non-negative.
Example:
./svm-train -W heart_scale.wgt heart_scale
For now only the python wrapper are supported.