Novelty detection, anomaly detection and one-class classification, all mean the same thing: Classify samples as positive or negative having only positive training examples.
The SVM-based framework used by novel enables the use of non-linear kernels that help capturing the shape of the training examples.
novel implements the one-class SVM classifier, based on the paper
Support Vector Data Description
of Tax et al.
moon-shaped dataset
Rational Quadratic kernel
- numpy
- cvxopt
~$ pip install novel