Fast, Efficient One Class Support Measure Machines

Repository intended to explore a fast and memory efficient implementation of One Class Support Measure Machines (OCSMM). Ideas from these papers are explored:

1. One-Class Support Measure Machines for Group Anomaly Detection, 2013, Proceedings of the Twenty-Ninth
2. Large Scale Online Kernel Learning. Journal of Machine Learning Research 2016

Conda environment

local_user@local_host$ bash setup.sh

Run OCSMM experiments

local_user@local_host$ python src/utils_experiment.py
local_user@local_host$ jupyter notebook src/group_outlier_detection.ipynb