V-detector

V-detector is an anomaly detection (one-class classification) algorithm and its implementation. It is an important part of my PhD dissertation.

This is where you can find useful information, including source code, about V-detector: http://vdetector.zhouji.net

Background

It is a general purpose one-class classification algorithm. As a development in the research area of Artificial Immune Systems (AIS), it is a negative selection algorithm, one of the earliest methods inspired by immune system. Negative selection algorithms are mainly based on the negative selection process in the thymus, in which immature T cells that recognize self cells are eliminated before the rest can mature and are deployed into the immune system.

The implementation posted here uses real-valued representation although the algorithm of V-detector should apply to other data representations as long as the matching rule can be expressed as a threshold of some distance measure.

Source code

V-detector has been re-implemented as an open source project hosted at SourceForge.net. I strongly suggest you use the code at SourceForge.net although the code originally posted here are still available upon request at zhou.ji [at] yahoo.com.

V-detector project is now at http://vdetector.sourceforge.net/

The first relrease can be downloaded at release package

I also cordially invite all interesting parties to join this project even if the only thing you want to do with V-detector is to ask questions. If you are requesting features, you're already at the second level! If you'd like to contribute more in any way, you're very welcome and greatly appreciated.

Implementation in C

Implementation in C is still available here, but I intend to migrate it into the sourceforge.net project as well.

Test Dataset

2-D Synthetic Dataset
Here is more details.

References

  1. Zhou Ji, Dipankar Dasgupta, Real-valued Negative Selection Algorithm with Variable-sized Detectors, GECCO 2004
  2. Zhou Ji, Dipankar Dasgupta, Estimating the Detector Coverage in a Negative Selection Algorithm, GECCO 2005
  3. Zhou Ji, Negative Selection Algorithms: from the Thymus to V-detector, Dissertation, The University of Memphis. 2006. (You can download the PDF file here or order a nicely bound hardcopy from Lulu.com.)
  4. Zhou Ji, Dipankar Dasgupta, V-Detector : An Efficient Negative Selection Algorithm with "Probably Adequate" Detector Coverage, Information Sciences 179 (2009), pp. 1390-1406

Comments

Please feel free to email me at zhou.ji [at] yahoo.com for any questions about the info/code you find here and about my papers.