dragon-dane / anomaly-detection_varitaional-autoencoders Goto Github PK

A small script to demonstrate the use of Variational Autoencoders (VAE) for detecting anomalies in MNIST dataset, the anomnalies being the absence of a particular digit in the training dataset and its presence in the test dataset. The VAE attempts to identify such a digit as an anomaly.

Other resources on anomaly detection based on Bayesian Inference:

Vessel Tracks: https://www.sciencedirect.com/science/article/pii/S0888613X13000728

Using Auto-Encoder: https://arxiv.org/pdf/1802.03903.pdf

In mixed Datasets: https://pdfs.semanticscholar.org/ddaf/628bcdb20c44f93477c5b367987b231db324.pdf

BNs + GMMs: http://www-sop.inria.fr/oasis/Vercors/papers/Submit-AAI05.pdf

Review: https://www.cse.iitb.ac.in/~comad/2010/ResearchTrack/paper%2017.pdf

Network IDS: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=4573059&tag=1

Cable Bridge: https://www.tandfonline.com/doi/abs/10.1080/15732479.2014.951867

http://www.dtic.mil/dtic/tr/fulltext/u2/a610860.pdf