Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Latent space conditioning for improved classification and anomaly detection

Norlander, Erik and Sopasakis, Alexandros LU (2019)
Abstract
We propose a variational autoencoder to perform improved pre-processing forclustering and anomaly detection on data with a given label. Anomalies howeverare not known or labeled. We call our method conditioned variationalautonencoder since it separates the latent space by conditioning on informationwithin the data. The method fits one prior distribution to each class in thedataset, effectively expanding the prior distribution to include a Gaussianmixture model. Our approach is compared against the capabilities of a typicalvariational autoencoder by measuring their V-score during cluster formationwith respect to the k-means and EM algorithms. For anomaly detection, we use a new metric composed of the mass-volume andexcess-mass curves which... (More)
We propose a variational autoencoder to perform improved pre-processing forclustering and anomaly detection on data with a given label. Anomalies howeverare not known or labeled. We call our method conditioned variationalautonencoder since it separates the latent space by conditioning on informationwithin the data. The method fits one prior distribution to each class in thedataset, effectively expanding the prior distribution to include a Gaussianmixture model. Our approach is compared against the capabilities of a typicalvariational autoencoder by measuring their V-score during cluster formationwith respect to the k-means and EM algorithms. For anomaly detection, we use a new metric composed of the mass-volume andexcess-mass curves which can work in an unsupervised setting. We compare theresults between established methods such as as isolation forest, local outlierfactor and one-class support vector machine. (Less)
Please use this url to cite or link to this publication:
author
and
organization
publishing date
type
Other contribution
publication status
published
subject
keywords
VAE, K-means, V-score, F1 score, conditioned, anomaly detection, Gaussian mixture model, isolation forest, EM algorithm, one class SVM, Unsupervised classification
language
English
LU publication?
yes
id
103007aa-61e7-40e4-9498-a1a00fa89c92
alternative location
https://arxiv.org/abs/1911.10599
date added to LUP
2019-11-26 06:57:04
date last changed
2020-12-03 13:25:34
@misc{103007aa-61e7-40e4-9498-a1a00fa89c92,
  abstract     = {{We propose a variational autoencoder to perform improved pre-processing forclustering and anomaly detection on data with a given label. Anomalies howeverare not known or labeled. We call our method conditioned variationalautonencoder since it separates the latent space by conditioning on informationwithin the data. The method fits one prior distribution to each class in thedataset, effectively expanding the prior distribution to include a Gaussianmixture model. Our approach is compared against the capabilities of a typicalvariational autoencoder by measuring their V-score during cluster formationwith respect to the k-means and EM algorithms.  For anomaly detection, we use a new metric composed of the mass-volume andexcess-mass curves which can work in an unsupervised setting. We compare theresults between established methods such as as isolation forest, local outlierfactor and one-class support vector machine.}},
  author       = {{Norlander, Erik and Sopasakis, Alexandros}},
  keywords     = {{VAE; K-means; V-score; F1 score; conditioned; anomaly detection; Gaussian mixture model; isolation forest; EM algorithm; one class SVM; Unsupervised classification}},
  language     = {{eng}},
  title        = {{Latent space conditioning for improved classification and anomaly detection}},
  url          = {{https://lup.lub.lu.se/search/files/87558700/1911.10599.pdf}},
  year         = {{2019}},
}