Support Measure Data Description for group anomaly detection

Abstract : We address the problem of learning a data description model from a dataset containing probability measures as observations. We estimate the data description model by optimizing volume-sets of probability measures where each volume-set is defined as a set of probability measures whose representative functions in a reproducing kernel Hilbert space (RKHS) belong to an enclosing ball. We present three data description models, which are functions in a RKHS depending only on some probability measures, named support measures in analogy to support vectors. An advantage of the method is that we do not consider any particular form for the probability measures. We validate our method in the task of group anomaly detection, with artificial and real datasets.
Document type :
Conference papers
Complete list of metadatas

Cited literature [35 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01330487
Contributor : Jorge Guevara <>
Submitted on : Friday, June 10, 2016 - 5:46:56 PM
Last modification on : Thursday, February 7, 2019 - 5:46:39 PM

File

sig-alternate.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01330487, version 1

Citation

Jorge Guevara, Stéphane Canu, R Hirata. Support Measure Data Description for group anomaly detection. ODDx3 Workshop on Outlier Definition, Detection, and Description at the 21st ACM SIGKDD INTERNATIONAL CONFERENCE ON KNOWLEDGE DISCOVERY AND DATA MINING (KDD2015), Aug 2015, Sydney, Australia. ⟨hal-01330487⟩

Share

Metrics

Record views

327

Files downloads

365