Anomaly detection for imbalanced datasets with deep generative models

N.R. Santos Buitrago, L.M.A. Tonnaer, V. Menkovski, Dimitrios Mavroeidis

Research output: Contribution to conferencePaperAcademic

Abstract

Many important data analysis applications present with severely imbalanced datasets with respect to the target variable. A typical example is medical image analysis, where positive samples are scarce, while performance is commonly estimated against the correct detection of these positive examples. We approach this challenge by formulating the problem as anomaly detection with generative models. We train a generative model without supervision on the ‘negative’ (common) datapoints and use this model to estimate the likelihood of unseen data. A successful model allows us to detect the ‘positive’ case as low likelihood
datapoints.
In this position paper, we present the use of state-of-the-art deep generative models (GAN and VAE) for the estimation of a likelihood of the data. Our results show that on the one hand both GANs and VAEs are able to separate the ‘positive’ and ‘negative’ samples in the MNIST case. On the other hand, for the NLST case, neither GANs nor VAEs were able to capture the complexity of the data and discriminate anomalies at the level that this task requires. These results show that even though there are a number of successes presented in the literature for using generative models in similar applications, there remain further challenges for broad successful implementation.

Conference

ConferenceBenelearn 2018 -- Annual Machine Learning Conference of Belgium and the Netherlands
Abbreviated titleBenelearn 2018
CountryNetherlands
City's-Hertogenbosch
Period8/11/189/11/18
Internet address

Fingerprint

Image analysis

Cite this

Santos Buitrago, N. R., Tonnaer, L. M. A., Menkovski, V., & Mavroeidis, D. (2018). Anomaly detection for imbalanced datasets with deep generative models. Paper presented at Benelearn 2018 -- Annual Machine Learning Conference of Belgium and the Netherlands, 's-Hertogenbosch, Netherlands.
Santos Buitrago, N.R. ; Tonnaer, L.M.A. ; Menkovski, V. ; Mavroeidis, Dimitrios. / Anomaly detection for imbalanced datasets with deep generative models. Paper presented at Benelearn 2018 -- Annual Machine Learning Conference of Belgium and the Netherlands, 's-Hertogenbosch, Netherlands.15 p.
@conference{e771ad5352914246bfcf04915ea62b87,
title = "Anomaly detection for imbalanced datasets with deep generative models",
abstract = "Many important data analysis applications present with severely imbalanced datasets with respect to the target variable. A typical example is medical image analysis, where positive samples are scarce, while performance is commonly estimated against the correct detection of these positive examples. We approach this challenge by formulating the problem as anomaly detection with generative models. We train a generative model without supervision on the ‘negative’ (common) datapoints and use this model to estimate the likelihood of unseen data. A successful model allows us to detect the ‘positive’ case as low likelihooddatapoints.In this position paper, we present the use of state-of-the-art deep generative models (GAN and VAE) for the estimation of a likelihood of the data. Our results show that on the one hand both GANs and VAEs are able to separate the ‘positive’ and ‘negative’ samples in the MNIST case. On the other hand, for the NLST case, neither GANs nor VAEs were able to capture the complexity of the data and discriminate anomalies at the level that this task requires. These results show that even though there are a number of successes presented in the literature for using generative models in similar applications, there remain further challenges for broad successful implementation.",
author = "{Santos Buitrago}, N.R. and L.M.A. Tonnaer and V. Menkovski and Dimitrios Mavroeidis",
year = "2018",
month = "9",
day = "8",
language = "English",
note = "Benelearn 2018 -- Annual Machine Learning Conference of Belgium and the Netherlands, Benelearn 2018 ; Conference date: 08-11-2018 Through 09-11-2018",
url = "https://bnaic2018.nl/",

}

Santos Buitrago, NR, Tonnaer, LMA, Menkovski, V & Mavroeidis, D 2018, 'Anomaly detection for imbalanced datasets with deep generative models' Paper presented at Benelearn 2018 -- Annual Machine Learning Conference of Belgium and the Netherlands, 's-Hertogenbosch, Netherlands, 8/11/18 - 9/11/18, .

Anomaly detection for imbalanced datasets with deep generative models. / Santos Buitrago, N.R.; Tonnaer, L.M.A.; Menkovski, V.; Mavroeidis, Dimitrios.

2018. Paper presented at Benelearn 2018 -- Annual Machine Learning Conference of Belgium and the Netherlands, 's-Hertogenbosch, Netherlands.

Research output: Contribution to conferencePaperAcademic

TY - CONF

T1 - Anomaly detection for imbalanced datasets with deep generative models

AU - Santos Buitrago,N.R.

AU - Tonnaer,L.M.A.

AU - Menkovski,V.

AU - Mavroeidis,Dimitrios

PY - 2018/9/8

Y1 - 2018/9/8

N2 - Many important data analysis applications present with severely imbalanced datasets with respect to the target variable. A typical example is medical image analysis, where positive samples are scarce, while performance is commonly estimated against the correct detection of these positive examples. We approach this challenge by formulating the problem as anomaly detection with generative models. We train a generative model without supervision on the ‘negative’ (common) datapoints and use this model to estimate the likelihood of unseen data. A successful model allows us to detect the ‘positive’ case as low likelihooddatapoints.In this position paper, we present the use of state-of-the-art deep generative models (GAN and VAE) for the estimation of a likelihood of the data. Our results show that on the one hand both GANs and VAEs are able to separate the ‘positive’ and ‘negative’ samples in the MNIST case. On the other hand, for the NLST case, neither GANs nor VAEs were able to capture the complexity of the data and discriminate anomalies at the level that this task requires. These results show that even though there are a number of successes presented in the literature for using generative models in similar applications, there remain further challenges for broad successful implementation.

AB - Many important data analysis applications present with severely imbalanced datasets with respect to the target variable. A typical example is medical image analysis, where positive samples are scarce, while performance is commonly estimated against the correct detection of these positive examples. We approach this challenge by formulating the problem as anomaly detection with generative models. We train a generative model without supervision on the ‘negative’ (common) datapoints and use this model to estimate the likelihood of unseen data. A successful model allows us to detect the ‘positive’ case as low likelihooddatapoints.In this position paper, we present the use of state-of-the-art deep generative models (GAN and VAE) for the estimation of a likelihood of the data. Our results show that on the one hand both GANs and VAEs are able to separate the ‘positive’ and ‘negative’ samples in the MNIST case. On the other hand, for the NLST case, neither GANs nor VAEs were able to capture the complexity of the data and discriminate anomalies at the level that this task requires. These results show that even though there are a number of successes presented in the literature for using generative models in similar applications, there remain further challenges for broad successful implementation.

M3 - Paper

ER -

Santos Buitrago NR, Tonnaer LMA, Menkovski V, Mavroeidis D. Anomaly detection for imbalanced datasets with deep generative models. 2018. Paper presented at Benelearn 2018 -- Annual Machine Learning Conference of Belgium and the Netherlands, 's-Hertogenbosch, Netherlands.