TY - GEN
T1 - Tail of Distribution GAN (TailGAN): Generative-Adversarial-Network-Based Boundary Formation
AU - Dionelis, Nikolaos
AU - Yaghoobi Vaighan, Mehrdad
AU - Tsaftaris, Sotirios
PY - 2020/11/30
Y1 - 2020/11/30
N2 - Generative Adversarial Networks (GANs) are a powerful methodology and can be used for unsupervised anomaly detection, where current techniques have limitations such as the accurate and robust detection of anomalies near the tail of a distribution. GANs generally do not guarantee the existence of a probability density and are susceptible to mode collapse, while few GANs use likelihood to reduce mode collapse. In this paper, we create a GAN-based tail formation model for anomaly detec- tion, the Tail of distribution GAN (TailGAN), to generate samples on the tail of the data distribution and detect anomalies near the support boundary. Using TailGAN, we use maximum entropy regularization and leverage GANs for anomaly detection. Using GANs that learn the probability of the underlying distribution has advantages in improving the anomaly detection methodology by allowing us to devise a generator for boundary samples, and use this model to characterize anomalies. TailGAN addresses supports with disjoint components and achieves competitive performance on images. We evaluate TailGAN for identifying Out-of-Distribution (OoD) data and its performance evaluated on MNIST, CIFAR-10, Baggage X-Ray, and OoD data shows competitiveness compared to methods from the literature.
AB - Generative Adversarial Networks (GANs) are a powerful methodology and can be used for unsupervised anomaly detection, where current techniques have limitations such as the accurate and robust detection of anomalies near the tail of a distribution. GANs generally do not guarantee the existence of a probability density and are susceptible to mode collapse, while few GANs use likelihood to reduce mode collapse. In this paper, we create a GAN-based tail formation model for anomaly detec- tion, the Tail of distribution GAN (TailGAN), to generate samples on the tail of the data distribution and detect anomalies near the support boundary. Using TailGAN, we use maximum entropy regularization and leverage GANs for anomaly detection. Using GANs that learn the probability of the underlying distribution has advantages in improving the anomaly detection methodology by allowing us to devise a generator for boundary samples, and use this model to characterize anomalies. TailGAN addresses supports with disjoint components and achieves competitive performance on images. We evaluate TailGAN for identifying Out-of-Distribution (OoD) data and its performance evaluated on MNIST, CIFAR-10, Baggage X-Ray, and OoD data shows competitiveness compared to methods from the literature.
UR - https://sspd.eng.ed.ac.uk/sites/sspd.eng.ed.ac.uk/files/attachments/sspd20eBook.pdf
UR - https://sspd.eng.ed.ac.uk/welcome-sspd-2020
UR - https://sspd.eng.ed.ac.uk/sspd-2020-presentations
UR - https://www.dropbox.com/sh/o7wyzrs94stoyip/AAAPweiKkEdcypunebOsPBPBa/Underpinning%20Signal%20Processing?dl=0&preview=WP3.1++Robust+Generative+Neural+Networks.mp4&subfolder_nav_tracking=1
U2 - 10.1109/SSPD47486.2020.9272100
DO - 10.1109/SSPD47486.2020.9272100
M3 - Conference contribution
SN - 978-1-7281-3811-4
SP - 1
EP - 5
BT - 2020 Sensor Signal Processing for Defence Conference (SSPD)
PB - Institute of Electrical and Electronics Engineers
ER -