Outlier detection using autoencoders

Outlier detection is a crucial part of any data analysis applications. The goal of outlier detection is to separate a core of regular observations from some polluting ones, called “outliers”. We propose an outlier detection method using deep autoencoder. In our research the invented method was applied to detect outlier points in the MNIST dataset of handwriting digits. The experimental results show that the proposed method has a potential to be used for anomaly detection.