eduzhai > Applied Sciences > Engineering >

Robust Variational Autoencoder for Tabular Data with Beta Divergence

  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: We propose a robust variational autoencoder with $ beta$ divergence fortabular data (RTVAE) with mixed categorical and continuous features.Variational autoencoders (VAE) and their variations are popular frameworks foranomaly detection problems. The primary assumption is that we can learnrepresentations for normal patterns via VAEs and any deviation from that canindicate anomalies. However, the training data itself can contain outliers. Thesource of outliers in training data include the data collection process itself(random noise) or a malicious attacker (data poisoning) who may target todegrade the performance of the machine learning model. In either case, theseoutliers can disproportionately affect the training process of VAEs and maylead to wrong conclusions about what the normal behavior is. In this work, wederive a novel form of a variational autoencoder for tabular data sets withcategorical and continuous features that is robust to outliers in trainingdata. Our results on the anomaly detection application for network trafficdatasets demonstrate the effectiveness of our approach.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...