Robust Deep Learning Methods for Anomaly Detection.
Abstract
Anomaly detection is an important problem that has been well-studied within diverse research areas and application domains. A robust anomaly detection system identifies rare events and patterns in the absence of labelled data. The identified patterns provide crucial insights about both the fidelity of the data and deviations in the underlying data-generating process. For example a surveillance system designed to monitor the emergence of new epidemics will use a robust anomaly detection methods to separate spurious associations from genuine indicators of an epidemic with minimal lag time.
The key concept in anomaly detection is the notion of "robustness", i.e., designing models and representations which are less-sensitive to small changes in the underlying data distribution. The canonical example is that the median is more robust than the mean as an estimator. The tutorial will primarily help researchers and developers design deep learning architectures and loss functions where the learnt representation behave more like the "median" rather than the "mean".
The tutorial will revisit well known unsupervised learning techniques in deep learning including autoencoders and generative adversarial networks (GANs) from the perspective of anomaly detection. This in turn will give the audience a more grounded perspective on unsupervised deep learning methods. All the methods will be introduced in a hands-on manner to demonstrate how high-level ideas and concepts get translated to practical real code.
Outline
| No | Robust Anomaly Detection | Topics | Notebook | Video |
|---|---|---|---|---|
| 1 | Deep learning Preliminaries for Anomaly Detection | Anomaly Detection Deep Models for Anomaly Detection |
Introduction | |
| 2 | Autoencoder for Anomaly Detection | Autoencoder Adversarial Auto-Encoders (AAE) Variational Auto-Encoders (VAE) Wasserstein Auto-Encoders (WAE) Deep Suport Vector Data Description (SVDD) One-Class Neural Network (OCNN) |
Notebook | 1.Setup 2. Slide Talk 3. Code Walk |
| 3 | Robust, Deep and Inductive Anomaly Detection | Robust Autoencoder Experiments |
Notebook | 1.Setup 2. Slide Talk 3. Code Walk |
| 4 | Real World UseCases | Motivation Use Cases Experiments Results |
Notebook | 1.Setup 2. Slide Talk 3. Code Walk |
Presenters
Sanjay Chawla
QCRI’s Data Analytics department.
Sanjay Chawla is Research Director of QCRI’s Data Analytics department. His research is in data mining and machine learning with a specialization in spatio-temporal data mining, outlier detection, class imbalanced classification, and adversarial learning.
Raghav Chalapathy
CSIRO Data61, Research Fellow in the Analytics and Decision Sciences Program at Data61 (CSIRO).
Dr. Raghav Chalapathy is involved with the research in Structural Health Monitoring and data analytics for asset management and energy demand forecasting. He holds a Ph.D. in data mining from the University of Sydney. His research interests are machine learning and data mining with a focus on anomaly detection, deep learning, online learning, concept drift anomaly detection, bayesian deep learning, clustering and predictive modelling. He has been driving several industrial projects investigating machine learning for real-world problems such as damage detection in civil structures (including the iconic Sydney Harbour Bridge) and energy demand forecasting for a multilevel model predictive controller used in optimal scheduling of air conditioning systems with renewable energy resource and thermal storage.
Dr. Khoa Nguyen
CSIRO Data61, Senior Research Scientist in the Analytics and Decision Sciences Program at Data61 (CSIRO).
Dr. Khoa Nguyen is a research team leader and a senior research scientist in Analytics and Decision Sciences Program at Data61 (CSIRO). He is leading a research team working on predictive analytics for asset management (including Structural Health Monitoring) and energy demand forecasting using machine learning techniques. He holds a PhD in computer science from the University of Sydney. His research interests are machine learning and data mining with a focus on tensor analysis for data fusion, anomaly detection, forecasting, imbalanced classification, data clustering and dimensionality reduction. He has been driving several industrial projects that investigate machine learning for real-world problems such as damage detection in civil structures (including the iconic Sydney Harbour Bridge), pothole detection in road pavements, fault detection and diagnosis for HVAC systems and energy demand forecasting. These project clients include government entities or private companies in transports and energy sectors (such as Transport for NSW, Transurban, Boeing, Department of the Environment and Energy and Australian Energy Market Operator).