A Review of Machine Learning Techniques for Anomaly Detection - Dr. David Green

Explore a comprehensive review of machine learning techniques for anomaly detection in this 22-minute seminar by Dr. David Green from the Alan Turing Institute. Delve into various aspects of anomaly detection, including point, contextual, and collective anomalies. Learn about traditional decomposition methods and the application of deep neural networks in this field. Discover the differences between supervised and unsupervised learning approaches, and understand how they apply to anomaly detection. Examine clustering techniques, including traditional and spectral methods, as well as time series analysis. Address challenges and risks associated with anomaly detection in large-scale projects, one-shot projects, IT infrastructure security, and smart cities. Gain insights into the latest technology trends and their impact on machine learning for anomaly detection.

Introduction
Technology trends
What is machine learning
Traditional decomposition
Point anomalies
Contextual anomalies
Collective anomalies
Deep neural networks
Two styles of explanation
Training a neural network
Hierarchical classification
Background problem categories
Supervised learning
Project forward in time
Unsupervised learning
Traditional clustering
Time series type analysis
Spectral clustering
False positives
Challenges and risks
Large projects
Oneshot projects
IT infrastructure security
Smart cities
The Churring