Lecture
This lecture by the instructor covers the challenges and solutions in analyzing big multi-dimensional data, focusing on the deluge of complex data types like time series, text, audio, images, and videos. It explores the ubiquity of data sequences and the need for automated tools to detect anomalies in streaming data. The talk delves into the SAND framework for streaming subsequence anomaly detection, highlighting the complexities of time-series anomalies and the importance of context-aware solutions. It also discusses state-of-the-art methods like Product Quantization and Variance-Aware Quantization to improve accuracy and runtime in similarity search tasks.