Sequential Clustering of Data Streams from Unknown Distributions

Learn about sequential clustering techniques for data streams in this technical lecture from Prof. Srikrishna Bhashyam of IIT Madras. Explore methods for clustering multiple data streams into groups based on their underlying distribution similarities, with a focus on sequential testing where new sample sets arrive at each time step. Discover universal nonparametric clustering tests that work independently of distribution configurations for both known and unknown numbers of clusters. Examine how these sequential tests achieve finite stopping times and universal exponential consistency, while outperforming fixed sample size tests in terms of expected sample numbers for given error probabilities. Gain insights from the speaker's extensive experience in communication theory, signal processing, and wireless networks, developed through his work at IIT Madras and Qualcomm Inc.