Stochastic Approximation vis-a-vis Online Learning for Big Data Analytics [Lecture Notes]

We live in an era of data deluge, where data translate to knowledge and can thus contribute in various directions if harnessed and processed intelligently. There is no doubt that signal processing (SP) is of uttermost relevance to timely big data applications such as real-time medical imaging, smart cities, network state visualization and anomaly detection (e.g., in the power grid and the Internet), health informatics for personalized treatment, sentiment analysis from online social media, Web-based advertising, recommendation systems, sensor-empowered structural health monitoring, and e-commerce fraud detection, just to name a few. Accordingly, abundant chances unfold to SP researchers and practitioners for fundamental contributions in big data theory and practice.