Introduction to Online Optimization

In a world where automatic data collection becomes ubiquitous, statisticians must update their paradigms to cope with new problems. Whether we discuss the Internet network, consumer data sets, or financial market, a common feature emerges: huge amounts of dynamic data that need to be understood and quickly processed. This state of affair is dramatically different from the classical statistical problems, with many observations and few variables of interest. Over the past decades, learning theory tried to address this issue. One of the standard and thoroughly studied models for learning is the framework of statistical learning theory. We start by briefly reviewing this model.