From Signal Processing to Machine Learning

This chapter reviews the main landmarks of signal processing in the 20th century from the perspective of algorithmic developments. It focuses on cross‐fertilization with the field of statistical (machine) learning in the last decades. In the 21st century, model and data assumptions as well as algorithmic constraints are no longer valid, and the field of machine‐learning signal processing has erupted, with many successful stories to tell. The chapter also focuses on digital signal processing (DSP), which deals with the analysis of digitized and discrete sampled signals. Machine learning is a branch of computer science and artificial intelligence that enables computers to learn from data. Machine learning adequately fits the constraints and solution requirements posed by DSP problems: from computational efficiency, online adaptation, and learning with limited supervision, to their ability to combine heterogeneous information, to incorporate prior knowledge about the problem, or to interact with the user to achieve improved performance.