Using AI for Radio (Big) Data

The use of artificial intelligence, more specifically of algorithms based on modern machine learning and neural networks, has recently seen accelerated use in astronomy. At the very same time we enter a true big data era in radio astronomy, where data-rates of terabits per second are reached in the years to come. In this chapter we present key applications of modern machine learning, from the detection and characterisation of radio signals of different domains (image, time, tomography), the classification of complex radio source morphologies, and inference of both radio source and large-scale radio map properties. We show that these methods are able to process radio astronomical data both fast and accurate, while offering flexibility and versatility to learn complex representations and shapes of radio signals. We highlight how dimensionality reduction, anomaly detection and uncertainty estimation based on machine learning controls the quality, robustness and understanding of AI-derived radio astronomical results.