Machine learning algorithms in astronomy

We analyze the current state and challenges of machine learning techniques in astronomy, in order to keep up to date with the exponentially increasing amount of observational data and future instrumentation of at least a few orders of magnitude higher than from current instruments. We present the latest cutting-edge methods and new algorithms for extracting knowledge from large and complex astronomical data sets, as a versatile and valuable tool in a new era of data-driven science. As telescopes and detectors become more powerful, the volume of available data will enter the petabyte regime, in need for new algorithms aimed at better modeling of sky brightness, requiring more computing and providing more promising data analysis for billions of sky objects. We emphasize the most important current trends and future directions in machine learning currently adapted to astronomy, pushing forward the frontiers of knowledge through better data collection, manipulation, analysis and visualizing. We also evaluate the emergent techniques and latest approaches for various types and sizes of data sets and show the vast potential and versatility of machine learning algorithms in front of the new challenge of the Fourth Paradigm.