《nature14541-Probabilistic machine learning and artificial intelligence》.pdf

REVIEW doi:10.1038/nature14541 Probabilistic machine learning and artificial intelligence Zoubin Ghahramani1 How can a machine learn from experience? Probabilistic modelling provides a framework for understanding what learn- ing is, and has therefore emerged as one of the principal theoretical and practical approaches for designing machines that learn from data acquired through experience. The probabilistic framework, which describes how to represent and manipulate uncertainty about models and predictions, has a central role in scientific data analysis, machine learning, robotics, cognitive science and artificial intelligence. This Review provides an introduction to this framework, and dis - cusses some of the state-of-the-art advances in the field, namely, probabilistic programming, Bayesian optimization, data compression and automatic model discovery. he key idea behind the probabilistic framework to machine learn- observed data. Typical examples of such tasks might include detecting ing is that learning can be thought of as inferring plausible models pedestrians in images taken from an autonomous vehicle, classifying Tto explain observed data. A machine can use such models to make gene-expression patterns from leukaemia patients into subtypes by clin- predictions about future data, and take decisions that are rational given ical outcome, or translating English sentences into French. However, as these predictions. Uncertainty plays a fundamental part in all of this. I discuss, the scope of machine-learning tasks is even broader than these Observed data can be consistent with many models, and therefore which pattern classification or mapping tasks, and can include optimization model is appropriate