Processing of Large Volumes of Data

This chapter addresses the processing of large volumes of data that raises theoretical, algorithmic, and computational issues. It discusses parallel computing, which can, under certain conditions, accelerate calculations and then turns to distributed computing which aims at treating massive data by distributing them on several machines. The chapter then explores cloud computing, which allows the greatest number of people to access machine learning and deep learning resources, and a great deal of computing power. It also presents some processors which are particularly adapted to deep learning. The chapter describes the open source tools R and Python: they are both software and languages, and they are currently the most used in data science. Finally, it mentions the promises of quantum computing. A parsimonious model is a model that uses only a part of the available data: the explanatory variables deemed most relevant to the desired objective are selected.