The next generation of modeling & simulation: integrating big data and deep learning

Big data allows users to cope with data that are huge in regards to volume, velocity, variety, and veracity. It provides methods and tools to extract aggregates and new information out of heterogeneously structured data or even completely unstructured data. Deep learning is a collection of algorithms that allows us to discover correlations and learn --- supervised and unsupervised --- from information provided. This contribution introduces the main ideas and methods of big data and deep learning and shows how they can be applied to various phases of the traditional modeling and simulation process. Big data supports obtaining data for the initialization as well as evaluating the results of the simulation experiment. Deep learning can help with the conceptual modeling phase as well as with the discovery of correlations in the results. Examples of existing applications will be given to prove the feasibility of such ideas. This leads to the observation that big data, deep learning, and modeling and simulation have the potential to lead to a new generation of modeling and simulation applications that provide computational scientific support on a new scale beyond the current capabilities.