Greater scientific insight is fundamental to enhancing our understanding of the earth and environment, health and well-being. Scientists adopt a variety of tools to improve our knowledge. In 2009 Microsoft Research produced a series of essays around the topic of Data-Driven Discovery. This concept of extracting knowledge directly from data at an extreme scale was termed the “fourth paradigm”. The first three paradigms were experimental, theoretical, and, more recently, computational science. Dr. Jim Gray, a database software pioneer and a Microsoft researcher summarised the first three paradigms in the excellent graphic.

It illustrates the major shift that has taken place in the past decade from the computational era towards the data-driven era. Since the 1960's immense effort has gone into creating (and maintaining) complex computer simulations of all aspects of our world such as weather, biological, and molecular. Because complex, detailed, and refined physics-based models can be prohibitively computationally expensive, a new approach to estimating their simulations with machine-learning (ML) is growing in popularity. Excellent textbooks are available to those interested in the theory and development of machine learning (Goodfellow et. al. (2016), LeCun et. al. (2015), and Schmidhuber (2015)). 

Machine learning is founded on the concept of learning complex, nonlinear relationships between inputs and outputs. Instead of explicitly programming these complex relationships as we did in the past, we instead let the model learn these patterns by showing it multiple features and corresponding labels. Once trained, the model can predict the correct output or label for new inputs or features that are provided.

In our next tutorial, we explore how Julia can be used to develop a number of different machine learning models

References:

Goodfellow, Ian, Yoshua Bengio, and Aaron Courville. Deep learning. MIT press, 2016.

LeCun, Yann, Yoshua Bengio, and Geoffrey Hinton. "Deep learning." nature 521.7553 (2015): 436-444.

Schmidhuber, Jürgen. "Deep learning in neural networks: An overview." Neural networks 61 (2015): 85-117.