Exploring complex pattern formation with convolutional neural networks

Abstract

Many nonequilibrium systems, such as biochemical reactions and socioeconomic interactions, can be described by reaction-diffusion equations that demonstrate a wide variety of complex spatiotemporal patterns. The diversity of the morphology of these patterns makes it difficult to classify them quantitatively and they are often described visually. Hence, searching through a large parameter space for patterns is a tedious manual task. We discuss how convolutional neural networks can be used to scan the parameter space, investigate existing patterns in more detail, and aid in finding new groups of patterns. As an example, we consider the Gray-Scott model for which training data is easy to obtain. Due to the popularity of machine learning in many scientific fields, well maintained open source toolkits are available that make it easy to implement the methods we discuss in advanced undergraduate and graduate computational physics projects.

References in corpus (6)

• Solving high-dimensional partial differential equations using deep learning
• Unmasking Clever Hans Predictors and Assessing What Machines Really Learn
• Complex Patterns in a Simple System
• Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps
• The importance of transparency and reproducibility in artificial intelligence research
• Stable localized moving patterns in the 2-D Gray-Scott model