(462a) Invited Talk: Data Mining and Machine Learning in Colloidal Science*

There is an enormous design space available to modern colloidal matter, where particle shape and interparticle interactions may be manipulated both in experiment and in computer simulation to span a nearly infinite range of possibilities. Traditional computer simulation methods, such as Monte Carlo and molecular dynamics, have revealed much about the self-assembly behavior of colloidal and patchy particles, and with fast GPU acceleration are producing huge amounts of data on more than 60,000 shapes to date. So what to do with all that data?  Although data mining and machine learning techniques such as SVM, neural networks, and genetic and evolutionary algorithms have been used often in bioinformatics and now in the discovery of new “hard” materials, they have yet to be fully exploited in soft matter systems. In this talk we discuss the potential of these powerful techniques in the context of colloid and nanoparticle assembly, and particle design, and present results on a range of systems where data-driven techniques are helping to provide important physical insight.

*with Carl Simon Adorf, Mayank Agrawal, Pablo Damasceno, Paul Dodd, Yina Geng, and Matthew Spellings.