The tutorial series is aimed at researchers who have little or no experience using AI tools in their research, with a goal of providing background, tools, and hands-on experience that will enable researcher to add AI tools to their standard tool sets. Throughout the tutorials, we will use data sets generated at Argonne with examples drawn from research done at Argonne. The first tutorial used an example of going from data sets to descriptors, discussion about how to generate “good” descriptors, building machine learning models using several different regression techniques, such as random forest and kernel ridge regression, evaluating the results, and making predictions outside of the domain of the training data. The second tutorial in the series introduced convolutional neural networks and showed how these can be used to de-noise data. An example was given using TomoGAN that also includes generative adversarial for more advanced de-noising. The third of a series of hands-on AI4Science tutorials will be given in two parts on offered on May 19 and May 26 from 1:30 pm to 3:30 pm and will deal with Bayesian optimization and Gaussian process regression. These are powerful tools that can help guide an experiment or a computational campaign to points in parameter space that add best value in terms of reducing uncertainties of a model used to represent the results. On May 19, we will go through an introduction to Bayesian optimization, and May 26 will be dedicated to an example based on research at Argonne. The example, adapted from one of the top 5 publications in JCESR in 2020, is a step-by-step demonstration of how Bayesian optimization may be used to accelerate the search for optimal electrolyte candidates in non-aqueous redox flow batteries.
- Please visit the following page, set up your google account to use the colab platform https://colab.research.google.com/
- Please follow the instructions on the page before you come to the tutorial: https://github.com/AIScienceTutorial/Material_Science
*Please note, the AI Tutorial For Science Hands-On Session is open to Argonne employees only*
