Our mission at the Data Science Institute (DSI) is to enable excellence in data science research and applications across the Laboratory's core missions.
Data science has become an essential discipline paving the path of LLNL's key program areas, and the Laboratory is home to some of the largest, most unique, and most interesting data and supercomputers in the world. The DSI acts as the central hub for all data science activity—in areas of artificial intelligence, big-data analytics, computer vision, machine learning, predictive modeling, statistical inference, uncertainty quantification, and more—at LLNL working to help lead, build, and strengthen the data science workforce, research, and outreach to advance the state-of-the-art of our nation's data science capabilities. Read more about the DSI.
Data Scientist Spotlight
To understand the science of nuclear weapons without underground testing, researchers at LLNL’s National Ignition Facility conduct inertial confinement fusion (ICF) experiments and optimize target designs for higher energy yields. Design physicist Kelli Humbird has developed machine learning models that accurately predict energy yield and reveal the surprising potential of ovoid-shaped targets. She explains, “This is a great example of what machine learning can do because it has no biases. It directed us to a design space we would not have typically considered.” She presented her research on transfer learning for ICF applications at LLNL’s 2020 Women in Data Science regional conference and says, “I feel really lucky to be a part of such cool science.” A former Lab intern and Livermore Graduate Scholar, Humbird holds a PhD in Nuclear Engineering and Physics from Texas A&M University and recently received her alma mater’s Nuclear Engineering 2020/2021 Young Former Student Award.
New Research in AI
Deep Learning (DL) models are proving useful for a number of materials science applications including materials discovery, microstructure analysis, and property predictions. In a recent publication, LLNL researchers propose a unified framework that leverages the predictive uncertainty from deep neural networks to answer challenging questions materials scientists usually encounter in machine learning–based material application workflows. Specifically, the team demonstrates that predictive uncertainty from uncertainty-aware DL approaches (particularly deep ensembles) can be used to determine the number of required training data to achieve the desired prediction accuracy without relying on labelled data. Further, the team shows that the predictive uncertainty-guided decision referral is highly effective in detecting and refraining deep neural networks from making wrong predictions on confusing material samples and out-of-distribution samples that deviate from the training data. The newly proposed uncertainty-enabled decision-making method is quite generic and can be used in a wide range of scientific domains to ensure trust, dependability, and usefulness of DL models.