We enable better decisions through the development and application of state-of-the-art techniques in machine learning, statistics, and decision sciences. Combining expert knowledge with data and modeling, we give decision makers the tools required to turn large, highly-complex and multi-modal data sets into information that can be used to make actionable decisions, optimally deploy resources, and make informed tradeoffs.
The Applied Statistics group's mission is to bring statistical rigor and innovation to bear on complex problems of national importance in a highly collaborative environment. The group is distinguished by its strong focus on interdisciplinary research, which is made possible by the opportunity to work with a diverse group of leading scientists and engineers on challenging problems of national importance. In addition to the statistical research staff, we host a number of post-doctoral researchers, students, and visiting scholars maintaining strong ties to the academic community as well as our partners at other national laboratories.
The group's work ranges from introducing existing statistical methodology in novel application areas to developing new statistical theory motivated by unique and challenging data sets. Some examples of active research areas include: design of computer experiments, climate modeling, cyber security, image and video analysis, energy analysis, computational biology, lasers and optics, uncertainty quantification, computer model validation/calibration, and statistical computing on HPC. A common theme is that statistical applications at LLNL require scaling statistics and machine learning algorithms to operate on extremely large data sets generated by state-of-the-art collection technologies.
Group members work with a wide variety of collaborators and sponsors on projects ranging from fast-paced consulting activities to multi-year collaborations. Statistical analysis and methodologies developed at LLNL have been used by government decision makers at the national and international levels, and the group has also formed successful partnerships in the energy and healthcare industries.
The Computer Vision group provides technical expertise in computer vision and video analytics—ranging from object recognition and tracking to unsupervised feature learning for multimedia data-to address problems of national interest. Our focus application areas include processing and exploitation of overhead imagery & video as well as large-scale video search, retrieval, and indexing.
The Cyber Operations group focuses on programmatic applied research in the characterization and defense of computer hosts and networks. We work closely with the network mapping and machine learning groups. Emphases include distributed data processing, sensor fusion, scalable attribute graphs, cloud processing technologies, stream processing technologies, host-based security, vulnerability characterization, network simulation/virtualization/visualization, protocol parsing, and malware analysis.
The Energy and Earth Systems group is focused on computational modeling of (1) physical and chemical processes in the earth and atmosphere, and (2) energy systems and resources. They have a strong presence in the areas of critical infrastructure protection and systems security. Energy and Earth Systems skills include expertise in fluid flow and transport, reactive chemistry, uncertainty quantification, as well as systems/operations integration and optimization. Primary programmatic activity for Energy and Earth Systems covers the breadth of Global Security programs including NProgram (NARAC, NCT), SProgram (EISDHS/ S&T and FEMA), and EProgram energy projects..
The Machine Learning group has expertise in developing and tailoring Machine Learning algorithms to solve classification, anomaly/change detection, prediction, and clustering tasks on "Big Data". Our researchers have extensive experience with Random Forests, Deep Neural Networks, SVMs, Mixture Models, HMMs and Dynamic Bayesian Networks, Ensemble Methods, and Big Data frameworks like Hadoop, Storm, and Spark.
The Operations Research and Systems Analysis group is a multidisciplinary team of engineers and applied mathematicians who formulate and analyze mathematical models of complex systems in order to enhance performance. We use a wide range of mathematical and computational approaches including optimization, stochastic modeling, simulation, algorithms, and graph theory. The modeling and analysis performed is broadly applicable and is currently being used to solve problems in areas such as chemical, biological, radiological and nuclear terrorism, cyber security, power systems and energy grids, and nuclear enterprise planning.
The Quantitative Risk Analysis group utilizes computational methods to analyze a variety of risks associated with the design and operation of radiological, nuclear, and chemical facilities. Our expertise includes Monte Carlo simulations of radiation transport, safety compliance for packaging and transportation of radioactive materials, nuclear safeguards technology development, system engineering and statistical analysis for nuclear material management, gamma spectroscopic analysis, hazard and accident analyses, safety basis documentation, quantification of margin and uncertainty, and sensitivity analysis.