Topic: AI/ML

The 2021 Conference on Computer Vision and Pattern Recognition, the premier conference of its kind, will feature two papers co-authored by an LLNL researcher targeted at improving the understanding of robust machine learning models.

The ADAPD program held a two-day virtual meeting to highlight science-based and data-driven analysis work to advance AI innovation and develop AI-enabled systems to enhance the U.S. capability to detect nuclear proliferation activities around the globe.

New research debuting at ICLR 2021 demonstrates a learning-by-compressing approach to deep learning that outperforms traditional methods without sacrificing accuracy.

LLNL is looking for participants and attendees from industry, research institutions and academia for the first-ever Machine Learning for Industry Forum (ML4I), a three-day virtual event starting Aug. 10. The event is sponsored by LLNL’s High Performance Computing Innovation Center and the Data Science Institute.

Highlights include scalable deep learning, high-order finite elements, data race detection, and reduced order models.

This project aims to tackle the complexities of HPC software integration with dependency compatibility models, binary analysis tools, efficient logic solvers, and configuration optimization techniques.

Led by computational scientist Youngsoo Choi, the Data-Driven Physical Simulation reading group has been meeting biweekly since October 2019. The pandemic almost disbanded the group... until it turned into a virtual seminar series.

In his opening keynote address at the AI Systems Summit, LLNL CTO Bronis de Supinski described integration of two AI-specific systems to achieve system level heterogeneity.

The Accelerating Therapeutics for Opportunities in Medicine consortium, of which LLNL is part, announced the U.S. Department of Energy’s Argonne, Brookhaven and Oak Ridge national labs are joining the consortium to further develop ATOM’s AI-driven drug discovery platform.

Three papers address feature importance estimation under distribution shifts, attribute-guided adversarial training, and uncertainty matching in graph neural networks.

StarSapphire is a collection of scientific data mining projects focusing on the analysis of data from scientific simulations, observations, and experiments.

The 34th Conference on Neural Information Processing Systems features two papers advancing the reliability of deep learning for mission-critical applications at LLNL.

LLNL will collaborate with Machina Labs to apply ML to aluminum sheet metal processing for aerospace and automotive applications. Five recently announced LLNL-led projects will be funded by HPC4EI.

LLNL has installed a new AI accelerator into the Corona supercomputer, allowing researchers to run simulations while offloading AI calculations from those simulations to the AI system.

CASC researcher Harsh Bhatia thrives in the Lab’s versatile research environment. “At the Lab, no two problems are the same. Therefore, as a team, researchers deliver hundreds of new data science solutions each year. We are very fortunate to have access to many high-impact projects so we can really make a difference with our data science or data analysis solutions," he says.

Computing’s summer hackathon was held virtually on August 6–7 and featured presentations from teams who tested software technologies, expanded project features, or explored new ways of analyzing data.

Ian Karlin on AI hardware integration into HPC systems, workflows, followed by a talk about software integration of AI accelerators in HPC with Brian Van Essen.

Lawrence Livermore National Lab has named Stefanie Guenther as Computing’s fourth Sidney Fernbach Postdoctoral Fellow in the Computing Sciences. This highly competitive fellowship is named after LLNL’s former Director of Computation and is awarded to exceptional candidates who demonstrate the potential for significant achievements in computational mathematics, computer science, data science, or scientific computing.

Two papers featuring LLNL scientists were accepted in the 2020 International Conference on Machine Learning (ICML), one of the world’s premier conferences of its kind.

Highlights include response to the COVID-19 pandemic, high-order matrix-free algorithms, and managing memory spaces.

A team led by an LLNL computer scientist proposes a deep learning approach aimed at improving the reliability of classifier models for predicting disease types from diagnostic images.

LLNL scientists have taken a step forward in the design of future materials with improved performance by analyzing its microstructure using artificial intelligence.

LLNL's Jay Thiagarajan joins the Data Skeptic podcast to discuss his recent paper "Calibrating Healthcare AI: Towards Reliable and Interpretable Deep Predictive Models." The episode runs 35:50.

An LLNL team developed ML tools that extract and structure information from the text and figures of nanomaterials articles using NLP, image analysis, computer vision, and visualization techniques.

In this video from the Stanford HPC Conference, Katie Lewis presents "The Incorporation of Machine Learning into Scientific Simulations at Lawrence Livermore National Laboratory."