Skip to main content

News

AI-based Approach Speeds Diagnosis of I/O Performance Bottlenecks in HPC

US-RSE Works to Bring Research Software Engineers into the Spotlight

The first US-RSE member meeting was held in Chicago in 2023. (Credit: Keith Beattie)

Simulating Plasma, NERSC Systems Enable Efficient Microchip Production

STRUDEL Aims to Improve Scientific Software Design with Unique UX/UI Tool

2024 CSA Postdoc Symposium Showcases 26 Early Career Researchers

During the symposium, postdoctoral researchers currently working at the Lab shared 10-minute slide presentations on their projects with an audience of peers, mentors, and coworkers, followed by interactive Q&As throughout the day

Machine Learning Yields More Efficient Hydrogen Combustion Reactivity Modeling

Berkeley Lab Affiliate Giulia Guidi Awarded SIAM’s Supercomputing Early Career Prize

Machine Learning Opens New Doors in Traumatic Brain Injury Research

TBI patient outcomes can be predicted from patient intake data with a high degree of precision by applying a novel combination of non-parametric unsupervised and supervised machine learning. Left: Fold-over-chance accuracy of predicting outcomes into a different number of clusters. Right: Two-dimensional UMAP (Uniform Manifold Approximation and Projection) embedding clustered into 19 clusters. (Credit: Andrew Tritt and Kristofer Bouchard, Lawrence Berkeley National Laboratory)

In a Warming World, Climate Scientists Consider Category 6 Hurricanes

Hurricane Patricia Bears Down on Mexico's Pacific Coast on October 23, 2015. Image Credit: NASA Goddard Spaceflight Center

50 Years of NERSC Firsts

The Cray-1 sold for about $8 million in 1978 (around $38 million in inflation-adjusted dollars).

Perlmutter Provides Peek into Interior of Ice Giant Planets

This photograph of Neptune, one of the ice giant planets addressed in a paper recently published by researchers at UC Berkeley, was reconstructed from two images taken by NASA Voyager 2. Credit: NASA

New Tomographic Reconstruction Algorithm Developed at Berkeley Lab Sets World Record

Reconstructions of a material sample from a sparse set of projection data using TomoCAM, conventional MBIR, and direct approximation methods. TomoCAM delivers higher-quality reconstructions compared to direct approximation methods while being 15 times faster than conventional MBIR methods.
Computing Sciences logo

Connect

Our Organization