Tokyo, Japan – November 21th, 2023– Morpho, Inc. (hereinafter, “Morpho”), a global leader in image processing and imaging AI solutions, announced today that its deep learning inference engine “SoftNeuro®” has been featured in a paper on “3D-Spatiotemporal forecasting the expansion of supernova shells using deep learning towards high-resolution galaxy simulations” promoted by The University of Tokyo, Tohoku University, and Kobe University, where it contributed to accelerating inference.
In this research, Morpho’s “SoftNeuro” was provided in enabling faster inference using newly developed deep learning to eliminate the bottleneck in the calculation of supernova explosions in large-scale galaxy formation simulations. It contributed to an approx.19.2-times faster inference time used in simulations (*1) and a 93% reduction in power consumption (*2).
The new deep learning model developed through the research will be incorporated into the galaxy formation simulation code “ASURA-FDPS”, and “SoftNeuro” will continue to be used in studies of galaxy formation.
This new approach will make it possible to efficiently simulate the detailed motion of individual stars in relatively large galaxies, such as the Milky Way.
Morpho will contribute to the realization of diversified ecosystems and energy saving through the expected new synergy between the IT and AI industries and astronomical research as well as the high convenience and versatility of “SoftNeuro”.
Public repository of deep learning model 3D-MIM
https://github.com/kyaFUK/3D-MIM
3D-Spatiotemporal forecasting the expansion of supernova shells using deep learning towards high-resolution galaxy simulations
https://academic.oup.com/mnras/article-lookup/doi/10.1093/mnras/stad2864
26/10/2023 Press Release (The University of Tokyo’s website)
Deep learning speeds up galactic calculations
“SoftNeuro” supports major deep learning frameworks and performs faster processing in various edge-device environments. Since it is a general-purpose inference engine, it can be used not only for image recognition but also for speech recognition and text analysis. Morpho has proposed and provided “SoftNeuro” for multi-platform and high-speed inference for various detection applications based on image data.
– Product page: https://www.morphoinc.com/en/technology/sie
12/05/2022
Morpho Provides “SoftNeuro” to Highly Resolved Galaxy Simulations Project for Universities: Supporting 3D Simulation with Supercomputer Fugaku
https://www.morphoinc.com/en/news/20221205-epr-sn
02/08/2023
Morpho’s “SoftNeuro” Enables 19x Faster Inference of 3D Simulation on Fugaku
https://www.morphoinc.com/en/news/20230208-epr-sn
*1, 2: Executed one at a time (each execution calls the inference function only once.) five times. The average of the measured values across five executions is the result. For details, please refer to the press release issued on February 8, 2023.https://www.morphoinc.com/en/news/20230208-epr-sn.
Copyright© Morpho, Inc. All rights reserved.
