Python, Brain Circuits, Wildfires & More

[ad_1]

On this bimonthly function, HPCwire highlights newly revealed analysis within the high-performance computing neighborhood and associated domains. From parallel programming to exascale to quantum computing, the small print are right here.


Charting the tendencies of machine studying in Python

“Deep neural networks, together with developments in classical [machine learning] and scalable general-purpose GPU computing, have develop into vital elements of synthetic intelligence,” these authors – a staff from the College of Wisconsin-Madison, the College of Maryland, and Nvidia – write. On this paper, they survey the sphere of machine studying with Python, figuring out core {hardware} and software program paradigms that enabled it to be the popular language for a lot of this exercise.

Authors: Sebastian Raschka, Joshua Patterson and Corey Nolet.

Figuring out wildfires with HPC and deep studying

As local weather change looms, harmful wildfires have develop into sadly prevalent in lots of elements of the world. These researchers, a duo from the P.G. Demidov Yaroslavl State College in Russia, used an Nvidia DGX-1 system to coach a neural community to determine wildfires  in a sequence of almost two thousand satellite tv for pc pictures. The researchers report that the ensuing algorithm is appropriate for early wildfire detection in the actual world.

Authors: Vladimir Khryashchev and Roman Larionov.

Simulating large-scale fashions of mind circuits with Google Cloud Platform

Increasingly, researchers are turning to cloud-based HPC to satisfy their intensive computing wants. This staff of researchers from a number of universities, two hospitals and Google developed an in depth mannequin of the mind motor cortex circuits, together with greater than 10,000 neurons and 30 million connections, utilizing Google Cloud Platform. Every simulated second, the authors say, required 50 core hours.

Authors: Subhashini Sivagnanam, Wyatt Gorman, Donald Doherty, Samuel A. Neymotin, Stephen Fang, Hermine Hovhannisyan, William W. Lytton and Salvador Dura-Bernal.

Benchmarking Microsoft Azure digital machines for HPC purposes

With cloud HPC on the rise, researchers are turning their consideration to the relative efficiency of various cloud choices for HPC purposes. On this paper, written by a staff from the College of Jordan, the authors conduct a efficiency evaluation of Microsoft Azure digital machines utilizing NASA’s NAS parallel benchmarks for HPC. 

Authors: Rawan Aljamal, Ali El-Mousa and Fahed Jubair.

Simulating 100 million atoms on Summit

These authors, a staff from 4 universities in China and the U.S., current a GPU adaptation of the DeePMD instrument, which makes use of a deep neural community to drive extraordinarily giant molecule dynamics simulations. After testing the brand new model of the instrument on Summit – the world’s quickest publicly ranked supercomputer – the authors concluded that “the GPU model is seven instances sooner than the CPU model with the identical energy consumption.”

Authors: Denghui Lu, Han Wang, Mohan Chen, Jiduan Liu, Lin Lin, Roberto Automobile, Weinan E, Weile Jia and Linfeng Zhang.

Scaling neural networks for geophysics on supercomputers

Numerical simulation is commonly computationally costly, main geophysics researchers to seek for surrogate modeling strategies by means of deep studying. On this paper, researchers from Argonne Nationwide Laboratory focus on the event of a scalable neural structure for a temperature forecasting mannequin. The mannequin scaled up utilizing a number of totally different strategies on over 500 nodes of Argonne’s Theta supercomputer.

Authors: Romain Egele, Bethany Lusch and Prasanna Balaprakash.

Creating an app for smoothed particle hydrodynamics at exascale

Smoothed particle hydrodynamics (SPH) are utilized in numerical simulations of fluids for astrophysics, computational fluid dynamics and lots of different fields. This paper, written by a staff from Switzerland, presents the SPH-EXA undertaking, which goals to develop an exascale-ready SPH app. As a place to begin, the authors study three totally different SPH codes and work to consolidate them right into a well-optimized mini-app for exascale.

Authors: Danilo Guerrera, Rubén M. Cabezón, Jean-Guillaume Piccinali, Aurélien Cavelan, Florina M. Ciorba, David Imbert, Lucio Mayer and Darren Reed.


Are you aware about analysis that needs to be included in subsequent month’s checklist? In that case, ship us an e-mail at [email protected]. We sit up for listening to from you.

[ad_2]
Source link

Total
0
Shares
Leave a Reply

Your email address will not be published.

Previous Post

3 Reasons SEO is a Crucial Cost-Effective Medium for Your Struggling Business

Next Post

Voy Media, A Preeminent Marketing Agency, Announces Launch Of New Google Ads Service For E-Commerce Brands – Press Release

Related Posts