[LA-SiGMA] HPC Training: Distributed Workload

[Feng Chen]

Dear All,

The Schedule for the Fall 2016 HPC Training is available at http://www.hpc.lsu.edu/training/tutorials.php.

Our next HPC training will be held on Wednesday, November 2 at 9:00 AM in 307 Frey Computing Service Center and broadcast through WebEx for remote users. All HPC trainings will start at 9:00AM.

The training sessions are subject to cancellation due to lack of registrations, so please register at http://www.hpc.lsu.edu/training/tutorials.php as early as possible if you plan on attending. Registration closes in the afternoon at 4:30pm on the day prior to the training.

Wednesday, November 2,2016: Distributed Workload
In many scientific disciplines such as bioinformatics and computational biology, researchers often need to process workloads composed of a large number of loosely-coupled or independent tasks, using tools that are serial in nature. While the execution of each of these tasks may be simple and straightforward, processing the entire workload efficiently on an HPC cluster can be challenging. Distributed Workload is running multiple tasks independently with no information exchanging among them also known as "embarrassingly parallel". In fact, those tasks are extremely parallelizable using this embarrassingly parallel concept. This training focuses on how to parallelize the processing of such an "embarrassingly parallel" workload, i.e. how to run multiple independent tasks in parallel using one job submission script.
The presentation starts with a review of some shell job control features, followed by the discussion of several approaches of how to submit embarrassingly parallel jobs using WQ, GNU Parallel and Swift. Examples will include serial, multi-threaded, and small MPI tasks.
Prerequisites: Beginner level knowledge of shell scripting and PBS job submission is assumed, but not required.

Next training:

Wednesday, November 9,2016: Introduction to Machine Learning
Machine learning, which learns from and generates predictions on data, has become one of the most important technology impacting every application domain including e-commerce, medical informatics, science and engineering. As data get massive and more complex, a deep learning framework emerges as a powerful tool to learn robust and complex features directly from data without domain-specific knowledge. Deep neural networks have shown great success in various fields such as computer vision, natural language processing, and speech recognition, and been widely used in big tech companies such as Google, Microsoft and Facebook. Existing tools such as Keras, Tensorflow and Theanos are being developed to build and evaluate deep learning models. This training demonstrates how to build your first deep learning model for image classification tasks using existing python library Keras. We invite Dr. Mingxuan Sun, Assistant Professor in Computer Science, to present insights on current work of applying deep learning to recommender systems.

Please visit http://www.hpc.lsu.edu/training/tutorials.php for more details and register using the link provided.
Users who plan on joining remotely will be provided with a WebEx Link in their registration confirmation email. Please see the system requirements at https://grok.lsu.edu/Categories.aspx?parentCategoryId=3381.

Please distribute to faculty, staff and students that might be interested.

Thanks,

Feng Chen, PhD
IT Consultant
High Performance Computing
Louisiana State University
329 Frey Computing Services Center, Baton Rouge, LA  70803
office 225-578-2924
fchen14 at lsu.edu<mailto:fchen14 at lsu.edu>


-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mail.loni.org/mailman/private/lasigma/attachments/20161031/ef6aeb0c/attachment.html