Jad El Karchi (Storm)

Hi all,

All good things must come to an end, next week is Jad's last week. Join us on April 12 (sorry it is a Wednesday..) to learn everything about his internship.

Speaker

Jad El Karchi

When

Wednesday, April 12, 2023 at 1:30 pm.

Where

room George Boole 2 or online

Title

Detecting errors in MPI applications with a deep learning approach

Abstract

Message Passing Interface (MPI ) is a broadly utilized parallel programming show that permits different forms to communicate and facilitate with each other in parallel computing situations. As MPI applications develop in complexity and scale, it gets to be progressively troublesome to identify mistakes and bugs, which can lead to execution errors and bugs. Conventional approaches for identifying mistakes in MPI applications depend on manual assessment and investigation, which can be time-consuming and error-prone. In recent years, machine learning techniques have risen as a promising approach for recognizing blunders in complex frameworks. Machine learning models can learn from huge sums of information to distinguish designs and peculiarities which will demonstrate blunders and bugs in MPI applications. By leveraging this innovation, designers can progress the exactness and effectiveness of bug detection and diminish the time and exertion required for manual investigating.

This research explores the use of machine learning to detect errors in MPI applications. It provides an overview of MPI code and challenges related to detecting errors in code, detailing how to detect patterns and anomalies in applications. We also introduce ML models and their possible applications in the context of MPI error detection with different approaches. Finally, we discuss and analyze some of the current research and approaches to detecting errors in MPI applications using current models. This includes data preprocessing, feature selection, label selection, and model selection