Do You Pick the Right Device for Data Science?
Abstract
The increasing application of Machine Learning (ML) techniques in various areas has led designers to leverage ML accelerators like GPGPUs, TPUs and many more. However, choosing the most appropriate accelerator for such algorithm is very challenging as they commonly should adhere to tight constraints e.g., low power consumption, performance, and low cost. As a consequence, designing such application-specific devices becomes a non-trivial and difficult task.
In this talk I give an overview about which ML accelerators are available an the market and which is best for the different ML algorithm.
About the Speaker
Christopher Metz acquired a Bachelor degree in Business Informatics (2015) followed by a Master degree in Informatics – Distributed and Mobile Applications (2017) at Osnabrück University of Applied Science. After his studies, he worked as software developer at CKS Systeme GmbH, where one of his main tasks was to implement interfaces for e.g. medical equipment, web services, police offices or fire departments. Moreover, he was responsible for the development of products for the police and fire department. From October 2018 until April 2020, Christopher Metz worked as research associate in the project Environment, Health and Safety based on Artificial Intelligence at Osnabrück University of Applied Science. He was responsible for the establishment of a GPU cluster for deep learning models as well as the development of AI models for NLP (Natural Language Processing).
Christopher Metz contributes to the DSC with his extensive knowledge in developing and implementing data interfaces and neuronal networks, in particular deep learning. Additionally, he is responsible for the establishment of the DSC infrastructure as well as for the collection, storage, archiving, and processing of data. Moreover, Christopher Metz supports scientist from various disciplines with the development, programming, and implementation of data science technologies. His main research interests lie in AI-methods and infrastructure for high-performance computing.
Zoom-Link zur Veranstaltung.
0 Kommentare