Webinar | Self Supervised learning : Approaches and Applications

Organized by:
SDAIA-KFUPM Joint Research Center for AI

Abstract:
Developing intelligent systems using machine learning approaches has relied heavily on high-quality labeled data. The unavailability of labeled data or complexities involved in the annotation process opens the door to proposing methods that do not require labeled data in the modeling process. Unsupervised learning techniques have been extensively developed to learn patterns from unlabeled data and have been employed in many learning applications. Unsupervised learning aims to focus on clustering approaches; however, Self-supervised learning (SSL) is a form of unsupervised technique where data provides supervision. SSL is an evolving technique and has the potential to address the issues brought about by the high dependence on annotated data. This talk presents the motivation behind self-supervised learning, some state-of-the-art techniques, and their potential benefits in challenging learning applications. The techniques covered in this talk will be based on pre-text tasks, invariance and contrast, and generation-based approaches. The applications will be presented from images, videos, and NLP. ​

Speaker:
Dr. Muhammad Azam
Dr. Muhammad Azam received his Ph.D. degree in electrical and computer engineering with a research focus on artificial intelligence from Concordia University. .He worked as Principal Data Scientist in the industry before his postdoc at Concordia University Montreal, Canada. He has authored or co-authored about 40 publications in reputed journals, conferences, and contributions to peer-reviewed books. His research interest includes machine learning, statistical modeling, speech recognition, computer vision, information retrieval, signal processing, energy efficiency, and time-series analysis

For more information about this event or the upcoming, visit our website:
Or follow our social media accounts: