An Introduction to Contrastive Learning for Visual Representations

preview_player
Показать описание
Recording from Knowledge Sharing Session #4 from the KTH AI Society.

Title: Contrastive Learning for Visual Representations

Topic Description: In Machine Learning, the performance of a model is dependent on the data representation. Since most of the data we use to train the models are unlabelled, it is crucial to find techniques that can leverage the unlabelled datasets. In this domain, representation learning is an important technique. In this session, we will learn more about contrastive learning, a type of representation learning technique where we will understand the process of parametric mapping from raw input data to feature vectors to be able to learn important representations useful for various downstream tasks.

Speaker Information: Kiran is a Ph.D. candidate at the School of EECS, KTH. His research addresses the core challenges of designing new techniques to create and control interactive virtual characters using machine learning, procedural modeling, and perceptual evaluations. To discuss similar interests, you can reach out to him through his contact page.

KTH AI Society
Stockholm, Sweden.

(C) 2021, All rights reserved.
Рекомендации по теме