Mastering Self-Supervised Algorithms for Learning without Labels

An understanding of self-supervised learning and its advantage over unsupervised learning

Working knowledge of designing your self-supervised learning tasks/objectives

Hands-on experience implementing and modifying existing self-supervised learning objectives to learn from unlabelled data

Ability to transfer and evaluate your self-supervised network representations on a downstream task

Familiarity with core components of self-supervised learning, including pretext tasks, similarity maximization, redundancy reduction, and masked image modeling