Contrastive Learning Insights
Simon discusses the importance of diverse features in contrastive learning and the challenges of applying these techniques beyond natural image datasets. He highlights the need to consider specific tasks when designing augmentation policies for non-traditional datasets like medical or satellite images.In this clip
From this podcast
Machine Learning Street Talk (MLST)
#032- Simon Kornblith / GoogleAI - SimCLR and Paper Haul!
Related Questions