When:  9am PT

How:  Zoom at https://ucsc.zoom.us/j/272379932

Title: SimCLR is a simple framework for contrastive learning of visual representations

Abstract: SimCLR is a simple framework for contrastive learning of visual representations. It simplifies recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels.

Bio: Ting Chen is a research scientist in the Google Brain Team. His main research interest is on representation learning, discrete structures, and generative modeling. He joined Google in 2019 after finishing his PhD at University of California, Los Angeles.