SimCLR paper overview

This is video no. 3 in a series walking through simCLR research paper, going over the key ideas of simCLR:
* Composition of image transformation
* Nonlinear projection head
* Temperature and normalization
* Large batch sizes

William Falcon, PyTorch Lightning founder, and Ananya Harsh Ja, Lightning research engineer, deep dive into simCLR (“A Simple Framework for Contrastive Learning of Visual Representations”), self-supervised representation learning on images. Start here: https://www.youtube.com/watch?v=pDJx8i3jenA&list=PLaMu-SDt_RB4k8VXiB3hOdsn0Y3GoXo1k
In the next videos, we will go over key ideas of the paper, and how they are implemented with PyTorch Lightning.

Next up: Training without fine-tuning- https://youtu.be/a7-qwwAFs_s

Paper: https://arxiv.org/abs/2002.05709
PyTorch Lightning implementation: https://github.com/PyTorchLightning/pytorch-lightning-bolts/tree/master/pl_bolts/models/self_supervised/simclr
Colab: https://colab.research.google.com/drive/1UK8BD3xvTpuSj75blz8hThfXksh19YbA?usp=sharing

Thanks for watching! Please Subscribe!

Source of this PyTorch Lightning Video

AI video(s) you might be interested in …