While going through Part 2, I was specially intrigued by Label smoothing because it goes against the conventional practice of maximising the likelihood of ground truth labels. So, I dug a bit into it’s working and wrote a blog post about what I understood. Hope you find it useful. Please provide feedback since I’m a newbie in blogging and have a lot to learn.
What does Label smoothing do?- https://abhimanyu08.github.io/blog/deep-learning/2020/05/17/final.html