I was learning about GAN when the term "Label Smoothing" appears. From the video tutorial that I watch, they use the term "label smoothing" to change the binary labels when calculating the loss of discriminator network. Instead of using 0 or 1 label they use 0 or 0.9 labels. What is the main purpose of this label smoothing?
I've skimmed through the original paper, there is a lot of maths, that honestly, I got difficulties in understanding it. But I notice this paragraph in there:
We propose a mechanism for encouraging the model to be less confident. While this may not be desired if the goal is to maximize the log-likelihood of training labels, it does regularize the model and makes it more adaptable
And it gives me another question:
why "this may not be desired if the goal is to maximize the log-likelihood of training labels"?
what do they mean with "adaptable"?