Self Supervised Learning
Published:
What is Self Supervised Learning ?
Self Supervised Learning(SSL) is a method to train neural networks in which we don’t require any extra labels, we take labels from the input data itself. It’s not a very new concept, we are using SSL in NLP from quite some time to train language models, in which we train our model to predict next word of a sentence given a word. It’s proved to give accuracies a large boost in NLP tasks beacuse it’s observed that a model which learns the nature or the way to generate nature language, can be fine tuned for any task.