TLDR
Around 80% of modern word processing (NLP) consists of self-supervised learning. Self-Supervised machine learning is a way to teach a model a lot without manual markup, as well as an opportunity to avoid deep learning when setting a model up to solve a problem. In all experiments, almost simple models are trained on simple feature received on the downstream feature learned in self.Downstream task evaluates the quality of features learned by self. In this case, the task is a task (pseudo-labels), on which the model is trained to learn to form good representations of objects.via the TL;DR App
no story
