论文部分内容阅读
Recent progress in deep learning is essentially based on a“big data for small tasks”paradigm,under which massive amounts of data are used to train a classifier for a single narrow task.In this paper,we call for a shift that flips this paradigm upside down