The machine learning (ML) algorithms covered in part two work well on a wide variety of important problems, including- on text data, as demonstrated in part three. We have also seen how they can provide critical input to a trading strategy. They have been less successful, however, in solving central problems in AI such as recognizing speech or classifying objects in images. The limitations of traditional algorithms to generalize well on such tasks have contributed to the motivation for developing DL, and the numerous breakthroughs by DL have greatly contributed to a resurgence of interest in AI.
In this section, we outline how DL overcomes many of the limitations of other ML algorithms on AI tasks to clarify the assumptions DL makes about data and its relationship with the outcome. These limitations particularly constrain performance on high-dimensional and...
