In the final chapter we discovered that deep neural networks are sometimes much more durable to coach than shallow neural networks. That’s unfortunate, since we’ve good reason to consider that if we may prepare deep nets they’d be way more powerful than shallow nets. But while the news from the final chapter is discouraging, we won’t let it cease us. In this chapter, we’ll develop techniques which can be utilized to train deep networks, and apply them in practice. We’ll also look at the broader picture, briefly reviewing latest progress on using deep nets for image recognition, speech recognition, and other functions. And we’ll take a short, speculative take a look at what the future could maintain for neural nets, and for synthetic intelligence.
Deep-studying methods are representation-learning strategies with multiple levels of illustration, obtained by composing easy but non-linear modules that every rework the representation at one stage (beginning … Read more