Introducing Machine Learning

Chapter 18. Other Types of Neural Networks

Either mathematics is too big for the human mind or the human mind is more than a machine.

—Kurt Godel, author of the Incompleteness Theorem

In the preceding two chapters, we presented neural networks as an extremely sophisticated and accurate tool to use for classification and regression tasks when everything else fails. Unfortunately, there are some problems that the same neural networks can’t solve well with the required level of precision. Under which conditions can a feed-forward neural network underperform? And where else can you turn when a canonical feed-forward neural network doesn’t work as expected?

These are the questions that we’ll try to answer in this chapter, looking at newer and even ...