Structuring Data without Neural NetworksΒΆ

Deep learning with neural networks is very much at the forefront of the recent renaissance in machine learning. However, machine learning is not synonymous with neural networks. There is a wealth of machine learning approaches without neural networks, and the boundary between them and conventional statistical analysis is not always sharp.

It is a common misconception that neural network techniques would always outperform these approaches. In fact, in some cases, a simple linear method could achieve faster and better results. Even when we might eventually want to use a deep network, simpler approaches may help to understand the problem we are facing and the specificity of the data so as to better formulate our machine learning strategy. In this chapter, we shall explore machine learning approaches without the use of neural networks. This will further allow us to introduce basic concepts and the general form of a machine learning workflow.