In this chapter, we introduced the perceptron concept and how to solve a linear problem with a perceptron. We discovered different ways to update the parameters of our model, and we saw how to implement a perceptron from scratch and using Keras. We then introduced neural networks as a set of connected neurons that are able, theoretically, to approximate any function. We saw a few different activation functions and we discussed their advantages and disadvantages. We saw how to create a simple network from scratch, and how to do it using Keras.
In the next chapter, we will explain how to solve an image classification problem, using the concepts we just introduced.
