Summary

This chapter described forward propagation in a neural network. The neural network we explained in this chapter is the same as a perceptron in the previous chapter in that the signals of the neurons are transmitted hierarchically. However, a large difference exists in the activation functions that change signals when they are transmitted to the next neurons. As an activation function, a neural network uses a sigmoid function, which changes signals smoothly, and a perceptron uses a step function, which changes signals sharply. This difference is important in neural network training and will be described in the next chapter. This chapter covered the following points:

• A neural network uses a function that changes smoothly, such as a sigmoid function or a ReLU function, as an activation function.
• By using NumPy's multidimensional arrays, you can implement a neural network efficiently.
• Machine learning problems can be broadly divided into classification...