This chapter introduced and described the concept of artificial neural networks. We first discussed ANNs from a conceptual standpoint. You learned that neural networks are made of individual neurons, which are simple weighted adding machines that can apply an activation function to their output. You learned that neural networks can have many topologies, and it is the topology and the weights and biases between neurons in the network that do the actual work. You also learned about the backpropagation algorithm, which is the method by which neural networks are automatically trained.
We also looked at the classic XOR problem and looked at it through the lens of neural networks. We discussed the challenges and the approach to solving XOR with ANNs, and we even built—by hand!—a fully-trained ANN that solves the XOR problem. We then introduced the TensorFlow.js...