In this chapter, we explored the complete life cycle of a neural network in Pytorch, starting from constituting different types of layers, adding activations, calculating cross-entropy loss, and finally optimizing network performance (that is, minimizing loss), by adjusting the weights of layers using the SGD optimizer.

We have studied how to apply the popular ResNET architecture to binary or multi-class classification problems.

While doing this, we have tried to solve the real-world image classification problem of classifying a cat image as a cat and a dog image as a dog. This knowledge can be applied to classify different categories/classes of entities, such as classifying species of fish, identifying different kinds of dogs, categorizing plant seedlings, grouping together cervical cancer into Type 1, Type 2, and Type 3, and much more.

In the next chapter, we will go through the fundamentals of machine learning.