Chapter 12: PyTorch and AutoML
Automated machine learning (AutoML) provides methods to find the optimal neural architecture and the best hyperparameter settings for a given neural network. We have already covered neural architecture search in detail while discussing the RandWireNN model in Chapter 5, Hybrid Advanced Models.
In this chapter, we will look more broadly at the AutoML tool for PyTorch—Auto-PyTorch—which performs both neural architecture search and hyperparameter search. We will also look at another AutoML tool called Optuna that performs hyperparameter search for a PyTorch model.
At the end of this chapter, non-experts will be able to design machine learning models with little domain experience, and experts will drastically speed up their model selection process.
This chapter is broken down into the following topics:
- Finding the best neural architectures with AutoML
- Using Optuna for hyperparameter search
