Summary

In this chapter, we covered important AI/ML concepts that are typically used by more advanced practitioners who have specific needs or preferences in terms of how they want to implement their ML workloads, such as by using specific frameworks, or by parallelizing their workloads across multiple processors.

We started by discussing BQML, focusing on the close relationship between data processing, data analytics, and ML. We learned how to train and evaluate a model using BQML, and how to get predictions from that model, all by using SQL query syntax.

Then, we discussed different types of hardware that we can use for our AI/ML workloads, and popular tools and frameworks that we had not previously covered in this book, such as Spark MLlib, Ray, and PyTorch.

Next, we dived into CNNs and their use in computer vision before moving on to discuss neural network architectures that are particularly useful for sequential data and use cases such as NLP, such as RNNs, LSTMs, and...