In this chapter, we've reviewed topic modeling techniques, including linear and non-linear learning methods. We explained how the NTM from SageMaker works by discussing its architecture and inner workings. We also looked at distributed training of the NTM model, where the dataset is divided into chunks for parallel training. Finally, we deployed a trained NTM model as an endpoint and ran the inference, interpreting topics from Enron emails. It is essential to synthesize information and themes from large volumes of unstructured data for any data scientist. NTM from SageMaker provides a flexible approach to doing this.
In the next chapter, we will cover the classification of images using SageMaker.