Chapter 9. A peek at performance

 

This chapter covers

  • Setting up DL models in production
  • Optimizing performance and deployment
  • Getting real-life neural search systems to work with data streams

After reading the previous eight chapters, you hopefully have gained a broad understanding of deep learning and how it can improve search. At this point, you should be ready to make the most out of DL when setting up successful search engine systems for your users. Along the way, however, you may have wondered about applying these ideas to real-world production systems:

  • How are these approaches applied in practice in a production scenario?
  • Will adding these DL algorithms have a serious impact on the time and space constraints of your systems?
  • How big is that impact, and which parts or processes (such as searching versus indexing) will be affected?

In this chapter, I’ll address these practical concerns and discuss the considerations you’ll need to think about as you apply DL and neural networks to your search engine. We’ll look at the performance bits when search engines and neural networks work side by side, and I’ll provide some example-driven suggestions for applying these DL techniques in practice.

9.1. Performance and the promises of deep learning

9.2. Indexes and neurons working together

9.3. Working with streams of data

Summary

Looking forward