Machine Learning Engineering in Action

Appendix A. Big O(no) and how to think about runtime performance

Runtime complexity, for ML use cases, is no different than it is for any other piece of software. The impact of inefficient and poorly optimized code affects processing tasks in ML jobs the same as it does any other engineering project. The only material difference that sets ML tasks apart from traditional software is in the algorithms employed to solve problems. The computational and space complexity of these algorithms is typically obscured by high-level APIs that encapsulate recursive iterations, which can dramatically increase runtimes.

The goal of this appendix is to focus on understanding both the runtime characteristics of control code (all the code in your project that isn’t ...