Engineering AI Systems: Architecture and DevOps Essentials
by Len Bass, Qinghua Lu, Ingo Weber, Liming Zhu
8
Performance
You can’t manage what you don’t measure.
—Peter Drucker
IN SOFTWARE ENGINEERING, performance is typically taken to mean computation efficiency, which includes speed and scalability. How fast do computations execute? How much of the requisite resources, such as CPUs and energy, do they need? In AI, performance is typically taken to mean “how accurate are the model outputs?” In many ways, these two interpretations of performance are contradictory. Achieving accuracy frequently requires more computation. Achieving higher efficiency, but operating with the same resources, typically reduces accuracy. Which is more important is a decision that drives the tradeoffs that the designer must make. In this chapter, we discuss these two interpretations ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access