back
Blog

What does Scalability mean in ML

Share on facebook
Share on twitter
Share on linkedin

Everybody claims to be building “scalable system”. That is a lie.
“Scale” really refers to two things: efficient system as the problem size grows, and efficient as the system size (measured in numbers of cores or compute nodes) grows.
Efficient system is one that wastes very little resource (compute)
You have to inspect the word scalable (whenever someone claims it) very carefully within the context of the a specific application or situation.
Here is a great presentation from @david dunson https://lnkd.in/gF9iMNf on “Scalable Bayesian Inference” presented at Neurips 2018 conference. https://lnkd.in/gjBhZZV
I particularly like his point about bandwagons!
“Bandwagons: most people work on very similar problems, whilecritical open problems remain untouched”
Scalability is one of those bandwagons!!
Share on facebook
Share on twitter
Share on linkedin

Request