Heuristics for knee finding in Python

The knee finding problem is everywhere

The theme: How much to spend before diminishing returns

In ML, selecting the number of clusters in an unsupervised model is a classic example

It has also come up in my career in making business decisions, for example figuring out the amount to spend on an ad campaign before diminishing returns kick in

References

Finding a “Kneedle” in a Haystack: Detecting Knee Points in System Behavior https://raghavan.usc.edu/papers/kneedle-simplex11.pdf

Written on th, by Louis Cialdella