Quantify the amount of differences of one data set to another

Basically I'm seeking the best method for my case scenario here. For example with the image shown below, the left one is poincare maps dataset #1, and the right is dataset #2.

What's the best method of quantifying the differences of the right one to the left? Standard deviation won't be of much help here, since these are two separate data sets. Appreciate any inputs, thanks in advance!

1 Answer

Just average (or pool, moree generally aggregate) some of the standard measures:

if you look closly python implementation is already general for some of them, meaning that you can apply them directly on dataset and not individually on the vectors

Examples: Cosine similarity, Eucledian distance, Manhattan/city-block distance, Chebychev distance, Minkowsky dustance. For computer vision, similarity measures are: Hausdorff Distance (the two data vectors will be considered close if every point of each set is close to some point of the other set), Bhattacharyya Distance (measure of similarity of two probabilty distributions), Bhattacharyya coefficient (measure of relative closeness of the two vectors) and Mahalanobis Distance (a specific case of Bhattacharyya Distance).

• That's great head start for me, thank you Noah! Mar 6 '20 at 8:00