What is non-euclidean data in machine learning?

1.24K views

What is non-euclidean data in machine learning?

In: Mathematics

Anonymous 0 Comments

In non-euclidean data shortest path between two data entries are not necessarily “straight line” – meaning, that their closeness can’t be calculated just by comparing the numerical values this data is encoded in. Say we have three pictures – two with dogs and one with a cat. If we calculate the numerical values of pixels we might get the result that one picture of a dog is more similar to the picture of a cat, than the picture of the other dog – just because the colours are more similar. This example is non-euclidean data because the numerical values of this data does not represent the meaning of the contents.

More detailed discussion can be found here:

https://ai.stackexchange.com/questions/11226/what-is-non-euclidean-data