How can I perform this complex action in pinecone?

Support
1

Total newbie here.
My question is: is there a way that allows me to get the center-most vector(each center-most vector for each group of vectors) from all the groups of vectors.

eg. vector A, vector B and vector C lie close together. vector B lies in the center of this group of vector. So, vector B would be returned.

however if there are other groups of vectors, the centermost vector of all the groups should be returned. the groups may be grouped according to provided radius.

Hope my question is clear.

2

Hi @theDude!

There is no way to get what you are looking for :slight_smile: But it depends why you are needing this in the first place. Maybe you can make a way around and only store these “centroids” into Pinecone and have all the other vectors stored elsewhere :slight_smile: or you could have them all stored inside Pinecone, but separated by “namespace”.

Hope this helps

3

turns out what i’m looking for is a popular algorithm called k-means clustering (but with radius instead of k).
is there a way to perform regular k-means clustering in the vectors stored in pinecone?

4

Not by default, as far as I know. You could maybe use Pinecone in your own implementation of k-means, but other than that… I don’t think this will help you much, sorry :sweat_smile:

5

Thanks for the quick reply. : )

1 Like