Storing and comparing multidimensional vectors

Question

I have multidimensional vectors with n-degrees. What is the best way to handle them.

• Storage (what is the best mode to store them, SQL, Mongo, Document?)
• Comparison (How do I compare two of them and find similarity).

Example:

a{1,3,4,1,-1},
b{2,3,3,0,0},
c{2,3,2,8,9}

If we compare the above two, a and b are more similar.

How do do this in large scale? I want to store thousands of such vectors.

Similarily can be the vector distance.

Answer 1

My bet for storage would be on a Graph database, like Neo4j. You can calculate the Euclidean distance (and probably implement k-means clustering to get closest points) using Cypher query.

For details on implementation, check out the below articles --

• http://heidi.morkland.org/tag/neo4j/
• http://mikelam.azurewebsites.net/beer-recommendations-with-user-based-collaborative-filtering/