How to learn a "complete order function" from a set of partial order relations?

Question

Input: vector data x_1,...,x_n and a partial ranking function e..g r_p(x_1) > r_p(x_2).

Output: a "complete order function" r_c that projects the data onto 1D (e.g. r_c(x_1)=0.5, r_c(x_2)=0.3)?

Question: what's the standard model/algorithm to use? (e.g. parametric family of functions)

score 1 · Answer 1 · answered Nov 18 '15 at 15:49

Draw the partial ranking data as a directed graph, and then observe if the resulting graph really is linear? If it is linear, you can read the complete order off the graph, if it is not linear, such complete order (compatible with your ranking data) do not exist.

How to learn a "complete order function" from a set of partial order relations?

1 Answers1