I am trying to implement Novelty search; I understand why it can work better than the standard Genetic Algorithm based solution which just rewards according to the objective. I am working on a problem which requires to generate a fixed number of points in a 2d box centered at the origin. In this problem, how can I identify which is a novel configuration of points?
Note: I have thought of one way of doing this: We call the mean of one configuration of points to be the mean of all points in that configuration (let's say this tuple is $(m_x, m_y)$, we store the mean of all configurations generated till now, now for a new configuration it's novelty can be defined as the distance of the mean of this new configuration with $(m_x, m_y)$.
But I think it will not work greatly as some very different configuration of points can also have the same mean.