Calculates the novelty of the individual based on the behaviors in the population and in the novelty-archive. Returns the individual with the :novelty key set, and if :novelty is a meta-error-category, also sets that.

Takes two behavior vectors and finds the distance between them. Differences in vectors are based on the data type(s) they contain. Distance metric is based on the arg :novelty-distance-metric. Note that there is no limit on behavior differences, which will only be limited by their max bounds based on things like maximum integer size.

Calculates a map storing the distances between any two behaviors, of the form: {behavior1 {behavior1 dist11 behavior2 dist12 behavior3 dist13 …} behavior2 {behavior1 dist21 behavior22 dist2 behavior3 dist23 …} …} Note: Only has outer-level keys for population behaviors, not archive behaviors. But, has inner-level keys for both population and archive behaviors.

Calculates the sparseness/novelty of an individual by averaging together the distances between it and its k nearest neighbors. First, it must look up those distances using the behavior-distance-map.

Calculates novelty for each individual in the population with respect to the rest of the population and the novelty-archive. Sets novelty to meta-error if necessary.

Returns an individual that does the best out of a tournament based on novelty.

Returns a number of individuals to be added to the novelty archive. Number of indviduals are :individuals-for-novelty-archive-per-generation.