Takes a genome and a map of transformation/probability pairs. Picks a transformation probabilistically, and then applies it to the genome by silencing/unsilencing/no-oping random genes that aren’t already of that type.

Auto-simplifies the provided individual.

Automatically simplifies the genome of an individual without changing its error vector on the training set, based on the error-function. steps is the number of hill-climbing evaluations to test. print-progress-interval is how often to print progress of the simplification; if it is set to 0, then nothing will be printed. simplification-step-probabilities is a map of probabilities that are used to select what change to make during each step of the simplification. Each change is represented as a map with the following options for the keys, each of which has an integer of how many of those changes to make: :silence - number of unsilenced or no-op genes to set :silent = true :unsilence - number of silenced or no-op genes to set :silent = false :no-op - number of unsilenced or silenced genes to set :silent = :no-op

Changes the value of the :silent tag on the gene at each index of indices into new-value.

Selects k random elements of input list, without replacement. If fewer than k exist, selects all.

Removes one random pair of parens from a program. Cannot remove outermost pair.

Given map of item/probability key/values, picks an item probabilistically based on the probabilities.