Hands-On Neuroevolution with Python

NoveltyArchive

This class maintains a list of relevant novelty items and provides methods to evaluate the novelty scores of individual genomes as well as of the entire population of genomes as a whole. It has the following fields defined in the constructor:

    def __init__(self, threshold, metric):        self.novelty_metric = metric        self.novelty_threshold = threshold        self.novelty_floor = 0.25        self.items_added_in_generation = 0        self.time_out = 0        self.neighbors = KNNNoveltyScore        self.generation = 0        self.novel_items = []        self.fittest_items = []

Note that novelty_metric is a reference to the function that can be used to estimate the novelty metric or distance between two novelty items.

Furthermore, novelty_threshold defines the current minimal novelty ...