Deterministic taxonomy based on genetics

By Robin Dowling · a year ago

Each simulated evolution begins with a random initial seed. I wanted to avoid any preconceived model of the world and its evolutionary processes. For this approach to succeed, the taxonomy classification needed to be a deterministic generalization of an organism's genetics across all taxonomic levels, from Domain to Species.

I began by constructing a new Taxonomy class that accepts an array of numbers, with each value representing a component of an organism's genetic makeup. This genetic array is then transformed and reduced at varying levels of precision corresponding to each taxonomic level.

Through extensive experimentation with precision settings and transformation methods, I tried to find the ideal balance, a sweet spot that would yield a reasonable distribution of species based on the diversity of randomly assigned genetics across all organisms.

This reduction process organized organisms not only into species groups but also into progressively fewer clusters of genus, family, order, class, phylum, kingdom, and domain, each level encompassing larger populations within its hierarchy. Importantly, all groupings remained deterministically derived from the randomly assigned genetic foundations.

Finding an equilibrium that effectively sorted the vast array of organisms with their uniquely individual genetic profiles into coherent groups turned out to be more challenging than expected.

I was committed to avoiding manual classifications or special cases. The solution needed to be universally applicable across all possible initial random seeds, relying only on the input genetics and the varying precision thresholds for each taxonomic level.

Ultimately, the solution emerged with elegant simplicity: just a few lines of mathematical expressions and carefully balanced weights.

A single Taxonomy algorithm capable of sorting and classifying any collection of organisms with randomly assigned compositions of genetics.