To understand how the brain works, you need a detailed map of neuron connections — the so-called connectome. The traditional approach is like trying to figure out a huge garden irrigation system by opening each tap one by one and watching where the water flows. In experiments, this means stimulating neurons with light one at a time: it's laborious, requires millions of attempts, and risks damaging living tissue due to prolonged exposure.

The OPhELIA algorithm does it differently, like an experienced gardener who, by turning on water at just a few strategically chosen points, mentally reconstructs the entire irrigation layout from the resulting streams and splashes. Its "intuition" is based on the concept of entropy — a measure of uncertainty or lack of information about a system. The algorithm deliberately selects neurons for stimulation that, according to calculations, will yield the most new information about the connection structure, progressively reducing the entropy of the connectome model. It even accounts for spontaneous neuron activity, the equivalent of random "drips" in the garden analogy, extracting additional data from it.

This approach allowed researchers, using zebrafish, to reconstruct nearly the full map of their connections by stimulating only about 5% of neurons — roughly 20 times faster than traditional brute-force methods. This resembles how astronomers, without directly observing dark matter, infer its distribution from the motion of visible stars and galaxies, and provides a kind of Standard Model — a reliable, reproducible foundation for further research into neural networks.