EvolutionIslandSummary

Models & Types · AiDotNet.Models.Results

How one island of an evolution run ended: its coverage, its best entry, and how far it advanced.

For Beginners

Splitting the search into islands is like running several small competitions instead of one big one, letting each explore its own direction before the best members are exchanged. This tells you how each competition went: how many distinct good answers it collected, which was its best, and how many rounds it managed.

How It Works

An island is one independent sub-population with its own archive. Comparing these summaries is how you tell a healthy island model from a degenerate one: islands that all report the same BestGenomeId have converged on a single idea, and an island whose EliteCount stays near zero while others fill up is being starved of proposals.

Constructors

Constructor Summary
EvolutionIslandSummary()

Properties

Property Summary
double Coverage { get; set; } Gets or sets the fraction of the grid this island filled, from 0 to 1.
double? BestQuality { get; set; } Gets or sets this island’s best quality, or null when it archived nothing scored.
double? MeanQuality { get; set; } Gets or sets the mean quality across this island’s entries, or null when there are none.
int EliteCount { get; set; } Gets or sets how many archive cells this island filled.
int HistoryCount { get; set; } Gets or sets how many past evaluations this island retained beyond its elites.
int Island { get; set; } Gets or sets the island index, counted from zero.
long Generation { get; set; } Gets or sets how many variation proposals this island produced.
long TotalCells { get; set; } Gets or sets how many cells the archive grid has in total.
string BestGenomeId { get; set; } Gets or sets the identifier of this island’s best entry, or null when it archived nothing.