EvolutionEliteSummary

Models & Types · AiDotNet.Models.Results

One archived elite, described by identity and coordinates rather than by its contents.

For Beginners

Think of the archive as a shelf with one slot per combination of properties — for instance “short and fast” or “long and accurate”. This is the label on one slot: what scored there, how well it scored, and where in the grid the slot sits. Descriptors gives the measured values by name and Cell gives the bin each value fell into.

How It Works

An evolution archive keeps the best candidate found for each combination of behaviour values. This is the redacted record of one of them: which cell it occupies, how it scored, and which island found it. It carries no genome, because a genome can be arbitrarily large and is often the very thing that should not travel into a log or a saved model file. Program evolution’s own summary keeps a size-capped copy of the source alongside this when the caller asks for it.

Constructors

Constructor Summary
EvolutionEliteSummary()

Properties

Property Summary
IDictionary<string, double> Descriptors { get; set; } Gets or sets the measured behaviour values, keyed by descriptor name.
IDictionary<string, double> Metrics { get; set; } Gets or sets every number the evaluation reported, keyed by metric name.
IList<int> Cell { get; set; } Gets or sets the archive cell coordinates, one bin index per descriptor, in descriptor order.
double? Quality { get; set; } Gets or sets the quality that was recorded, or null when the candidate was never scored.
int Island { get; set; } Gets or sets the island the elite was archived on.
long EvaluationId { get; set; } Gets or sets the sequential evaluation identifier that produced this entry.
string GenomeId { get; set; } Gets or sets the canonical identifier of the archived candidate.