self-validation-loop

Triadic color validation compares predicted colors from a color system against actual observed values using a self-validation loop. Feed seed, indices, and color sources (splitmix_ternary, xoroshiro_3color, gay_mcp) into the loop with a comparator to measure accuracy and surprise, outputting a structured log with error metrics and pass/fail results based on your threshold.

Accuracy measures how closely predictions match observations; surprise quantifies unexpected deviations. The self-validation loop aggregates both metrics across evaluations to guide learning, enforcing thresholds like accuracy ≥ 0.99 or zero surprise to determine whether the color system passes validation.

Yes. The self-validation loop is designed to work with indices-driven color streams from these three sources. It applies reafference or a comparator module to measure prediction accuracy and surprise across all three, generating auditable JSON logs for each evaluation run.

Provide seed (random initialization), indices (range to validate, e.g., 1..20), color sources (splitmix_ternary, xoroshiro_3color, or gay_mcp), and a comparator to measure predicted vs. observed alignment. The loop outputs accuracy, surprise, errors, and a pass/fail decision based on your threshold.

A structured JSON log containing seed, indices, predicted colors, observed colors, error values, accuracy percentage, surprise metric, and a pass/fail result. This auditable record enables tracking validation performance and refining color system parameters.

Automate validation when you need repeatable, metric-driven evaluation across multiple indices and color sources. Self-validation loops enforce consistent pass thresholds and generate comparable logs, scaling beyond manual spot-checks to catch systemic prediction drift.