Algorithm Rehaul

Author: ItCodinTime

benchmarks/adaptive_neat_vs_baselines.py

@@ -0,0 +1,59 @@
+"""Run the confirmed adaptive-NEAT comparison against Adam and SGD."""
+
+from __future__ import annotations
+
+import json
+from pathlib import Path
+
+from benchmarks.tasks.keras_mlp import (
+    BenchmarkConfig,
+    OptimizerSpec,
+    run_optimizer_suite,
+)
+
+
+def main() -> None:
+    specs = (
+        OptimizerSpec("sgd_momentum", "sgd_momentum", {"learning_rate": 1e-2}),
+        OptimizerSpec("adam", "adam", {"learning_rate": 1e-3}),
+        OptimizerSpec(
+            "neat_adaptive_best",
+            "neat",
+            {
+                "learning_rate": 0.008,
+                "alpha": 0.25,
+                "beta": 0.9,
+                "opponent_source": "previous_gradient",
+                "nce_mode": "projection",
+                "nce_clip_ratio": 1.0,
+                "adaptive_correction": True,
+                "adaptive_correction_decay": 0.9,
+                "adaptive_correction_min_scale": 1.0,
+                "adaptive_correction_max_scale": 2.5,
+                "adaptive_preconditioning": True,
+                "second_moment_beta": 0.999,
+                "bias_correction": True,
+                "precondition_nce": True,
+                "correction_warmup_steps": 0,
+                "conflict_threshold": 0.0,
+            },
+        ),
+    )
+    result = run_optimizer_suite(
+        specs,
+        config=BenchmarkConfig(seeds=(7, 11, 19), epochs=20, batch_size=64),
+    )
+    output = Path("benchmarks/results/neat_adaptive_confirm_2026-04-05.json")
+    output.write_text(json.dumps(result, indent=2))
+    print(output)
+    for row in result["summary"]:
+        print(
+            row["optimizer"],
+            row["mean_test_accuracy"],
+            row["mean_test_loss"],
+            row["mean_seconds"],
+        )
+
+
+if __name__ == "__main__":
+    main()