Enhance MAP-Elites feature binning and diversity handling

This update adds per-dimension feature bin support, adaptive feature scaling, and a diversity reference set with caching for MAP-Elites in ProgramDatabase. The config now allows feature_bins to be an int or dict, and includes diversity_reference_size. Related tests and evaluator logic are updated for improved cascade validation and feature map coverage.

Author: codelion

openevolve/_version.py

@@ -1,3 +1,3 @@
 """Version information for openevolve package."""
 
-__version__ = "0.0.19"
+__version__ = "0.0.20"

openevolve/config.py

@@ -166,9 +166,10 @@ class DatabaseConfig:
     diversity_metric: str = "edit_distance"  # Options: "edit_distance", "feature_based"
 
     # Feature map dimensions for MAP-Elites
-    # feature_dimensions: List[str] = field(default_factory=lambda: ["score", "complexity"])
-    feature_dimensions: List[str] = field(default_factory=lambda: ["complexity"])
-    feature_bins: int = 10
+    # Default to complexity and diversity for better exploration
+    feature_dimensions: List[str] = field(default_factory=lambda: ["complexity", "diversity"])
+    feature_bins: Union[int, Dict[str, int]] = 10  # Can be int (all dims) or dict (per-dim)
+    diversity_reference_size: int = 20  # Size of reference set for diversity calculation
 
     # Migration parameters for island-based evolution
     migration_interval: int = 50  # Migrate every N generations

openevolve/database.py

@@ -89,7 +89,7 @@ def _load_evaluation_function(self) -> None:
 
             self.evaluate_function = module.evaluate
             logger.info(f"Successfully loaded evaluation function from {self.evaluation_file}")
-            
+
             # Validate cascade configuration
             self._validate_cascade_configuration(module)
         except Exception as e:
@@ -99,16 +99,16 @@ def _load_evaluation_function(self) -> None:
     def _validate_cascade_configuration(self, module) -> None:
         """
         Validate cascade evaluation configuration and warn about potential issues
-        
+
         Args:
             module: The loaded evaluation module
         """
         if self.config.cascade_evaluation:
             # Check if cascade functions exist
             has_stage1 = hasattr(module, "evaluate_stage1")
-            has_stage2 = hasattr(module, "evaluate_stage2") 
+            has_stage2 = hasattr(module, "evaluate_stage2")
             has_stage3 = hasattr(module, "evaluate_stage3")
-            
+
             if not has_stage1:
                 logger.warning(
                     f"Configuration has 'cascade_evaluation: true' but evaluator "
@@ -123,7 +123,9 @@ def _validate_cascade_configuration(self, module) -> None:
                     f"multi-stage evaluation for better cascade benefits."
                 )
             else:
-                logger.debug(f"Cascade evaluation properly configured with available stage functions")
+                logger.debug(
+                    f"Cascade evaluation properly configured with available stage functions"
+                )
 
     async def evaluate_program(
         self,
@@ -305,7 +307,9 @@ def get_pending_artifacts(self, program_id: str) -> Optional[Dict[str, Union[str
         """
         return self._pending_artifacts.pop(program_id, None)
 
-    async def _direct_evaluate(self, program_path: str) -> Union[Dict[str, float], EvaluationResult]:
+    async def _direct_evaluate(
+        self, program_path: str
+    ) -> Union[Dict[str, float], EvaluationResult]:
         """
         Directly evaluate a program using the evaluation function with timeout
 
@@ -616,22 +620,23 @@ async def _llm_evaluate(self, program_code: str, program_id: str = "") -> Dict[s
     def _create_cascade_error_context(self, stage: str, error: Exception) -> dict:
         """
         Create rich error context for cascade failures
-        
+
         Args:
             stage: The stage where the error occurred
             error: The exception that was raised
-            
+
         Returns:
             Dictionary with enhanced error context
         """
         import time
+
         return {
             "failure_stage": stage,
             "error_type": type(error).__name__,
             "error_message": str(error),
             "timestamp": time.time(),
             "cascade_config": self.config.cascade_evaluation,
-            "cascade_thresholds": getattr(self.config, 'cascade_thresholds', []),
+            "cascade_thresholds": getattr(self.config, "cascade_thresholds", []),
             "timeout_config": self.config.timeout,
             "evaluation_file": self.evaluation_file,
         }