evonet 0.1.0.dev12__tar.gz → 0.1.0.dev13__tar.gz

{evonet-0.1.0.dev12 → evonet-0.1.0.dev13}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: evonet
-Version: 0.1.0.dev12
+Version: 0.1.0.dev13
 Summary: Evolvable neural network core for integration with EvoLib
 Author-email: EvoLib <evolib@dismail.de>
 License: MIT License

{evonet-0.1.0.dev12 → evonet-0.1.0.dev13}/evonet/core.py

@@ -230,12 +230,12 @@ class Nnet:
         Args:
             source (Neuron): Source neuron.
             target (Neuron): Target neuron.
-            weight (float | None): Initial weight. If None, random value is used.
+            weight (float | None): Initial weight. If None, 0.0 is used.
             conn_type (ConnectionType): Type of connection (e.g. standard, recurrent).
         """
 
         if weight is None:
-            weight = np.random.randn() * 0.5
+            weight = 0.0
 
         conn = Connection(source, target, weight=weight, conn_type=conn_type)
         source.outgoing.append(conn)

{evonet-0.1.0.dev12 → evonet-0.1.0.dev13}/evonet/mutation.py

@@ -152,21 +152,44 @@ def mutate_biases(net: Nnet, probability: float = 1.0, std: float = 0.1) -> None
 def add_random_connection(
     net: Nnet,
     allowed_recurrent: Optional[Collection[RecurrentKind | str]] = None,
+    connection_init: Literal["zero", "random", "near_zero"] = "zero",
 ) -> bool:
     """
     Add a valid connection between two randomly chosen neurons.
 
     Rules:
-    - disallow connections into INPUT neurons
-    - disallow duplicate (source, target) pairs
-    - classify connection type by layer order:
-      * src_layer < dst_layer  -> STANDARD
-      * src_layer >= dst_layer -> RECURRENT
-
-    Recurrent edges are filtered by allowed_recurrent
-    (list of 'direct' | 'lateral' | 'indirect').
+    - Disallow connections into INPUT neurons.
+    - Disallow duplicate (source, target) pairs.
+    - Classify connection type by layer order:
+        * src_layer < dst_layer  -> STANDARD
+        * src_layer >= dst_layer -> RECURRENT
+    - Recurrent edges are filtered by `allowed_recurrent`
+      (list of 'direct' | 'lateral' | 'indirect').
+
+    Args:
+        net (Nnet): Target network to modify.
+        allowed_recurrent (Collection[RecurrentKind | str] | None):
+            Allowed recurrent connection kinds. Default: None (no recurrent edges).
+        connection_init (Literal["zero", "random", "near_zero"]):
+            Weight initialization mode for the new connection.
+            - "zero":     Weight = 0.0 (neutral; ideal with HELI)
+            - "random":   Weight ~ N(0, 0.5)
+            - "near_zero": Weight ~ U(-0.05, 0.05)
+
+    Returns:
+        bool: True if a connection was added, False otherwise.
     """
 
+    # Weight initialization
+    if connection_init == "zero":
+        weight = 0.0
+    elif connection_init == "random":
+        weight = np.random.randn() * 0.5
+    elif connection_init == "near_zero":
+        weight = np.random.uniform(-0.05, 0.05)
+    else:
+        raise ValueError(f"Unknown init mode: {connection_init}")
+
     # normalize to set[RecurrentKind]
     kinds: set[RecurrentKind] = set()
     if allowed_recurrent:
@@ -223,7 +246,7 @@ def add_random_connection(
         else ConnectionType.RECURRENT
     )
 
-    net.add_connection(src, dst, conn_type=conn_type)
+    net.add_connection(src, dst, weight=weight, conn_type=conn_type)
     return True
 
 

{evonet-0.1.0.dev12 → evonet-0.1.0.dev13}/evonet.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: evonet
-Version: 0.1.0.dev12
+Version: 0.1.0.dev13
 Summary: Evolvable neural network core for integration with EvoLib
 Author-email: EvoLib <evolib@dismail.de>
 License: MIT License

{evonet-0.1.0.dev12 → evonet-0.1.0.dev13}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "evonet"
-version = "0.1.0.dev12"
+version = "0.1.0.dev13"
 description = "Evolvable neural network core for integration with EvoLib"
 authors = [
     { name = "EvoLib", email = "evolib@dismail.de" }