Update network.py

network.py

@@ -7,13 +7,13 @@ from typing import Dict, Optional
 from .node import CognitiveNode
 
 class DynamicCognitiveNet(nn.Module):
-    """Jaringan dengan manajemen koneksi yang robust"""
+    """Arsitektur jaringan dengan manajemen tensor yang robust"""
     def __init__(self, input_size: int, output_size: int):
         super().__init__()
         self.input_size = input_size
         self.output_size = output_size
         
-        # Node input/output
+        # Node dengan input size 1
         self.input_nodes = nn.ModuleList([
             CognitiveNode(i, 1) for i in range(input_size)
         ])
@@ -25,7 +25,7 @@ class DynamicCognitiveNet(nn.Module):
         self.connections = nn.ParameterDict()
         self._init_base_connections()
         
-        # Konteks pembelajaran
+        # Sistem pembelajaran
         self.emotional_state = nn.Parameter(torch.tensor(0.0))
         self.optimizer = optim.AdamW(self.parameters(), lr=0.001)
         self.loss_fn = nn.MSELoss()
@@ -40,6 +40,9 @@ class DynamicCognitiveNet(nn.Module):
                 )
 
     def forward(self, x: torch.Tensor) -> torch.Tensor:
+        # Validasi dimensi input
+        x = x.view(-1)
+        
         # Pemrosesan input
         activations = {}
         for i, node in enumerate(self.input_nodes):
@@ -61,8 +64,8 @@ class DynamicCognitiveNet(nn.Module):
         return torch.stack(outputs).squeeze()
 
     def structural_update(self, global_reward: float):
-        """Update struktur dengan validasi koneksi"""
-        # Adaptasi kekuatan koneksi
+        """Update struktur jaringan"""
+        # Update kekuatan koneksi
         for conn_id in list(self.connections.keys()):
             new_weight = self.connections[conn_id] + 0.1 * global_reward
             self.connections[conn_id].data = new_weight.clamp(-1, 1)
@@ -74,7 +77,7 @@ class DynamicCognitiveNet(nn.Module):
                 self.connections[new_conn] = nn.Parameter(torch.randn(1) * 0.1)
 
     def _find_underutilized_connection(self) -> Optional[str]:
-        """Mencari koneksi input-output yang underutilized"""
+        """Mencari pasangan node yang kurang aktif"""
         input_act = {n.id: np.mean(n.recent_activations) 
                     for n in self.input_nodes if n.recent_activations}
         output_act = {n.id: np.mean(n.recent_activations)
@@ -92,10 +95,10 @@ class DynamicCognitiveNet(nn.Module):
         self.optimizer.zero_grad()
         
         try:
-            pred = self(x)
-            loss = self.loss_fn(pred, y)
-        except RuntimeError as e:
-            print(f"Error selama forward pass: {e}")
+            pred = self(x.view(-1))
+            loss = self.loss_fn(pred, y.view(-1))
+        except Exception as e:
+            print(f"Error forward: {e}")
             return float('nan')
         
         # Regularisasi struktural
@@ -105,16 +108,14 @@ class DynamicCognitiveNet(nn.Module):
         try:
             total_loss.backward()
             self.optimizer.step()
-        except RuntimeError as e:
-            print(f"Error selama backpropagation: {e}")
+        except Exception as e:
+            print(f"Error backward: {e}")
             return float('nan')
         
-        # Update state emosional
+        # Update emosi
         self.emotional_state.data = torch.sigmoid(
             self.emotional_state + (0.5 - loss.item()) * 0.1
         )
-        
-        # Update struktur
         self.structural_update(0.5 - loss.item())
         
         return total_loss.item()