Delete holographic_transformer_model.py

holographic_transformer_model.py

@@ -1,114 +0,0 @@
-import torch
-import torch.nn as nn
-import torch.optim as optim
-from torch.optim.lr_scheduler import CosineAnnealingLR
-from tqdm import tqdm
-
-class HolographicMasterCodeTransformer(nn.Module):
-    def __init__(self, input_dim=8, d_model=128, nhead=8, num_layers=6, output_dim=1, n_harmonics=4):
-        super().__init__()
-        self.d_model = d_model
-        self.n_harmonics = n_harmonics
-        self.alpha_0 = 1.0 / 137.035
-        
-        self.input_proj = nn.Linear(input_dim, d_model)
-        
-        encoder_layer = nn.TransformerEncoderLayer(
-            d_model=d_model, 
-            nhead=nhead, 
-            dim_feedforward=d_model * 4,
-            dropout=0.1,
-            activation='gelu',
-            batch_first=True,
-            norm_first=True
-        )
-        self.transformer = nn.TransformerEncoder(encoder_layer, num_layers=num_layers)
-        
-        self.harmonic_weights = nn.Parameter(torch.randn(n_harmonics, d_model))
-        
-        self.output_head = nn.Sequential(
-            nn.LayerNorm(d_model),
-            nn.Linear(d_model, d_model // 2),
-            nn.GELU(),
-            nn.Dropout(0.1),
-            nn.Linear(d_model // 2, output_dim)
-        )
-        
-    def forward(self, x, alpha=None):
-        if alpha is None:
-            alpha = self.alpha_0
-            
-        h = self.input_proj(x)
-        h = h.unsqueeze(1)
-        h = self.transformer(h)
-        h = h.squeeze(1)
-        
-        # Master Code Ψ
-        psi = torch.zeros_like(h)
-        for n in range(self.n_harmonics):
-            phase = 2 * torch.pi * (n + 1) * (alpha - self.alpha_0) * 800.0   # Yumuşatıldı
-            psi += self.harmonic_weights[n] * torch.cos(h * (n + 1) + phase)
-        
-        # Güvenlik Kilidi (daha yumuşak)
-        alpha_dev = torch.abs(alpha - self.alpha_0)
-        security_factor = torch.exp(-300.0 * alpha_dev)   # Daha yumuşak
-        
-        h = h + 0.35 * psi * security_factor
-        return self.output_head(h)
-    
-    def holographic_reg(self):
-        reg = sum(torch.norm(w, p=1) for w in self.harmonic_weights)
-        return 0.0015 * reg / self.n_harmonics   # Daha dengeli
-
-
-# Eğitim fonksiyonu (300 epoch)
-def train_holographic(alpha_values, epochs=300):
-    results = {}
-    for name, alpha_val in alpha_values.items():
-        print(f"\n→ Eğitim: {name} (α = {alpha_val:.8f})")
-        model = HolographicMasterCodeTransformer()
-        optimizer = optim.AdamW(model.parameters(), lr=0.001, weight_decay=1e-5)
-        scheduler = CosineAnnealingLR(optimizer, T_max=epochs, eta_min=1e-6)
-        criterion = nn.MSELoss()
-        
-        X = torch.randn(300, 8)
-        y = torch.sin(X.sum(dim=1, keepdim=True)) * 0.9 + 0.05 * torch.randn(300, 1)
-        
-        best_loss = float('inf')
-        for epoch in tqdm(range(epochs)):
-            optimizer.zero_grad()
-            pred = model(X, alpha=alpha_val)
-            loss = criterion(pred, y)
-            reg = model.holographic_reg()
-            total_loss = loss + reg
-            
-            total_loss.backward()
-            torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)
-            optimizer.step()
-            scheduler.step()
-            
-            if loss.item() < best_loss:
-                best_loss = loss.item()
-                
-            if epoch % 50 == 0 or epoch == epochs-1:
-                print(f"Epoch {epoch:3d} | Loss: {loss.item():.6f} | Reg: {reg.item():.6f}")
-        
-        results[name] = best_loss
-    return results
-
-# ====================== Çalıştır ======================
-if __name__ == "__main__":
-    alpha_values = {
-        "Nominal": 1/137.035,
-        "Physical": 1/137.035999,
-        "Sapma1": 1/137.000,
-        "Sapma2": 1/137.070
-    }
-    
-    results = train_holographic(alpha_values, epochs=300)
-    
-    # En iyi sonucu yazdır
-    best_alpha = min(results, key=lambda k: results[k]['final_loss'])
-    print("\n" + "="*70)
-    print(f"EN İYİ SONUÇ → {best_alpha} α ile Final Loss = {results[best_alpha]:.6f}")
-    print("="*70)