vG.0.1/train_quantum_classifier

import numpy as np
import torch
import torch.nn as nn
import torch.optim as optim
from qiskit import QuantumCircuit
from qiskit.circuit.library import RealAmplitudes, ZFeatureMap
from qiskit_aer import AerSimulator
from qiskit_machine_learning.connectors import TorchConnector
from qiskit_machine_learning.neural_networks import SamplerQNN
from sklearn.metrics import roc_auc_score

# 1. Load Data (The Optimized Set from Phase VI)
print("Loading validated data...")
data = np.load('qgan_data_optimized.npz')
X_train = data['X_train']
X_test = data['X_test']
y_train = data['y_train']
y_test = data['y_test']

# 2. Select Top 4 Features Only
# Quantum hardware is small. We take the 4 most important physics signals.
# (Beta_94, Beta_93, Beta_91, Beta_82 based on your previous output)
num_qubits = 4
X_train_q = X_train[:, :num_qubits]
X_test_q = X_test[:, :num_qubits]

# Convert to PyTorch Tensors
X_train_tensor = torch.tensor(X_train_q, dtype=torch.float32)
y_train_tensor = torch.tensor(y_train, dtype=torch.float32).reshape(-1, 1)
X_test_tensor = torch.tensor(X_test_q, dtype=torch.float32)

print(f"Training Quantum Model on {len(X_train)} shots using {num_qubits} Qubits.")

# 3. Build the Quantum Circuit
# A. Feature Map: Encodes classical data into quantum states (Rotation Z)
feature_map = ZFeatureMap(feature_dimension=num_qubits, reps=1)

# B. Ansatz: The "Neural Network" weights (Entanglement + Rotations)
ansatz = RealAmplitudes(num_qubits, reps=1)

# C. Combine
qc = QuantumCircuit(num_qubits)
qc.compose(feature_map, inplace=True)
qc.compose(ansatz, inplace=True)

# 4. Define the Quantum Neural Network
qnn = SamplerQNN(
    circuit=qc,
    input_params=feature_map.parameters,
    weight_params=ansatz.parameters,
    interpret=lambda x: x % 2, # Map quantum measurements to 0 or 1
    output_shape=2
)

# 5. Hybrid PyTorch Module
class HybridQNN(nn.Module):
    def __init__(self):
        super().__init__()
        self.qnn = TorchConnector(qnn)
        # Classical post-processing (Weighted sum of quantum output)
        self.fc = nn.Linear(1, 1) 
        self.sigmoid = nn.Sigmoid()

    def forward(self, x):
        # Scale input to [0, 2pi] for quantum rotation
        x_scaled = x * torch.pi 
        x_q = self.qnn(x_scaled)
        # We look at probability of state '1' (Disruption)
        x_out = self.fc(x_q[:, 1:2]) 
        return self.sigmoid(x_out)

model = HybridQNN()
optimizer = optim.Adam(model.parameters(), lr=0.01)
loss_func = nn.BCELoss()

# 6. Training Loop
epochs = 15 # Kept short for demonstration speed
loss_list = []

print(f"\n🚀 Starting Hybrid Training ({epochs} epochs)...")
model.train()
for epoch in range(epochs):
    optimizer.zero_grad()
    
    # Forward pass
    output = model(X_train_tensor)
    loss = loss_func(output, y_train_tensor)
    
    # Backward pass
    loss.backward()
    optimizer.step()
    
    loss_list.append(loss.item())
    print(f"   Epoch {epoch+1}/{epochs} | Loss: {loss.item():.4f}")

# 7. Evaluation
print("\nEvaluating Quantum Model...")
model.eval()
with torch.no_grad():
    test_preds = model(X_test_tensor).numpy()

auc = roc_auc_score(y_test, test_preds)
print("="*40)
print(f"🔬 QUANTUM MODEL RESULTS")
print("="*40)
print(f"✅ QNN AUC Score: {auc:.4f}")
print(f"📊 Classical Benchmark to Beat: 0.8951")
print("="*40)

if auc > 0.8951:
    print("🏆 RESULT: QUANTUM ADVANTAGE ACHIEVED (on this dataset)")
elif auc > 0.85:
    print("🥈 RESULT: Competitive Performance (Close to Classical)")
else:
    print("⚠️ RESULT: Underperformed (Expected for simple NISQ circuits)")