import cirq import numpy as np import matplotlib.pyplot as plt from time import sleep from sklearn.ensemble import RandomForestClassifier from sklearn.preprocessing import StandardScaler import uuid import json # === CONFIGURARE SIMULARE === NOISE_LEVEL = 0.05 REPETITIONS = 300 # redus pentru viteza BITSTREAM_INTERVAL = 0.0 # eliminat delay-ul TRAINING_WINDOW = 100 MIN_NOISE = 0.01 MAX_NOISE = 0.1 UPDATE_GRAPH_EVERY = 10 # === INITIALIZARE MODEL ML === model = RandomForestClassifier(n_estimators=50, max_depth=5, n_jobs=-1) scaler = StandardScaler() X_train = [] y_train = [] is_model_trained = False recent_bits = [] # === FUNCTII === def test_trc(bit): q0, q1 = cirq.LineQubit.range(2) circuit = cirq.Circuit([ cirq.H(q0), cirq.CNOT(q0, q1), ]) if bit == 1: circuit.append(cirq.X(q1)) noise = cirq.DepolarizingChannel(p=NOISE_LEVEL) circuit = circuit.with_noise(noise) circuit.append([ cirq.measure(q0, key='result'), cirq.measure(q1, key='check') ]) simulator = cirq.Simulator() result = simulator.run(circuit, repetitions=REPETITIONS) return result.histogram(key='result'), result.histogram(key='check') def extract_features(counts): total = sum(counts.values()) return np.array([counts.get(0, 0)/total, counts.get(1, 0)/total]) def error_correction(bits, repeat=3): corrected = [] for i in range(0, len(bits), repeat): chunk = bits[i:i+repeat] corrected.append(1 if sum(chunk) > len(chunk)/2 else 0) return corrected def collect_calibration_data(): counts_0, _ = test_trc(0) counts_1, _ = test_trc(1) p0_g0 = counts_0.get(0, 0) / REPETITIONS p1_g1 = counts_1.get(1, 0) / REPETITIONS return p0_g0, p1_g1 def draw_quantum_circuit(): q0, q1 = cirq.LineQubit.range(2) circuit = cirq.Circuit([ cirq.H(q0), cirq.CNOT(q0, q1), cirq.measure(q0), cirq.measure(q1) ]) print("\nCircuit cuantic folosit:") print(circuit) def text_to_bitstream(text): bits = ''.join(format(ord(c), '08b') for c in text) return [int(b) for b in bits] def bitstream_to_text(bitstream): bytes_list = [bitstream[i:i+8] for i in range(0, len(bitstream), 8) if len(bitstream[i:i+8]) == 8] chars = [] for byte in bytes_list: char_code = sum([b << (7-i) for i, b in enumerate(byte)]) try: chars.append(chr(char_code)) except: chars.append('?') return ''.join(chars) # === GRAFIC LIVE === plt.ion() fig, ax = plt.subplots() ax.set_title("Acuratețe în timp real") ax.set_xlabel("Bit transmis") ax.set_ylabel("Acuratețe (%)") ax.set_ylim(0, 100) ax.grid(True) line, = ax.plot([], [], 'g-', marker='o', markersize=4) # === MAIN LOOP === print("\U0001f4e1 Sistem activ: Transmitere Retrocauzala Cuantica cu Invatare Automata") draw_quantum_circuit() message = input("\n🌤 Introduceți un mesaj: ") bitstream = text_to_bitstream(message) length = len(bitstream) print(f"\n💌 Mesajul '{message}' va fi transmis ca {length} biți.") input("Press Enter to start transmission...") accuracy_log = [] correct = 0 total = 0 received_bits = [] p0_g0, p1_g1 = collect_calibration_data() log_file = open(f"received_messages_{uuid.uuid4()}.txt", "w", encoding="utf-8") training_data_file = open(f"training_data_{uuid.uuid4()}.json", "w", encoding="utf-8") try: round_idx = 0 while True: round_bits = [] noise_adjustment = NOISE_LEVEL round_idx += 1 print(f"\n=== Runda {round_idx} ===") for idx, sent_bit in enumerate(bitstream): measured_counts, check_counts = test_trc(sent_bit) features = extract_features(measured_counts) context_bits = np.array([recent_bits[-(i+1)] if len(recent_bits) > i else -1 for i in range(8)]) features = np.concatenate((features, context_bits)) if is_model_trained: features_scaled = scaler.transform([features])[0] decoded_bit = model.predict([features_scaled])[0] else: features_scaled = scaler.fit_transform([features])[0] decoded_bit = 1 if features_scaled[1] > (p0_g0 + p1_g1) / 2 else 0 is_correct = (decoded_bit == sent_bit) total += 1 if is_correct: correct += 1 accuracy = correct / total * 100 accuracy_log.append(accuracy) if not is_correct and noise_adjustment < MAX_NOISE: noise_adjustment += 0.005 elif is_correct and noise_adjustment > MIN_NOISE: noise_adjustment -= 0.005 globals()['NOISE_LEVEL'] = noise_adjustment X_train.append(features_scaled.tolist()) y_train.append(sent_bit) recent_bits.append(sent_bit) round_bits.append(decoded_bit) received_bits.append(decoded_bit) if len(X_train) > TRAINING_WINDOW: X_train.pop(0) y_train.pop(0) if len(X_train) >= 2: model.fit(X_train, y_train) is_model_trained = True if idx % UPDATE_GRAPH_EVERY == 0 or idx == len(bitstream) - 1: line.set_xdata(range(len(accuracy_log))) line.set_ydata(accuracy_log) ax.relim() ax.autoscale_view() fig.canvas.draw() fig.canvas.flush_events() print(f"Bit #{idx+1}/{length} | Trimis: {sent_bit} | Decodificat: {decoded_bit} | {'Corect' if is_correct else 'GRESIT'} | Acuratețe: {accuracy:.1f}% | Zgomot: {NOISE_LEVEL:.3f}") if BITSTREAM_INTERVAL > 0: sleep(BITSTREAM_INTERVAL) corrected_bits = error_correction(round_bits) reconstructed = bitstream_to_text(corrected_bits) print(f"\n📣 Mesaj reconstruit (cu corectare): {reconstructed}\n") log_file.write(f"Runda {round_idx} | Mesaj primit: {reconstructed} | Acuratețe: {accuracy:.1f}%\n") log_file.flush() except KeyboardInterrupt: print("\nOprit simularea.") training_data = {'X_train': X_train, 'y_train': y_train} json.dump(training_data, training_data_file, indent=2) print("\n=== Statistici finale ===") print(f"Total biți transmiși: {total}") print(f"Biți corecți: {correct}") print(f"Acuratețe finală: {accuracy:.1f}%") print(f"Nivel zgomot final: {NOISE_LEVEL:.3f}") log_file.write(f"\nStatistici finale:\n") log_file.write(f"Total biți: {total}\n") log_file.write(f"Biți corecți: {correct}\n") log_file.write(f"Acuratețe: {accuracy:.1f}%\n") log_file.close() training_data_file.close()