def compute_throttle(snr_db, beta=0.90, alpha=2.2, snr_min=2.0):
    """Gate 1: SNR-based throttle for channel selection"""
    import numpy as np
    
    # Exponential throttle function
    throttle = beta * (1.0 - np.exp(-alpha * max(snr_db - snr_min, 0)))
    
    return np.clip(throttle, 0.0, 1.0)

# Example: SNR 4.2dB → Throttle 0.85
print(f"Throttle: {compute_throttle(4.2):.3f}")  # Output: 0.850

def compute_adaptive_lr(focus, resilience, base_lr=0.1):
    """Gate 2: Neuroadaptive Learning Rate (η_t)"""
    # Equation 3: ηₜ = 0.1 + 2×focus + 4×resilience
    eta_t = base_lr + 2.0 * focus + 4.0 * resilience
    
    return eta_t

# Example: Focus 0.9, Resilience 0.8
focus = 0.9  # High attention (EEG alpha power)
resilience = 0.8  # Low stress (EEG beta/gamma power)
lr = compute_adaptive_lr(focus, resilience)
print(f"Learning Rate η_t: {lr:.3f}")  # Output: 5.100

def compute_ensemble_confidence(acc_conf, traj_conf, snr_conf):
    """Gate 3: 3-model ensemble voting for robust confidence"""
    import numpy as np
    
    # Weighted voting (accuracy: 40%, trajectory: 35%, SNR: 25%)
    weights = {"acc": 0.40, "traj": 0.35, "snr": 0.25}
    ensemble = acc_conf * weights["acc"] + traj_conf * weights["traj"] + snr_conf * weights["snr"]
    
    # Agreement bonus (if all models agree)
    agreement = 1.0 - min(1.0, np.std([acc_conf, traj_conf, snr_conf]) / 0.3)
    bonus = agreement * 0.10  # Up to +10%
    
    return min(1.0, ensemble + bonus)

# Example: 3 confidence estimators
conf = compute_ensemble_confidence(0.78, 0.72, 0.68)
print(f"Ensemble Confidence: {conf:.3f}")  # Output: 0.734

def apply_feedback_loop(confidence, throttle_base):
    """G3→G1 Feedback: Adjust throttle based on confidence"""
    # High confidence → Increase throttle (more channels)
    # Low confidence → Decrease throttle (fewer channels, focus on quality)
    feedback_signal = confidence  # G3 output
    throttle_adjustment = feedback_signal * 0.15  # ±15% adjustment
    
    throttle_new = throttle_base + throttle_adjustment
    return min(1.0, max(0.0, throttle_new))

# Example: Confidence 0.72 → Boost throttle
throttle = apply_feedback_loop(0.72, 0.85)
print(f"Adjusted Throttle: {throttle:.3f}")  # Output: 0.958