import numpy as np

def nmse(gt, pred):
    """Compute Normalized Mean Squared Error (NMSE)"""
    return np.linalg.norm(gt - pred) ** 2 / np.linalg.norm(gt) ** 2