Beyond the Known: Adversarial Autoencoders in Novelty Detection