Advancing Ante-Hoc Explainable Models through Generative Adversarial Networks