The paper introduces a novel framework for adversarial domain adaptation, called Adversarial Discriminative Domain Adaptation (ADDA), which combines discriminative modeling, untied weight sharing, and a GAN loss. The authors propose this method to address the limitations of existing approaches, which either focus on generative modeling or impose tied weights on the model. ADDA is designed to handle larger domain shifts and achieve state-of-the-art results on standard cross-domain digit classification tasks and a more challenging cross-modality object classification task. The framework unifies various design choices and facilitates the development of new adaptation methods. The experiments demonstrate that ADDA outperforms competing domain-adversarial methods in both unsupervised and cross-modality adaptation tasks, showing its effectiveness in mitigating the effects of domain shift.The paper introduces a novel framework for adversarial domain adaptation, called Adversarial Discriminative Domain Adaptation (ADDA), which combines discriminative modeling, untied weight sharing, and a GAN loss. The authors propose this method to address the limitations of existing approaches, which either focus on generative modeling or impose tied weights on the model. ADDA is designed to handle larger domain shifts and achieve state-of-the-art results on standard cross-domain digit classification tasks and a more challenging cross-modality object classification task. The framework unifies various design choices and facilitates the development of new adaptation methods. The experiments demonstrate that ADDA outperforms competing domain-adversarial methods in both unsupervised and cross-modality adaptation tasks, showing its effectiveness in mitigating the effects of domain shift.