Why GANs are overkill for NLP
This work addresses the inefficiency of GANs in NLP for researchers and practitioners, offering a theoretical justification for preferring maximum-likelihood methods, which is incremental as it builds on existing understanding of generative modeling.
The paper argues that maximum-likelihood approaches are more efficient than GANs for sequential tasks like natural language generation, showing that minimizing KL-divergence effectively minimizes the same distinguishability criteria as adversarial models, with reductions applicable to models like n-grams and neural networks with softmax outputs.
This work offers a novel theoretical perspective on why, despite numerous attempts, adversarial approaches to generative modeling (e.g., GANs) have not been as popular for certain generation tasks, particularly sequential tasks such as Natural Language Generation, as they have in others, such as Computer Vision. In particular, on sequential data such as text, maximum-likelihood approaches are significantly more utilized than GANs. We show that, while it may seem that maximizing likelihood is inherently different than minimizing distinguishability, this distinction is largely artificial and only holds for limited models. We argue that minimizing KL-divergence (i.e., maximizing likelihood) is a more efficient approach to effectively minimizing the same distinguishability criteria that adversarial models seek to optimize. Reductions show that minimizing distinguishability can be seen as simply boosting likelihood for certain families of models including n-gram models and neural networks with a softmax output layer. To achieve a full polynomial-time reduction, a novel next-token distinguishability model is considered.