GANS and Meta-Learning
Munish Kumar · Advances in computational intelligence and robotics book series · 2025
This chapter introduces GANs and Meta-Learning as fundamental approaches to Artificial Intelligence necessary for tasks from synthetic data generation to task-based sample learning. Nevertheless, extensive research shows that both GANs and Meta-Learning suffer from important drawbacks preventing practical application. The authors examine more than articles published in the last few years to provide an overview of challenges and open questions for both methods. The problems that GANs face are training non-convergence, trapping, scalability, understanding, and several others. There are deficits in scalability, generalization, task diverseness and high computational costs in Meta-Learning. This chapter also discusses the relationship between GANs and Meta-Learning and, although promoting enrichment and possible improvements for both, it discusses the possible benefits and challenges of combining different techniques. Hence, the find that the survey, by focusing on practical examples, has stressed on some key barriers to emphasise on the usability and effectiveness issues.