Leveraging Synthetic Data Generation for Military Decision Support Enhancement
Naveiro · 2024
Synthetic data generation has become an important approach to tackle challenges related to data scarcity, privacy concerns, and resource optimization in artificial intelligence applications. This paper examines recent advances in synthetic data generation methods, with a focus on generative learning, transfer learning, and modelling techniques. Generative learning uses machine learning models to replicate statistical patterns found in real-world data. Meanwhile, transfer learning allows for knowledge transfer across related tasks, reducing the impact of data scarcity. Modeling techniques, such as statistical and machine learning based approaches, creates synthetic data that closely mirrors real data distributions. This paper examines various methodologies and case studies and their significance in different application domains, with an emphasis on the military. Additionally, benchmarking analyses demonstrate the effectiveness of Generative Adversarial Networks and Variational Autoencoders in synthetic data generation tasks. Transfer learning strategies are evaluated considering their advantages/disadvantages and the field of application. Modelling techniques are evaluated to generate synthetic scenarios. The paper concludes by discussing the importance of synthetic data generation to enhance decision support in military domain.