Generative Adversarial Networks in Feature Engineering
Kavita Khobragade, Poonam Ponde · Advances in geospatial technologies book series · 2025
Feature engineering, the process of extracting informative representations from raw data, is crucial for successful remote sensing applications. An automated feature engineering is provided by Generative Adversarial Networks (GANs), which is a type of deep learning models based on unsupervised learning. In feature engineering, training strategies for GANs are also reviewed, with an emphasis on methods to enhance convergence, stabilize training, and reduce problems like mode collapse and overfitting. In addition, the chapter addresses issues like computing efficiency, scalability, and integration with current analytical frameworks, when implementing GAN-based feature engineering techniques in practical remote sensing applications. This chapter concludes with a strong argument that using GANs is an effective tool for unsupervised feature engineering. Through the utilization of GANs' capacity to acquire insightful latent representations from unlabeled data, scholars and professionals might open up novel avenues for the extraction of significant insights from remote sensing data.