The recent developments in the research area of wireless network optimization and design have shown a significant shift towards integrating advanced machine learning techniques with traditional optimization methods. This integration aims to enhance the adaptability, efficiency, and accuracy of network solutions, particularly in dynamic and real-time scenarios. Key innovations include the use of deep unfolding frameworks for joint optimization problems, such as beamforming and power control in multi-port matching networks, and the application of graph neural networks for solving NP-hard optimization problems in mobile edge computing networks. Additionally, there is a growing focus on developing fast adaptation algorithms that can quickly respond to abrupt changes in network conditions, leveraging zero-shot updates and online learning strategies. These advancements are crucial for improving the performance of next-generation wireless networks, including ultra-dense LEO satellite systems and integrated multi-tier vehicular networks. Notably, the integration of optimization theory with deep learning is proving to be a powerful approach, offering a balance between the interpretability of traditional methods and the adaptability of deep learning models. This trend is expected to continue, driving further innovations in the field.