My research is dedicated to advancing both the theoretical foundations and practical capabilities of deep neural networks. I focus on how these networks can effectively approximate complex functions and patterns across various domains. A key part of my work involves minimizing critical errors, such as approximation, generalization, and optimization errors, to bridge the gap between theoretical models and their practical applications.
In addition to these foundational aspects, I explore a wide range of activation functions and delve into the optimization processes within neural networks to improve their performance. My research also extends to investigating emerging neural network architectures, such as Transformer models and graph neural networks, and their unique expressive capabilities. Furthermore, I am interested in cutting-edge topics like transfer learning, neural architecture search, and model compression techniques. Through this multifaceted approach, my goal is to push the boundaries of deep learning, contributing to both theoretical advancements and practical innovations in artificial intelligence.