Optical neural networks (ONNs) have made significant progress in optical computing due to their advantages such as sub-nanosecond latency, low heat dissipation, and high parallelism. This review article discusses the design, principles, and challenges of ONNs, focusing on non-integrated and integrated ONNs. Non-integrated ONNs use volume optical components like 4f systems and diffractive elements, while integrated ONNs use on-chip optical components. The article summarizes the computational density, nonlinearity, scalability, and practical applications of ONNs, and discusses their future development trends. ONNs have the potential to provide support for the further development of artificial intelligence with a novel computing paradigm. The review highlights the challenges and perspectives of ONNs in future development, including the need for higher computational power, better energy efficiency, and more practical applications. The article also discusses the recent advancements in ONNs, including the use of diffractive metasurfaces, MRR weight banks, and other on-chip optical components. The review concludes that ONNs have the potential to revolutionize computing by providing a new paradigm for artificial intelligence.Optical neural networks (ONNs) have made significant progress in optical computing due to their advantages such as sub-nanosecond latency, low heat dissipation, and high parallelism. This review article discusses the design, principles, and challenges of ONNs, focusing on non-integrated and integrated ONNs. Non-integrated ONNs use volume optical components like 4f systems and diffractive elements, while integrated ONNs use on-chip optical components. The article summarizes the computational density, nonlinearity, scalability, and practical applications of ONNs, and discusses their future development trends. ONNs have the potential to provide support for the further development of artificial intelligence with a novel computing paradigm. The review highlights the challenges and perspectives of ONNs in future development, including the need for higher computational power, better energy efficiency, and more practical applications. The article also discusses the recent advancements in ONNs, including the use of diffractive metasurfaces, MRR weight banks, and other on-chip optical components. The review concludes that ONNs have the potential to revolutionize computing by providing a new paradigm for artificial intelligence.