Device parameter acquisition method and system based on neural network, and related components

• Main Authors: ZHAO YAQIAN, DONG GANG, WANG BINQIANG, LI RENGANG, XU ZHE, YANG HONGBIN, CHAO YINYIN
• Format: Patent
• Language: Chinese; English
• Published: 12.04.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention discloses a device parameter acquisition method, system and device based on a neural network and a computer readable storage medium, and the method comprises the steps: optimizing a corresponding device parameter for a target electromagnetic spectrum response through the neural network composed of an inverse network, a discrimination network and a forward prediction network; the inverse network and the discriminant network are combined for adversarial training, the generated device parameters are optimized to be close to real parameters, meanwhile, a training data set needed for generating the adversarial network is small, hardware resources are saved, the robustness of the neural network is improved, then the device parameters which are generated by the inverse network and meet real conditions are input into the forward prediction network, and the robustness of the neural network is improved. According to the method, the inverse network and the forward prediction network are combined for traini