面向数字创意产业的ANN-SNN混合神经网络架构研究OA

Aiming at the computational bottleneck and low energy efficiency of traditional neural networks when processing large-scale data,this study proposes an innovative hybrid neural network architecture that deeply integrates ANN(artificial neural network)and SNN(spiking neural network).This architecture improves the interactivity,intelligence and efficiency of virtual reality(VR)and animation creation.A dynamic weight allocation mechanism is introduced,which utilizes task-aware α and β coefficients to dynamically adjust the output contributions of ANN and SNN,thereby adapting to the real-time interaction requirements of multimodal data.At the same time,a hierarchical fusion strategy is designed to achieve efficient coupling of spatiotemporal features in the core network layer,solving the response delay problem of traditional models in dynamic scenes.Experimental results show that the ANN-SNN fusion architecture is superior to traditional ANN models and SNN models in multiple dimensions.