Convolutional Neural Network Analysis Framework based on Cyclic MCU Simulation Software Data: from the Perspective of Limit Testing

In this paper, an algorithm research on the limit test cycle microcontroller based on convolutional neural network is proposed, and a limit test classifier is obtained by training the network structure model of Network.In.Network (NIN). Since the global mean pooling is used in the model to replace the fully connected layer in the traditional network, it can effectively prevent the data over-fitting phenomenon in the training process. A virtual experiment scheme based on Proteus software simulation is proposed. Need to purchase a lot of hardware equipment, but use Proteus software simulation to complete the experiment and training of the microcontroller. At the same time, the virtual experiment of single-chip microcomputer using Proteus software is convenient to design and modify the peripheral circuit of single-chip microcomputer.