Quantum Information Protection Scheme Based on Reinforcement Learning for Periodic Surface Codes

Figure 4

The 7 layers (marked by blue circles), 14 layers (marked by red triangles), and 21 layers (marked by yellow circles) of the ResNet network are used as the medium training models. After increasing the number of training syndromes, the correction accuracy can be improved to over 95%. The effective error rate function is generated according to the number of iterations.