Fig. 3: Influence of low- and high-momentum parts on classification.
From: Heuristic machinery for thermodynamic studies of SU(N) fermions with neural networks
Classification accuracy with neural networks trained only with low a and high b momentum parts. Red and gray bars represent the accuracy of pre-trained and re-trained NNs, respectively. Retraining of the neural networks with only high- or low-momentum parts enhances the accuracy as shown in gray bars. For high-momentum parts, classification accuracy increases with N as shown in black ticks in b, where the average of the four investigated SU(N) cases is 64.30%. For the re-trained NN, the output probabilities of the NN are shown for input images.
