Research Article

Hybridizing Evolutionary Computation and Deep Neural Networks: An Approach to Handwriting Recognition Using Committees and Transfer Learning

Table 4

Summary of accuracies of the best 20 GE individuals after full training in EMNIST Letters.

# MeanStd. Dev.Min.MedianMax.

194.65850.07400494.66594.5194.76
294.73000.07093494.74094.6194.86
394.56350.08487094.58094.4194.70
495.12150.04955395.12095.0295.19
594.27900.06488094.27594.1694.38
694.72300.06883294.71094.6394.88
794.65400.05364694.65094.5394.74
893.72700.09442993.70593.6093.93
993.65150.07005893.64593.5093.75
1094.53050.13907594.55094.1894.75
1194.68900.04541094.69094.6194.78
1294.43350.07154594.44094.2894.55
1394.43300.07588894.43594.2894.61
1494.62600.13204594.63094.2494.81
1594.46050.06261694.46594.3594.57
1694.15900.09357894.16594.0094.29
1793.45050.15642393.45593.1493.81
1893.50950.14992993.50593.2593.77
1994.27250.07468494.27094.1394.38
2094.50850.07020894.49594.4194.64