神经网络训练结果分析：Epochs 与预测输出变化趋势

神经网络在训练过程中，会随着 epochs 的增加，逐渐降低误差，最终达到一个稳定状态。以下表格展示了该神经网络在不同 epochs 下的训练误差：

| Epochs | Error | |---|---| | 0 | 59.00997922718588 | | 1000 | 58.73531897202282 | | 2000 | 58.735309548897554 | | 3000 | 58.735306380290325 | | 4000 | 58.73530479068084 | | 5000 | 58.73530383519302 | | 6000 | 58.73530319747653 | | 7000 | 58.73530274160723 | | 8000 | 58.73530239950851 | | 9000 | 58.73530213331459 |

可以看到，训练误差在开始阶段下降得比较快，之后逐渐趋于平缓。

最终，该神经网络在 9000 个 epochs 后，预测输出结果为：

[ [ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ], 
[ 0.99999374, 1., 1. ] ]

对应的真实输出结果为：

[ [ 1.715, 93.52, 77.84 ], 
[ 2.81, 96.28, 83.2 ], 
[ 1.83, 93.93, 86.23 ], 
[ 2.91, 96.41, 86.65 ], 
[ 1.89, 94.12, 82.12 ], 
[ 3.425, 96.95, 94.33 ], 
[ 2.095, 94.7, 83.01 ], 
[ 3.4, 96.93, 82.88 ], 
[ 2.235, 95.03, 86.04 ], 
[ 3.845, 97.3, 75.97 ], 
[ 1.39, 92.01, 82.33 ], 
[ 2.015, 94.82, 82.19 ], 
[ 1.705, 93.48, 90.36 ], 
[ 2.75, 96.7, 83.05 ], 
[ 1.55, 92.83, 89.47 ], 
[ 3.165, 96.7, 89.29 ], 
[ 1.915, 94.2, 74.37 ], 
[ 2.81, 96.28, 91.49 ], 
[ 1.83, 93.93, 72.79 ], 
[ 2.725, 96.17, 85.73 ], 
[ 1.17, 90.5, 65.03 ], 
[ 2.365, 95.58, 90.2 ], 
[ 1.365, 91.86, 83.33 ], 
[ 2.935, 96.44, 92.31 ], 
[ 1.6, 93.06, 80.56 ], 
[ 2.445, 95.73, 90.11 ], 
[ 1.45, 92.34, 80.91 ], 
[ 2.525, 96.43, 84.24 ], 
[ 1.54, 92.78, 82.31 ], 
[ 2.725, 96.17, 94.59 ], 
[ 1.085, 89.76, 43.91 ], 
[ 1.775, 94.12, 88.76 ], 
[ 1.08, 93.83, 80.63 ], 
[ 2.05, 94.91, 82.65 ], 
[ 1.905, 94.17, 87.4 ], 
[ 2.115, 95.06, 80.98 ], 
[ 1.31, 91.52, 76.54 ], 
[ 3.105, 96.64, 83.82 ], 
[ 1.97, 94.36, 87.5 ], 
[ 2.96, 96.47, 86.47 ], 
[ 0.87, 87.23, 83.89 ], 
[ 1.74, 94., 90.12 ], 
[ 1., 88.89, 79.17 ], 
[ 2.02, 94.83, 86.21 ], 
[ 0.995, 88.83, 77.52 ], 
[ 2.06, 94.93, 87.12 ], 
[ 1.11, 89.99, 76.39 ], 
[ 2.085, 94.99, 88.83 ], 
[ 1.365, 91.86, 76.39 ], 
[ 2.425, 95.69, 84.38 ] ]

通过对比预测输出和真实输出，我们可以评估该神经网络的预测准确性。