样本

最速下降法traingd

弹性算法trainrp

动量算法traingdm

Fletcher-Reeves修正算法

N

[0,0,0.286,0.668,0.651]

[0,0,0.093,0.376,0.004]

[0,0,0.149,0.323,0.649]

[0,0,0.001,0.823,0.903]

O

[0,0,0.016,0.496,0.369]

[0,0,0.012,0.730,0.428]

[0,0,0.003,0.888,0.016]

[0,0,0.002,0.900,0.185]

P

[0,0,0.118,0.782,0.020]

[0,0,0.670,0.269,0.845]

[0,0,0.158,0.867,0.026]

[0,0,0.007,0.900,0.245]

Q

[0,0,0.027,0.054,0.081]

[0,0,0.539,0.004,e]

[0,0,0.007,0.893,e]

[0,0,0.877,0.892,0.000]

R

[0,0,0.650,0.896,0.001]

[0,0,0.873,0.334,0.002]

[0,0,0.644,0.383,0.060]

[0,0,0.800,0.598,0.002]

S

[0,0,0.159,0.828,0.085]

[0,0,0.898,0.208,0.006]

[0,0,0.835,0.153,0.200]

[0,0,0.633,0.192,0.006]

T

[0,0,0.094,0.684,0.566]

[0,0,0.030,0.859,0.002]

[0,0,0.165,0.727,0.044]

[0,0,0.002,0.900,0.565]

U

[0,0,0.100,0.693,0.256]

[0,0,0.028,0.847,0.002]

[0,0,0.189,0.719,0.091]

[0,0,0.008,0.900,0.422]

训练步数

348步收敛

141步收敛

973步收敛

26步收敛

评价结果

部分不准确，不稳定

接近准确，不太稳定

部分不准确，不稳定

稳定但不准确，偏差较大