r1
r2
<1
<1
吸引子
c
>1
>1
排斥子
p
>1
<1
鞍点
a
<1
>1
反鞍点
fa
本文制作一个二分类网络用来分类p和a，通过改变测试集p和a的比例观察网路分类能力的变化，并将得到的数据拟合成数学表达式。
训练集的比例p：a=1：1，
测试集p和a的比例为x：y.
让x：y的比例分别为0：10，1：9，2：8，3：7，4：6，5：5，6：4，7：3，8：2，9：1，10：0。
实验过程
二分类排斥子和鞍点
制作一个4*4*2的网络向这个的左侧输入排斥子p，并让左侧网络向1，0收敛；向右侧网络输入鞍点a让右侧向0，1收敛，并让4*4*2部分权重共享，前面大量实验表明这种效果相当于将两个弹性系数为k1，k2的弹簧并联成一个弹性系数为k的弹簧，并且让k1=k2=k/2的过程。
这个网络的收敛标准是
if (Math.abs(f2[0]-y[0])< δ && Math.abs(f2[1]-y[1])< δ )
因为对应每个收敛标准δ都有一个特征的迭代次数n与之对应因此可以用迭代次数曲线n(δ)来评价网络性能。
本文尝试了δ从0.5到1e-6在内的36个值.
具体进样顺序
进样顺序
迭代次数
δ=0.5
P
1
判断是否达到收敛
a
2
判断是否达到收敛
梯度下降
P
3
判断是否达到收敛
a
4
判断是否达到收敛
梯度下降
……
达到收敛标准测量准确率,记录迭代次数，将这个过程重复199次
δ=0.4
…
δ=1e-6
将这个网络简写成
d2（p，a）-4-4-2-（2*k）,k∈{0,1}
排斥子的初始化方式
Random rand1 =new Random();
int ti1=rand1.nextInt(99)+1;
x[0]=sig(1+ ((double)ti1/100) );
Random rand2 =new Random();
int ti2=rand2.nextInt(99)+1;
x[3]=sig ( 1+((double)ti2/100) );
鞍点的初始化方式
Random rand1 =new Random();
int ti1=rand1.nextInt(99)+1;
x[0]=sig(1+ ((double)ti1/100) );
Random rand2 =new Random();
int ti2=rand2.nextInt(99)+1;
x[3]=sig ( ((double)ti2/100) );
得到的数据
以第一组为例
训练集
0.5
p
0.5
a
测试集
0
p
1
a
f2[0]
f2[1]
迭代次数n
平均准确率p-ave
δ
耗时ms/次
耗时ms/199次
耗时 min/199
最大准确率p-max
0.514954
0.499634
178.9899
0
0.5
2.226131
444
0.0074
0
0.415923
0.588426
3992.095
0.876017
0.4
9.457286
1883
0.031383
0.951952
0.285787
0.717031
4830.568
0.899497
0.3
10
1992
0.0332
0.964965
0.185407
0.819002
5277.367
0.911172
0.2
10.07538
2006
0.033433
0.986987
0.089429
0.911647
6136.231
0.916952
0.1
11.36181
2261
0.037683
0.975976
0.008561
0.991586
8374.784
0.929714
0.01
15.9799
3180
0.053
0.981982
8.58E-04
0.999156
11611.27
0.943104
0.001
20.94975
4169
0.069483
0.988989
8.50E-05
0.999915
17998.72
0.955176
1.00E-04
30.72362
6114
0.1019
0.994995
7.79E-05
0.999922
18298.15
0.957319
9.00E-05
31.28141
6225
0.10375
0.993994
6.89E-05
0.999932
18879.48
0.955689
8.00E-05
32.14573
6412
0.106867
0.991992
6.06E-05
0.99994
19210.73
0.954809
7.00E-05
33.1407
6595
0.109917
0.992993
5.15E-05
0.999949
19577.49
0.956263
6.00E-05
33.39698
6648
0.1108
0.995996
4.33E-05
0.999957
20861.04
0.956414
5.00E-05
36.85427
7335
0.12225
0.998999
3.46E-05
0.999966
21738.86
0.955815
4.00E-05
37.68342
7499
0.124983
0.992993
2.59E-05
0.999974
22956.84
0.956001
3.00E-05
39.17085
7810
0.130167
0.994995
1.72E-05
0.999983
25275.3
0.957319
2.00E-05
43.54271
8665
0.144417
0.98999
8.63E-06
0.999991
29969.81
0.959698
1.00E-05
51.23618
10196
0.169933
0.991992
7.73E-06
0.999992
30247.04
0.957556
9.00E-06
51.62312
10274
0.171233
0.993994
6.84E-06
0.999993
30677.38
0.958536
8.00E-06
52.26633
10401
0.17335
0.988989
6.02E-06
0.999994
31932.25
0.958884
7.00E-06
55.18593
10997
0.183283
0.991992
5.20E-06
0.999995
33557.89
0.959105
6.00E-06
57.37186
11433
0.19055
0.995996
4.32E-06
0.999996
33544.95
0.959467
5.00E-06
57.32663
11409
0.19015
0.990991
3.42E-06
0.999997
35471.9
0.961368
4.00E-06
60.07035
11954
0.199233
0.997998
2.56E-06
0.999997
38427.08
0.95912
3.00E-06
65.20101
12975
0.21625
0.994995
1.70E-06
0.999998
41693.27
0.959955
2.00E-06
71.00503
14146
0.235767
0.991992
8.60E-07
0.999999
49787.77
0.959155
1.00E-06
85.94472
17103
0.28505
0.991992
这个网络的测试集p与a的比例是0：1，也就全是a。
平均准确率p-ave，是199次收敛的平均值
最大准确率p-max，是199次收敛的最大值
将得到的平均准确率画成图
可见网络的平均准确率没有随着迭代的增加而增加而是趋于一个常数，将1e-5到1e-6的平均准确率平均为0.959284。
也就是在训练集p：a=1：1的情况下如果测试集p：a=0：1网络可以接近恒定正确的区分出95.9284%的a。
其他的数据
a
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
平均准确率p-ave
p
δ
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.5
0
0.101202
0.199697
0.29948
0.402669
0.501768
0.599791
0.700816
0.799377
0.900126
1
0.4
0.876017
0.883587
0.900599
0.910423
0.919412
0.932903
0.946625
0.953763
0.967928
0.978853
0.990277
0.3
0.899497
0.908054
0.918833
0.921112
0.928144
0.936147
0.942691
0.948823
0.956036
0.963396
0.969366
0.2
0.911172
0.914271
0.918848
0.923667
0.929392
0.93247
0.936565
0.940554
0.94664
0.949195
0.956092
0.1
0.916952
0.919437
0.921152
0.923713
0.926494
0.928758
0.930322
0.93325
0.936097
0.936132
0.938154
0.01
0.929714
0.933305
0.930363
0.926343
0.923622
0.92085
0.91823
0.914553
0.910252
0.913813
0.905795
0.001
0.943104
0.939648
0.932847
0.928411
0.919684
0.914462
0.906615
0.903748
0.895654
0.88993
0.884875
1.00E-04
0.955176
0.946504
0.937043
0.92917
0.921137
0.909729
0.904699
0.893582
0.884543
0.875262
0.862712
9.00E-05
0.957319
0.943386
0.937329
0.927656
0.91997
0.910896
0.90159
0.890292
0.888235
0.873024
0.871288
8.00E-05
0.955689
0.944869
0.937249
0.928366
0.919352
0.911746
0.900599
0.892928
0.882541
0.874754
0.863804
7.00E-05
0.954809
0.944512
0.936892
0.928708
0.919236
0.909608
0.901826
0.891102
0.887657
0.874221
0.868844
6.00E-05
0.956263
0.945876
0.938164
0.928008
0.920639
0.910071
0.900398
0.895443
0.885649
0.876298
0.866203
5.00E-05
0.956414
0.948868
0.936177
0.927782
0.91994
0.909015
0.901369
0.88912
0.881374
0.869221
0.863632
4.00E-05
0.955815
0.947148
0.937354
0.929422
0.920282
0.910735
0.901872
0.892546
0.884291
0.872777
0.866319
3.00E-05
0.956001
0.94741
0.938516
0.929533
0.919427
0.910383
0.901671
0.889533
0.88153
0.868622
0.861138
2.00E-05
0.957319
0.949688
0.939729
0.929507
0.919824
0.911313
0.899316
0.888703
0.882999
0.869679
0.867254
1.00E-05
0.959698
0.94669
0.938733
0.92908
0.91999
0.911504
0.90076
0.890841
0.885172
0.873054
0.864281
9.00E-06
0.957556
0.947938
0.939628
0.929754
0.920911
0.909965
0.900011
0.892118
0.878351
0.869654
0.863567
8.00E-06
0.958536
0.949155
0.938728
0.930247
0.921067
0.911711
0.899377
0.892888
0.881681
0.873034
0.865152
7.00E-06
0.958884
0.950498
0.939301
0.931037
0.919497
0.911439
0.902656
0.889694
0.880046
0.876751
0.866404
6.00E-06
0.959105
0.947898
0.940881
0.930222
0.91913
0.911681
0.901328
0.892023
0.882948
0.873089
0.859528
5.00E-06
0.959467
0.948179
0.939236
0.93078
0.92074
0.911012
0.901359
0.892878
0.888557
0.87319
0.862541
4.00E-06
0.961368
0.948662
0.941072
0.929985
0.92071
0.911821
0.901826
0.892586
0.881811
0.870805
0.863758
3.00E-06
0.95912
0.949266
0.939854
0.931751
0.922566
0.910302
0.901258
0.890665
0.884422
0.871077
0.867184
2.00E-06
0.959955
0.950131
0.941026
0.930795
0.923617
0.912631
0.903401
0.895061
0.880634
0.874804
0.862033
1.00E-06
0.959155
0.951977
0.942078
0.931872
0.922495
0.912908
0.904553
0.895856
0.888783
0.875519
0.86898
0.959284
0.949039
0.940054
0.930552
0.921072
0.911497
0.901653
0.892461
0.883241
0.873098
0.864343
也就是在训练集p：a=1：1的情况下如果测试集p：a=1：0网络可以接近恒定正确的区分出86.4343%的p。
将最后一行准确率的平均值画成图是一条直线
如果用pp和pa表示测试集中p与a的比例则
d2（p，a）-4-4-2-（2*k）,k∈{0,1}的平均准确率的表达式为
平均准确率p-ave
pa
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
pp
δ
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
实测值
0.959284
0.949039
0.940054
0.930552
0.921072
0.911497
0.901653
0.892461
0.883241
0.873098
0.864343
计算值0.959284*pa+0.864343*pp
0.959284
0.94979
0.940296
0.930802
0.921308
0.911814
0.902319
0.892825
0.883331
0.873837
0.864343
将实测值和计算值画成图
图像是高度重合的。
实验参数
学习率 0.1
权重初始化方式
Random rand1 =new Random();
int ti1=rand1.nextInt(98)+1;
tw[a][b]=xx*((double)ti1/100);
迭代次数n
δ
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.5
178.9899
175.2211
175.6633
176.7286
174.9698
175.2814
176.5377
177.4121
174.9196
170.8995
178.8392
0.4
3992.095
4101.196
4076.523
4155.457
4006.583
4086.05
4030.116
4109.563
4115.342
4017.347
4051.658
0.3
4830.568
4713.518
4719.558
4833.392
4784.146
4760.03
4862.226
4740.181
4760.95
4721.533
4796.181
0.2
5277.367
5272.432
5314.251
5231.136
5421.779
5363.698
5488.477
5407.196
5349.799
5316.603
5451.658
0.1
6136.231
5991.387
6012.482
6147.387
6021.839
6068.291
6023.296
6141.548
5938.312
6031.166
6081.799
0.01
8374.784
8274.935
8297.186
8344.181
8307.709
8312.714
8373.266
8323.095
8314.472
8184.02
8227.578
0.001
11611.27
11731.6
11651.79
11758.99
11852.64
11708.52
11769.97
11895.17
11630.32
11934.5
11658.61
1.00E-04
17998.72
17924.94
18386.23
17823.12
17950.76
18032.69
18349.92
18039.44
17969.95
18146.08
18020.46
9.00E-05
18298.15
18370.07
18272.76
18671.64
18371.74
18455.81
18306.11
18449.29
18655.23
18287.8
18260.02
8.00E-05
18879.48
18877.53
18968.96
18814.26
19029.53
18907.17
18780.89
18454.62
18774.82
18842.83
19365.37
7.00E-05
19210.73
19618.94
19058.27
18864.31
19699.54
19192.85
19353.83
19463.43
19908.6
19146.66
19051.94
6.00E-05
19577.49
20321.44
19950.06
20230.75
20016.31
19637.51
20324.79
20284.9
19755.79
19900.13
20186.96
5.00E-05
20861.04
20918.89
21342.23
20212.31
20355.08
21041.91
20345.64
20506.18
20452
20773.77
21040.99
4.00E-05
21738.86
21696.41
22189.29
21664.93
21334.22
21458.21
21949.56
21906.86
22221.49
21510.16
22071.54
3.00E-05
22956.84
23914.19
23013.26
23628.01
23121.98
23419.3
22980.68
23162.81
23015.72
22580.6
23027.16
2.00E-05
25275.3
25512.2
25625.51
24909.99
25287.12
25467.01
25387.27
25386.75
25443.69
25225.02
25235.65
1.00E-05
29969.81
29470.97
29902.99
28973.61
29675.12
28929.45
30018.27
29183.94
29146.75
29139.24
29858.89
9.00E-06
30247.04
30513.24
30112.32
29809.58
30146.41
30596.92
30042.71
29992.44
30192.51
29208.69
29416.11
8.00E-06
30677.38
31836.22
31413.26
30242.05
31306.97
31467.45
31193.98
30287.49
31619.56
30228.18
30564.27
7.00E-06
31932.25
30926.25
32104.82
32418.75
31242.93
32308.31
32049.32
31156.3
31521.76
31077.35
31670.24
6.00E-06
33557.89
33163.54
32469.63
33315.13
32843.59
33142.06
33314.14
33318.47
32457.08
32820.7
32734.45
5.00E-06
33544.95
32903.76
34470.5
34408.14
34360.62
34622.05
33135.32
35836.14
34892.76
34875.83
33449.26
4.00E-06
35471.9
36459.97
35645.53
36694.53
36231.65
37457.67
34921.27
36361.22
36031.88
36771.91
35864.78
3.00E-06
38427.08
38564.55
38165.17
38486.48
38178.81
39145.41
37738.74
38041.16
38541.94
38024.93
38843.03
2.00E-06
41693.27
40956.78
42375.36
41608.88
42350.46
41515.22
42446.08
41794.77
41538.48
42297.98
41138.88
1.00E-06
49787.77
48948.52
49260.11
48418.2
48665.55
48841.56
46846.71
48042.7
50248.82
49040.7
49087.97
训练集
0.5
p
0.5
a
测试集
0
p
1
a
f2[0]
f2[1]
迭代次数n
平均准确率p-ave
δ
耗时ms/次
耗时ms/199次
耗时 min/199
最大准确率p-max
0.514954
0.499634
178.9899
0
0.5
2.226131
444
0.0074
0
0.415923
0.588426
3992.095
0.876017
0.4
9.457286
1883
0.031383
0.951952
0.285787
0.717031
4830.568
0.899497
0.3
10
1992
0.0332
0.964965
0.185407
0.819002
5277.367
0.911172
0.2
10.07538
2006
0.033433
0.986987
0.089429
0.911647
6136.231
0.916952
0.1
11.36181
2261
0.037683
0.975976
0.008561
0.991586
8374.784
0.929714
0.01
15.9799
3180
0.053
0.981982
8.58E-04
0.999156
11611.27
0.943104
0.001
20.94975
4169
0.069483
0.988989
8.50E-05
0.999915
17998.72
0.955176
1.00E-04
30.72362
6114
0.1019
0.994995
7.79E-05
0.999922
18298.15
0.957319
9.00E-05
31.28141
6225
0.10375
0.993994
6.89E-05
0.999932
18879.48
0.955689
8.00E-05
32.14573
6412
0.106867
0.991992
6.06E-05
0.99994
19210.73
0.954809
7.00E-05
33.1407
6595
0.109917
0.992993
5.15E-05
0.999949
19577.49
0.956263
6.00E-05
33.39698
6648
0.1108
0.995996
4.33E-05
0.999957
20861.04
0.956414
5.00E-05
36.85427
7335
0.12225
0.998999
3.46E-05
0.999966
21738.86
0.955815
4.00E-05
37.68342
7499
0.124983
0.992993
2.59E-05
0.999974
22956.84
0.956001
3.00E-05
39.17085
7810
0.130167
0.994995
1.72E-05
0.999983
25275.3
0.957319
2.00E-05
43.54271
8665
0.144417
0.98999
8.63E-06
0.999991
29969.81
0.959698
1.00E-05
51.23618
10196
0.169933
0.991992
7.73E-06
0.999992
30247.04
0.957556
9.00E-06
51.62312
10274
0.171233
0.993994
6.84E-06
0.999993
30677.38
0.958536
8.00E-06
52.26633
10401
0.17335
0.988989
6.02E-06
0.999994
31932.25
0.958884
7.00E-06
55.18593
10997
0.183283
0.991992
5.20E-06
0.999995
33557.89
0.959105
6.00E-06
57.37186
11433
0.19055
0.995996
4.32E-06
0.999996
33544.95
0.959467
5.00E-06
57.32663
11409
0.19015
0.990991
3.42E-06
0.999997
35471.9
0.961368
4.00E-06
60.07035
11954
0.199233
0.997998
2.56E-06
0.999997
38427.08
0.95912
3.00E-06
65.20101
12975
0.21625
0.994995
1.70E-06
0.999998
41693.27
0.959955
2.00E-06
71.00503
14146
0.235767
0.991992
8.60E-07
0.999999
49787.77
0.959155
1.00E-06
85.94472
17103
0.28505
0.991992
训练集
0.5
p
0.5
a
测试集
0.1
p
0.9
a
f2[0]
f2[1]
迭代次数n
平均准确率p-ave
δ
耗时ms/次
耗时ms/199次
耗时 min/199
最大准确率p-max
0.514946
0.499667
175.2211
0.101202
0.5
2.060302
412
0.006867
0.127127
0.412703
0.591256
4101.196
0.883587
0.4
9.417085
1906
0.031767
0.974975
0.285532
0.718176
4713.518
0.908054
0.3
8.849246
1792
0.029867
0.955956
0.184483
0.818452
5272.432
0.914271
0.2
10.0603
2003
0.033383
0.964965
0.089488
0.91179
5991.387
0.919437
0.1
11.44221
2277
0.03795
0.962963
0.008507
0.99162
8274.935
0.933305
0.01
16.43719
3286
0.054767
0.966967
8.55E-04
0.999162
11731.6
0.939648
0.001
21.16583
4230
0.0705
0.971972
8.48E-05
0.999916
17924.94
0.946504
1.00E-04
30.92462
6155
0.102583
0.972973
7.76E-05
0.999923
18370.07
0.943386
9.00E-05
32.36683
6441
0.10735
0.970971
6.94E-05
0.999931
18877.53
0.944869
8.00E-05
32.68342
6505
0.108417
0.973974
6.02E-05
0.999941
19618.94
0.944512
7.00E-05
34.26131
6850
0.114167
0.96997
5.23E-05
0.999948
20321.44
0.945876
6.00E-05
35.48744
7063
0.117717
0.970971
4.24E-05
0.999958
20918.89
0.948868
5.00E-05
35.77387
7119
0.11865
0.96997
3.45E-05
0.999966
21696.41
0.947148
4.00E-05
37.32663
7428
0.1238
0.97998
2.57E-05
0.999975
23914.19
0.94741
3.00E-05
41.29146
8217
0.13695
0.974975
1.70E-05
0.999983
25512.2
0.949688
2.00E-05
43.47236
8683
0.144717
0.971972
8.66E-06
0.999991
29470.97
0.94669
1.00E-05
50.72362
10094
0.168233
0.970971
7.83E-06
0.999992
30513.24
0.947938
9.00E-06
52.77387
10502
0.175033
0.971972
7.00E-06
0.999993
31836.22
0.949155
8.00E-06
55.95477
11135
0.185583
0.974975
6.03E-06
0.999994
30926.25
0.950498
7.00E-06
53.46734
10655
0.177583
0.974975
5.20E-06
0.999995
33163.54
0.947898
6.00E-06
56.86935
11319
0.18865
0.978979
4.29E-06
0.999996
32903.76
0.948179
5.00E-06
56.14573
11173
0.186217
0.974975
3.46E-06
0.999997
36459.97
0.948662
4.00E-06
62.64824
12467
0.207783
0.976977
2.61E-06
0.999997
38564.55
0.949266
3.00E-06
66.58794
13252
0.220867
0.973974
1.71E-06
0.999998
40956.78
0.950131
2.00E-06
69.95477
13923
0.23205
0.976977
8.55E-07
0.999999
48948.52
0.951977
1.00E-06
83.75879
16683
0.27805
0.976977
训练集
0.5
p
0.5
a
测试集
0.2
p
0.8
a
f2[0]
f2[1]
迭代次数n
平均准确率p-ave
δ
耗时ms/次
耗时ms/199次
耗时 min/199
最大准确率p-max
0.515768
0.499665
175.6633
0.199697
0.5
1.819095
409
0.006817
0.241241
0.405578
0.598913
4076.523
0.900599
0.4
9.452261
1898
0.031633
0.955956
0.285142
0.71966
4719.558
0.918833
0.3
9.331658
1857
0.03095
0.960961
0.18482
0.819326
5314.251
0.918848
0.2
10.21608
2033
0.033883
0.965966
0.089624
0.913356
6012.482
0.921152
0.1
11.44221
2277
0.03795
0.954955
0.008701
0.991542
8297.186
0.930363
0.01
16.62312
3308
0.055133
0.956957
8.51E-04
0.999157
11651.79
0.932847
0.001
20.77387
4134
0.0689
0.957958
8.60E-05
0.999915
18386.23
0.937043
1.00E-04
32.02513
6391
0.106517
0.95996
7.88E-05
0.999922
18272.76
0.937329
9.00E-05
31.78894
6328
0.105467
0.957958
7.00E-05
0.999931
18968.96
0.937249
8.00E-05
33.07035
6582
0.1097
0.960961
5.96E-05
0.999941
19058.27
0.936892
7.00E-05
33.44724
6672
0.1112
0.95996
5.22E-05
0.999948
19950.06
0.938164
6.00E-05
35.60302
7086
0.1181
0.95996
4.33E-05
0.999957
21342.23
0.936177
5.00E-05
37.52764
7484
0.124733
0.960961
3.44E-05
0.999966
22189.29
0.937354
4.00E-05
38.31658
7625
0.127083
0.963964
2.63E-05
0.999974
23013.26
0.938516
3.00E-05
39.90452
7942
0.132367
0.962963
1.72E-05
0.999983
25625.51
0.939729
2.00E-05
44.37186
8831
0.147183
0.95996
8.58E-06
0.999991
29902.99
0.938733
1.00E-05
51.89447
10329
0.17215
0.95996
7.72E-06
0.999992
30112.32
0.939628
9.00E-06
52.46231
10441
0.174017
0.961962
6.87E-06
0.999993
31413.26
0.938728
8.00E-06
56.63317
11271
0.18785
0.961962
6.03E-06
0.999994
32104.82
0.939301
7.00E-06
54.90452
10962
0.1827
0.960961
5.01E-06
0.999995
32469.63
0.940881
6.00E-06
54.08543
10763
0.179383
0.965966
4.25E-06
0.999996
34470.5
0.939236
5.00E-06
57.83417
11509
0.191817
0.965966
3.40E-06
0.999997
35645.53
0.941072
4.00E-06
60.23116
11986
0.199767
0.963964
2.60E-06
0.999997
38165.17
0.939854
3.00E-06
64.25628
12787
0.213117
0.961962
1.71E-06
0.999998
42375.36
0.941026
2.00E-06
72.03015
14349
0.23915
0.958959
8.69E-07
0.999999
49260.11
0.942078
1.00E-06
82.36683
16391
0.273183
0.962963
训练集
0.5
p
0.5
a
测试集
0.3
p
0.7
a
f2[0]
f2[1]
迭代次数n
平均准确率p-ave
δ
耗时ms/次
耗时ms/199次
耗时 min/199
最大准确率p-max
0.515474
0.499642
176.7286
0.29948
0.5
1.959799
422
0.007033
0.335335
0.402981
0.602061
4155.457
0.910423
0.4
9.035176
1798
0.029967
0.965966
0.288652
0.717348
4833.392
0.921112
0.3
9.201005
1831
0.030517
0.956957
0.185627
0.819707
5231.136
0.923667
0.2
10.07035
2005
0.033417
0.956957
0.089602
0.912096
6147.387
0.923713
0.1
11.65829
2320
0.038667
0.953954
0.008618
0.991644
8344.181
0.926343
0.01
15.45226
3090
0.0515
0.950951
8.50E-04
0.999161
11758.99
0.928411
0.001
20.22613
4025
0.067083
0.952953
8.45E-05
0.999916
17823.12
0.92917
1.00E-04
30.46734
6063
0.10105
0.948949
7.67E-05
0.999925
18671.64
0.927656
9.00E-05
33.24121
6616
0.110267
0.94995
6.92E-05
0.999931
18814.26
0.928366
8.00E-05
33.58794
6685
0.111417
0.94995
6.02E-05
0.99994
18864.31
0.928708
7.00E-05
33.33668
6649
0.110817
0.954955
5.12E-05
0.999949
20230.75
0.928008
6.00E-05
35.36181
7037
0.117283
0.94995
4.24E-05
0.999958
20212.31
0.927782
5.00E-05
35.52261
7101
0.11835
0.94995
3.46E-05
0.999966
21664.93
0.929422
4.00E-05
37.82915
7528
0.125467
0.948949
2.61E-05
0.999974
23628.01
0.929533
3.00E-05
41.34171
8227
0.137117
0.948949
1.71E-05
0.999983
24909.99
0.929507
2.00E-05
43.52261
8661
0.14435
0.952953
8.61E-06
0.999991
28973.61
0.92908
1.00E-05
50.82412
10115
0.168583
0.948949
7.76E-06
0.999992
29809.58
0.929754
9.00E-06
52.65327
10478
0.174633
0.952953
6.95E-06
0.999993
30242.05
0.930247
8.00E-06
52.93467
10534
0.175567
0.953954
6.06E-06
0.999994
32418.75
0.931037
7.00E-06
56.50251
11245
0.187417
0.955956
5.12E-06
0.999995
33315.13
0.930222
6.00E-06
57.41709
11427
0.19045
0.950951
4.30E-06
0.999996
34408.14
0.93078
5.00E-06
59.24623
11806
0.196767
0.953954
3.49E-06
0.999997
36694.53
0.929985
4.00E-06
63.9397
12724
0.212067
0.950951
2.62E-06
0.999997
38486.48
0.931751
3.00E-06
67.08543
13366
0.222767
0.951952
1.71E-06
0.999998
41608.88
0.930795
2.00E-06
71.47739
14225
0.237083
0.948949
8.64E-07
0.999999
48418.2
0.931872
1.00E-06
83.41206
16599
0.27665
0.94995
训练集
0.5
p
0.5
a
测试集
0.4
p
0.6
a
f2[0]
f2[1]
迭代次数n
平均准确率p-ave
δ
耗时ms/次
耗时ms/199次
耗时 min/199
最大准确率p-max
0.515551
0.49964
174.9698
0.402669
0.5
2.256281
450
0.0075
0.565566
0.409463
0.592821
4006.583
0.919412
0.4
9.090452
1811
0.030183
0.971972
0.287786
0.716621
4784.146
0.928144
0.3
9.366834
1866
0.0311
0.963964
0.186042
0.819252
5421.779
0.929392
0.2
10.51759
2094
0.0349
0.956957
0.089566
0.911688
6021.839
0.926494
0.1
11.61809
2314
0.038567
0.951952
0.00859
0.991607
8307.709
0.923622
0.01
16.71859
3329
0.055483
0.951952
8.50E-04
0.99916
11852.64
0.919684
0.001
20.45226
4070
0.067833
0.942943
8.58E-05
0.999916
17950.76
0.921137
1.00E-04
30.26633
6023
0.100383
0.94995
7.64E-05
0.999925
18371.74
0.91997
9.00E-05
31.56281
6283
0.104717
0.944945
6.75E-05
0.999933
19029.53
0.919352
8.00E-05
32.43719
6455
0.107583
0.944945
5.99E-05
0.999941
19699.54
0.919236
7.00E-05
34.17085
6816
0.1136
0.944945
5.14E-05
0.999949
20016.31
0.920639
6.00E-05
35.60302
7089
0.11815
0.945946
4.26E-05
0.999958
20355.08
0.91994
5.00E-05
35.63317
7097
0.118283
0.942943
3.46E-05
0.999966
21334.22
0.920282
4.00E-05
37.80905
7527
0.12545
0.945946
2.58E-05
0.999974
23121.98
0.919427
3.00E-05
39.78392
7949
0.132483
0.940941
1.73E-05
0.999983
25287.12
0.919824
2.00E-05
42.67337
8499
0.14165
0.947948
8.63E-06
0.999991
29675.12
0.91999
1.00E-05
49.61809
9882
0.1647
0.941942
7.86E-06
0.999992
30146.41
0.920911
9.00E-06
50.54774
10066
0.167767
0.94995
6.90E-06
0.999993
31306.97
0.921067
8.00E-06
52.32663
10415
0.173583
0.94995
5.92E-06
0.999994
31242.93
0.919497
7.00E-06
52.58291
10467
0.17445
0.948949
5.16E-06
0.999995
32843.59
0.91913
6.00E-06
55.69849
11088
0.1848
0.948949
4.30E-06
0.999996
34360.62
0.92074
5.00E-06
57.33166
11411
0.190183
0.943944
3.45E-06
0.999997
36231.65
0.92071
4.00E-06
60.28141
12003
0.20005
0.947948
2.64E-06
0.999997
38178.81
0.922566
3.00E-06
63.62312
12667
0.211117
0.952953
1.73E-06
0.999998
42350.46
0.923617
2.00E-06
70.25628
13987
0.233117
0.946947
8.61E-07
0.999999
48665.55
0.922495
1.00E-06
81.49749
16226
0.270433
0.945946
训练集
0.5
p
0.5
a
测试集
0.5
p
0.5
a
f2[0]
f2[1]
迭代次数n
平均准确率p-ave
δ
耗时ms/次
耗时ms/199次
耗时 min/199
最大准确率p-max
0.515364
0.49962
175.2814
0.501768
0.5
2.20603
439
0.007317
0.551552
0.407196
0.597051
4086.05
0.932903
0.4
9.020101
1811
0.030183
0.974975
0.285219
0.719742
4760.03
0.936147
0.3
9.291457
1849
0.030817
0.95996
0.185463
0.818207
5363.698
0.93247
0.2
10.19095
2028
0.0338
0.954955
0.088918
0.911725
6068.291
0.928758
0.1
11.57286
2303
0.038383
0.955956
0.008788
0.991436
8312.714
0.92085
0.01
15.46734
3078
0.0513
0.947948
8.52E-04
0.999165
11708.52
0.914462
0.001
21.22613
4224
0.0704
0.944945
8.48E-05
0.999916
18032.69
0.909729
1.00E-04
30.88442
6178
0.102967
0.93994
7.58E-05
0.999925
18455.81
0.910896
9.00E-05
31.38191
6245
0.104083
0.93994
6.98E-05
0.999931
18907.17
0.911746
8.00E-05
32.30151
6428
0.107133
0.941942
6.00E-05
0.999941
19192.85
0.909608
7.00E-05
33.25628
6618
0.1103
0.93994
5.10E-05
0.99995
19637.51
0.910071
6.00E-05
33.23618
6614
0.110233
0.938939
4.33E-05
0.999957
21041.91
0.909015
5.00E-05
35.39196
7043
0.117383
0.948949
3.45E-05
0.999966
21458.21
0.910735
4.00E-05
36.67839
7299
0.12165
0.947948
2.56E-05
0.999975
23419.3
0.910383
3.00E-05
39.48241
7888
0.131467
0.936937
1.73E-05
0.999983
25467.01
0.911313
2.00E-05
43.12563
8582
0.143033
0.946947
8.62E-06
0.999991
28929.45
0.911504
1.00E-05
49.22613
9827
0.163783
0.947948
7.76E-06
0.999992
30596.92
0.909965
9.00E-06
51.05528
10192
0.169867
0.942943
7.03E-06
0.999993
31467.45
0.911711
8.00E-06
53.44724
10651
0.177517
0.943944
6.02E-06
0.999994
32308.31
0.911439
7.00E-06
53.8794
10722
0.1787
0.950951
5.21E-06
0.999995
33142.06
0.911681
6.00E-06
55.39196
11039
0.183983
0.946947
4.35E-06
0.999996
34622.05
0.911012
5.00E-06
58.53266
11648
0.194133
0.940941
3.42E-06
0.999997
37457.67
0.911821
4.00E-06
63.24121
12585
0.20975
0.937938
2.53E-06
0.999997
39145.41
0.910302
3.00E-06
65.72864
13080
0.218
0.942943
1.72E-06
0.999998
41515.22
0.912631
2.00E-06
69.12563
13787
0.229783
0.93994
8.68E-07
0.999999
48841.56
0.912908
1.00E-06
81.88945
16311
0.27185
0.946947
训练集
0.5
p
0.5
a
测试集
0.6
p
0.4
a
f2[0]
f2[1]
迭代次数n
平均准确率p-ave
δ
耗时ms/次
耗时ms/199次
耗时 min/199
最大准确率p-max
0.515048
0.499661
176.5377
0.599791
0.5
2.19598
437
0.007283
0.642643
0.405811
0.599618
4030.116
0.946625
0.4
8.728643
1769
0.029483
0.974975
0.287959
0.717861
4862.226
0.942691
0.3
-503.367
-100154
-1.66923
0.968969
0.187403
0.816228
5488.477
0.936565
0.2
10.29146
2048
0.034133
0.957958
0.089773
0.912661
6023.296
0.930322
0.1
11
2190
0.0365
0.957958
0.008599
0.991567
8373.266
0.91823
0.01
15.43719
3072
0.0512
0.953954
8.56E-04
0.999161
11769.97
0.906615
0.001
21.0804
4195
0.069917
0.940941
8.67E-05
0.999914
18349.92
0.904699
1.00E-04
450.9648
89742
1.4957
0.948949
7.72E-05
0.999924
18306.11
0.90159
9.00E-05
31.36683
6242
0.104033
0.93994
6.85E-05
0.999933
18780.89
0.900599
8.00E-05
32.02513
6373
0.106217
0.938939
5.99E-05
0.999941
19353.83
0.901826
7.00E-05
32.85427
6539
0.108983
0.936937
5.19E-05
0.999949
20324.79
0.900398
6.00E-05
34.42211
6850
0.114167
0.942943
4.29E-05
0.999958
20345.64
0.901369
5.00E-05
34.73367
6912
0.1152
0.951952
3.42E-05
0.999966
21949.56
0.901872
4.00E-05
37.62814
7504
0.125067
0.941942
2.60E-05
0.999974
22980.68
0.901671
3.00E-05
40.47236
8054
0.134233
0.941942
1.73E-05
0.999983
25387.27
0.899316
2.00E-05
43.0603
8585
0.143083
0.942943
8.60E-06
0.999991
30018.27
0.90076
1.00E-05
51.17085
10183
0.169717
0.948949
7.72E-06
0.999992
30042.71
0.900011
9.00E-06
51.28141
10220
0.170333
0.941942
6.86E-06
0.999993
31193.98
0.899377
8.00E-06
52.92462
10532
0.175533
0.935936
5.99E-06
0.999994
32049.32
0.902656
7.00E-06
54.29146
10819
0.180317
0.942943
5.12E-06
0.999995
33314.14
0.901328
6.00E-06
57.77387
11497
0.191617
0.940941
4.29E-06
0.999996
33135.32
0.901359
5.00E-06
56.50754
11260
0.187667
0.945946
3.40E-06
0.999997
34921.27
0.901826
4.00E-06
59.0402
11779
0.196317
0.944945
2.60E-06
0.999997
37738.74
0.901258
3.00E-06
63.25628
12588
0.2098
0.93994
1.72E-06
0.999998
42446.08
0.903401
2.00E-06
72.0603
14340
0.239
0.944945
8.57E-07
0.999999
46846.71
0.904553
1.00E-06
80.15075
15950
0.265833
0.941942
训练集
0.5
p
0.5
a
测试集
0.7
p
0.3
a
f2[0]
f2[1]
迭代次数n
平均准确率p-ave
δ
耗时ms/次
耗时ms/199次
耗时 min/199
最大准确率p-max
0.514892
0.499679
177.4121
0.700816
0.5
2.120603
422
0.007033
0.741742
0.416277
0.588445
4109.563
0.953763
0.4
9.19598
1830
0.0305
0.977978
0.289018
0.714385
4740.181
0.948823
0.3
9.281407
1847
0.030783
0.973974
0.185205
0.818787
5407.196
0.940554
0.2
10.13568
2017
0.033617
0.967968
0.089925
0.913099
6141.548
0.93325
0.1
11.48241
2285
0.038083
0.966967
0.008521
0.991705
8323.095
0.914553
0.01
17.06533
3412
0.056867
0.954955
8.59E-04
0.999152
11895.17
0.903748
0.001
20.65829
4111
0.068517
0.945946
8.55E-05
0.999916
18039.44
0.893582
1.00E-04
31.88945
6362
0.106033
0.958959
7.67E-05
0.999924
18449.29
0.890292
9.00E-05
31.45729
6275
0.104583
0.947948
6.87E-05
0.999932
18454.62
0.892928
8.00E-05
31.54774
6278
0.104633
0.938939
6.13E-05
0.99994
19463.43
0.891102
7.00E-05
33.33166
6648
0.1108
0.946947
5.13E-05
0.999949
20284.9
0.895443
6.00E-05
34.80905
6928
0.115467
0.953954
4.31E-05
0.999957
20506.18
0.88912
5.00E-05
35.17588
7016
0.116933
0.942943
3.49E-05
0.999965
21906.86
0.892546
4.00E-05
37.39698
7458
0.1243
0.93994
2.57E-05
0.999975
23162.81
0.889533
3.00E-05
39.74874
7925
0.132083
0.942943
1.71E-05
0.999983
25386.75
0.888703
2.00E-05
43.89447
8735
0.145583
0.938939
8.59E-06
0.999991
29183.94
0.890841
1.00E-05
50.56784
10063
0.167717
0.938939
7.68E-06
0.999992
29992.44
0.892118
9.00E-06
50.74874
10115
0.168583
0.943944
6.96E-06
0.999993
30287.49
0.892888
8.00E-06
51.13568
10191
0.16985
0.93994
5.89E-06
0.999994
31156.3
0.889694
7.00E-06
52.75377
10498
0.174967
0.944945
5.19E-06
0.999995
33318.47
0.892023
6.00E-06
56.57286
11273
0.187883
0.948949
4.35E-06
0.999996
35836.14
0.892878
5.00E-06
60.17085
11974
0.199567
0.941942
3.44E-06
0.999997
36361.22
0.892586
4.00E-06
62.66332
12486
0.2081
0.946947
2.53E-06
0.999997
38041.16
0.890665
3.00E-06
64.01005
12754
0.212567
0.93994
1.73E-06
0.999998
41794.77
0.895061
2.00E-06
71.06533
14144
0.235733
0.938939
8.62E-07
0.999999
48042.7
0.895856
1.00E-06
80.42211
16004
0.266733
0.942943
训练集
0.5
p
0.5
a
测试集
0.8
p
0.2
a
f2[0]
f2[1]
迭代次数n
平均准确率p-ave
δ
耗时ms/次
耗时ms/199次
耗时 min/199
最大准确率p-max
0.51531
0.499628
174.9196
0.799377
0.5
1.959799
406
0.006767
0.835836
0.407105
0.597558
4115.342
0.967928
0.4
9.311558
1855
0.030917
0.986987
0.287128
0.715635
4760.95
0.956036
0.3
9.175879
1826
0.030433
0.978979
0.18441
0.820183
5349.799
0.94664
0.2
10.1809
2026
0.033767
0.974975
0.089271
0.911917
5938.312
0.936097
0.1
11.12563
2214
0.0369
0.970971
0.008534
0.991737
8314.472
0.910252
0.01
15.31156
3047
0.050783
0.972973
8.60E-04
0.99916
11630.32
0.895654
0.001
20.23116
4026
0.0671
0.950951
8.59E-05
0.999915
17969.95
0.884543
1.00E-04
30.97487
6164
0.102733
0.951952
7.70E-05
0.999923
18655.23
0.888235
9.00E-05
32.68844
6505
0.108417
0.947948
6.80E-05
0.999933
18774.82
0.882541
8.00E-05
31.8794
6344
0.105733
0.945946
6.07E-05
0.99994
19908.6
0.887657
7.00E-05
34.21106
6841
0.114017
0.942943
5.11E-05
0.999949
19755.79
0.885649
6.00E-05
33.36683
6672
0.1112
0.943944
4.36E-05
0.999957
20452
0.881374
5.00E-05
34.92462
6965
0.116083
0.946947
3.46E-05
0.999966
22221.49
0.884291
4.00E-05
39.75377
7911
0.13185
0.947948
2.59E-05
0.999974
23015.72
0.88153
3.00E-05
39.38693
7838
0.130633
0.957958
1.73E-05
0.999983
25443.69
0.882999
2.00E-05
43.39196
8651
0.144183
0.94995
8.53E-06
0.999992
29146.75
0.885172
1.00E-05
50.31156
10012
0.166867
0.946947
7.67E-06
0.999992
30192.51
0.878351
9.00E-06
51.97487
10343
0.172383
0.941942
6.90E-06
0.999993
31619.56
0.881681
8.00E-06
53.47236
10641
0.17735
0.942943
6.09E-06
0.999994
31521.76
0.880046
7.00E-06
53.63819
10705
0.178417
0.942943
5.19E-06
0.999995
32457.08
0.882948
6.00E-06
55.30151
11005
0.183417
0.947948
4.34E-06
0.999996
34892.76
0.888557
5.00E-06
59.30151
11801
0.196683
0.942943
3.41E-06
0.999997
36031.88
0.881811
4.00E-06
61.28643
12196
0.203267
0.94995
2.59E-06
0.999997
38541.94
0.884422
3.00E-06
65.96482
13127
0.218783
0.941942
1.72E-06
0.999998
41538.48
0.880634
2.00E-06
70.19095
13983
0.23305
0.942943
8.66E-07
0.999999
50248.82
0.888783
1.00E-06
84.68342
16852
0.280867
0.955956
训练集
0.5
p
0.5
a
测试集
0.9
p
0.1
a
f2[0]
f2[1]
迭代次数n
平均准确率p-ave
δ
耗时ms/次
耗时ms/199次
耗时 min/199
最大准确率p-max
0.516515
0.49963
170.8995
0.900126
0.5
2.221106
442
0.007367
0.942943
0.410117
0.594254
4017.347
0.978853
0.4
9.477387
1901
0.031683
0.991992
0.287948
0.715286
4721.533
0.963396
0.3
9.221106
1835
0.030583
0.98999
0.186425
0.81875
5316.603
0.949195
0.2
10.06533
2003
0.033383
0.981982
0.089713
0.913228
6031.166
0.936132
0.1
11.01005
2191
0.036517
0.97998
0.008667
0.991602
8184.02
0.913813
0.01
15.10553
3006
0.0501
0.967968
8.47E-04
0.999165
11934.5
0.88993
0.001
20.75879
4131
0.06885
0.967968
8.44E-05
0.999916
18146.08
0.875262
1.00E-04
30.81407
6132
0.1022
0.966967
7.57E-05
0.999925
18287.8
0.873024
9.00E-05
31.19598
6208
0.103467
0.95996
6.81E-05
0.999932
18842.83
0.874754
8.00E-05
32.52764
6474
0.1079
0.970971
5.90E-05
0.999942
19146.66
0.874221
7.00E-05
33.21106
6609
0.11015
0.955956
5.21E-05
0.999949
19900.13
0.876298
6.00E-05
33.57286
6681
0.11135
0.94995
4.26E-05
0.999958
20773.77
0.869221
5.00E-05
35.47236
7074
0.1179
0.955956
3.45E-05
0.999966
21510.16
0.872777
4.00E-05
36.89447
7342
0.122367
0.943944
2.55E-05
0.999975
22580.6
0.868622
3.00E-05
38.20101
7602
0.1267
0.943944
1.70E-05
0.999983
25225.02
0.869679
2.00E-05
42.81407
8536
0.142267
0.955956
8.73E-06
0.999991
29139.24
0.873054
1.00E-05
49
9751
0.162517
0.950951
7.68E-06
0.999992
29208.69
0.869654
9.00E-06
49.03015
9757
0.162617
0.943944
6.90E-06
0.999993
30228.18
0.873034
8.00E-06
50.19095
9988
0.166467
0.938939
6.00E-06
0.999994
31077.35
0.876751
7.00E-06
52.10553
10385
0.173083
0.958959
5.23E-06
0.999995
32820.7
0.873089
6.00E-06
55.94472
11133
0.18555
0.946947
4.30E-06
0.999996
34875.83
0.87319
5.00E-06
58.28141
11598
0.1933
0.945946
3.41E-06
0.999997
36771.91
0.870805
4.00E-06
61.10553
12190
0.203167
0.947948
2.56E-06
0.999997
38024.93
0.871077
3.00E-06
63.25126
12587
0.209783
0.948949
1.73E-06
0.999998
42297.98
0.874804
2.00E-06
70.18593
13967
0.232783
0.961962
8.62E-07
0.999999
49040.7
0.875519
1.00E-06
82.47739
16413
0.27355
0.950951
训练集
0.5
p
0.5
a
测试集
1
p
0
a
f2[0]
f2[1]
迭代次数n
平均准确率p-ave
δ
耗时ms/次
耗时ms/199次
耗时 min/199
最大准确率p-max
0.514793
0.499623
178.8392
1
0.5
1.964824
391
0.006517
1
0.408172
0.596096
4051.658
0.990277
0.4
9.874372
1965
0.03275
1
0.285396
0.7193
4796.181
0.969366
0.3
10.18593
2043
0.03405
1
0.183454
0.819462
5451.658
0.956092
0.2
10.19095
2028
0.0338
0.994995
0.088923
0.912634
6081.799
0.938154
0.1
11.08543
2206
0.036767
0.995996
0.008548
0.991621
8227.578
0.905795
0.01
15.42714
3086
0.051433
0.97998
8.61E-04
0.999152
11658.61
0.884875
0.001
20.33668
4077
0.06795
0.963964
8.50E-05
0.999916
18020.46
0.862712
1.00E-04
31.12563
6194
0.103233
0.964965
7.70E-05
0.999924
18260.02
0.871288
9.00E-05
31.11558
6207
0.10345
0.971972
6.93E-05
0.999932
19365.37
0.863804
8.00E-05
32.24623
6448
0.107467
0.978979
6.03E-05
0.99994
19051.94
0.868844
7.00E-05
32.05025
6378
0.1063
0.948949
5.22E-05
0.999948
20186.96
0.866203
6.00E-05
34.21106
6808
0.113467
0.956957
4.33E-05
0.999957
21040.99
0.863632
5.00E-05
35.67839
7100
0.118333
0.955956
3.46E-05
0.999966
22071.54
0.866319
4.00E-05
36.88945
7357
0.122617
0.966967
2.59E-05
0.999974
23027.16
0.861138
3.00E-05
38.36181
7634
0.127233
0.941942
1.72E-05
0.999983
25235.65
0.867254
2.00E-05
43.06533
8570
0.142833
0.945946
8.70E-06
0.999991
29858.89
0.864281
1.00E-05
50.26633
10035
0.16725
0.942943
7.57E-06
0.999992
29416.11
0.863567
9.00E-06
48.88945
9729
0.16215
0.952953
6.94E-06
0.999993
30564.27
0.865152
8.00E-06
51.44221
10237
0.170617
0.950951
6.03E-06
0.999994
31670.24
0.866404
7.00E-06
52.32663
10428
0.1738
0.951952
5.17E-06
0.999995
32734.45
0.859528
6.00E-06
54.27136
10800
0.18
0.952953
4.23E-06
0.999996
33449.26
0.862541
5.00E-06
55.37186
11034
0.1839
0.943944
3.43E-06
0.999997
35864.78
0.863758
4.00E-06
60.89447
12133
0.202217
0.950951
2.57E-06
0.999997
38843.03
0.867184
3.00E-06
64.20101
12776
0.212933
0.945946
1.69E-06
0.999998
41138.88
0.862033
2.00E-06
68.14573
13561
0.226017
0.940941
8.67E-07
0.999999
49087.97
0.86898
1.00E-06
81.9196
16302
0.2717
0.95996
