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
制作一个二分类网络用来分类c和a，通过改变测试集c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c：a=1：1，
测试集c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
二分类吸引子和鞍点
制作一个4*4*2的网络向这个的左侧输入吸引子，并让左侧网络向1，0收敛；向右侧网络输入鞍点让右侧向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
c
1
判断是否达到收敛
a
2
判断是否达到收敛
梯度下降
c
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（c，a）-4-4-2-（2*k）,k∈{0,1}
得到的数据
以第一组为例
训练集
0.5
c
0.5
a
测试集
0
c
10
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
0.499549
0.516406
181.2261
1
0.5
1.984925
410
0.006833
1
0.608557
0.395061
4200.583
0.972752
0.4
10.57286
2119
0.035317
1
0.719274
0.285924
4900.789
0.940176
0.3
9.477387
1886
0.031433
0.984985
0.81995
0.185329
5614.688
0.919266
0.2
10.49246
2088
0.0348
0.988989
0.913052
0.088778
6392.407
0.897858
0.1
11.28141
2260
0.037667
0.967968
0.991559
0.008561
9046.377
0.860297
0.01
16.69347
3323
0.055383
0.957958
0.999172
8.40E-04
13557.47
0.845785
0.001
23.15075
4607
0.076783
0.947948
0.999918
8.24E-05
22123.3
0.841002
1.00E-04
38.13065
7589
0.126483
0.941942
0.999927
7.41E-05
22794.29
0.838778
9.00E-05
39.96482
7953
0.13255
0.923924
0.999935
6.59E-05
23001.94
0.842139
8.00E-05
39.59799
7880
0.131333
0.925926
0.999944
5.69E-05
23585.63
0.835926
7.00E-05
40.05528
7972
0.132867
0.918919
0.99995
5.01E-05
25008.55
0.844488
6.00E-05
42.19095
8396
0.139933
0.933934
0.999959
4.14E-05
25913.07
0.843602
5.00E-05
44.65327
8901
0.14835
0.930931
0.999967
3.33E-05
27440.44
0.840413
4.00E-05
47.28141
9409
0.156817
0.935936
0.999976
2.44E-05
29798.31
0.832471
3.00E-05
51.21106
10206
0.1701
0.937938
0.999984
1.57E-05
32360.2
0.830001
2.00E-05
55.69849
11084
0.184733
0.914915
0.999992
8.29E-06
40696.32
0.828668
1.00E-05
68.1407
13560
0.226
0.916917
0.999993
7.35E-06
40860.31
0.820514
9.00E-06
69.38693
13808
0.230133
0.905906
0.999993
6.54E-06
43086.26
0.822501
8.00E-06
72.56784
14441
0.240683
0.90991
0.999994
5.59E-06
45369.46
0.817717
7.00E-06
76.98492
15352
0.255867
0.91992
0.999995
4.88E-06
48723.63
0.817586
6.00E-06
83.66834
16651
0.277517
0.908909
0.999996
4.11E-06
51922.08
0.810378
5.00E-06
86.40201
17210
0.286833
0.894895
0.999997
3.24E-06
57409.06
0.805051
4.00E-06
96.86935
19277
0.321283
0.886887
0.999998
2.37E-06
67244.38
0.787682
3.00E-06
113.3116
22549
0.375817
0.883884
0.999998
1.57E-06
87439.24
0.763432
2.00E-06
147.5025
29370
0.4895
0.873874
0.999999
7.30E-07
135984.2
0.724051
1.00E-06
228.2261
45417
0.75695
0.874875
这个网络的测试集c与a的比例是0：1，也就全是a。
将得到的平均准确率画成图
网络的准确率没有随着迭代的增加而增加，将1e-5到1e-6的平均准确率平均为0.799758。
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
c
δ
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.5
1
0.899814
0.800947
0.699901
0.600656
0.500707
0.401804
0.299521
0.201734
0.102912
6.09E-04
0.4
0.972752
0.960589
0.947369
0.93407
0.919779
0.907405
0.89401
0.8816
0.863985
0.855051
0.84492
0.3
0.940176
0.928129
0.920393
0.913049
0.904261
0.895856
0.889005
0.876962
0.871897
0.864402
0.856746
0.2
0.919266
0.908833
0.902908
0.895871
0.890448
0.88653
0.88076
0.874825
0.868648
0.860725
0.857264
0.1
0.897858
0.890167
0.884533
0.882762
0.879337
0.874789
0.87154
0.865982
0.862174
0.856802
0.857128
0.01
0.860297
0.858492
0.861545
0.861354
0.860916
0.863416
0.86322
0.863959
0.863783
0.865841
0.866193
0.001
0.845785
0.849136
0.853804
0.856238
0.857425
0.862772
0.865655
0.868713
0.873124
0.877003
0.876867
1.00E-04
0.841002
0.847118
0.853859
0.858099
0.862878
0.868139
0.872898
0.877858
0.884749
0.886022
0.893491
9.00E-05
0.838778
0.847712
0.855489
0.85398
0.863869
0.867088
0.874558
0.878406
0.883321
0.887737
0.896248
8.00E-05
0.842139
0.84985
0.850147
0.857531
0.862667
0.869241
0.872938
0.879306
0.885262
0.888532
0.893949
7.00E-05
0.835926
0.843215
0.853391
0.859095
0.861877
0.869422
0.874804
0.880363
0.88493
0.887994
0.895368
6.00E-05
0.844488
0.843839
0.850861
0.858331
0.864664
0.869216
0.874382
0.87991
0.886756
0.889281
0.897646
5.00E-05
0.843602
0.838819
0.851419
0.855725
0.863321
0.868698
0.876173
0.88232
0.887304
0.890514
0.897335
4.00E-05
0.840413
0.84147
0.850931
0.856711
0.864135
0.869442
0.876485
0.882244
0.886736
0.896303
0.901082
3.00E-05
0.832471
0.846424
0.849905
0.8591
0.864261
0.869251
0.877858
0.884784
0.889709
0.896992
0.901761
2.00E-05
0.830001
0.838784
0.848864
0.856565
0.864085
0.871007
0.877843
0.8857
0.892244
0.903004
0.909412
1.00E-05
0.828668
0.837451
0.846173
0.854623
0.862948
0.872123
0.881937
0.890539
0.89988
0.909452
0.918768
9.00E-06
0.820514
0.829664
0.845036
0.855076
0.864312
0.871153
0.882174
0.892093
0.901535
0.910735
0.919377
8.00E-06
0.822501
0.832969
0.840952
0.853502
0.865111
0.870458
0.882782
0.892833
0.902294
0.910327
0.922888
7.00E-06
0.817717
0.832692
0.8407
0.853467
0.859422
0.872355
0.883391
0.892898
0.90414
0.91329
0.925745
6.00E-06
0.817586
0.823265
0.843416
0.851927
0.865222
0.872234
0.882063
0.894608
0.904774
0.915996
0.931062
5.00E-06
0.810378
0.824458
0.835816
0.846429
0.860388
0.869437
0.884145
0.89579
0.90825
0.920347
0.933843
4.00E-06
0.805051
0.811706
0.82903
0.843195
0.857068
0.871218
0.884181
0.897878
0.909331
0.926902
0.938179
3.00E-06
0.787682
0.810997
0.822259
0.836621
0.855494
0.87069
0.888698
0.902184
0.917566
0.93419
0.95077
2.00E-06
0.763432
0.794402
0.805494
0.824241
0.845353
0.865046
0.883803
0.905192
0.927214
0.946389
0.965694
1.00E-06
0.724051
0.754337
0.783356
0.806228
0.835685
0.852511
0.884135
0.911686
0.934241
0.959251
0.985453
0.799758
0.815194
0.829223
0.842531
0.8571
0.868722
0.883731
0.89757
0.910922
0.924688
0.939178
当c：a=0:1时网络的准确率是79.9758%当c：a=1：0时准确率是0.939178。
也就是当c与a是1:1进样的时候可以识别出79.9758%的a，可以识别出93.9178%的c。画成图
如果用pc和pa表示测试集中c与a的比例则
d2（c，a）-4-4-2-（2*k）,k∈{0,1}的平均准确率的表达式为
a
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
c
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
平均准确率p-ave
实测值
0.799758
0.815194
0.829223
0.842531
0.8571
0.868722
0.883731
0.89757
0.910922
0.924688
0.939178
计算值
0.799758
0.8137
0.827642
0.841584
0.855526
0.869468
0.88341
0.897352
0.911294
0.925236
0.939178
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);
训练集
0.5
c
0.5
a
测试集
0
c
10
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
0.499549
0.516406
181.2261
1
0.5
1.984925
410
0.006833
1
0.608557
0.395061
4200.583
0.972752
0.4
10.57286
2119
0.035317
1
0.719274
0.285924
4900.789
0.940176
0.3
9.477387
1886
0.031433
0.984985
0.81995
0.185329
5614.688
0.919266
0.2
10.49246
2088
0.0348
0.988989
0.913052
0.088778
6392.407
0.897858
0.1
11.28141
2260
0.037667
0.967968
0.991559
0.008561
9046.377
0.860297
0.01
16.69347
3323
0.055383
0.957958
0.999172
8.40E-04
13557.47
0.845785
0.001
23.15075
4607
0.076783
0.947948
0.999918
8.24E-05
22123.3
0.841002
1.00E-04
38.13065
7589
0.126483
0.941942
0.999927
7.41E-05
22794.29
0.838778
9.00E-05
39.96482
7953
0.13255
0.923924
0.999935
6.59E-05
23001.94
0.842139
8.00E-05
39.59799
7880
0.131333
0.925926
0.999944
5.69E-05
23585.63
0.835926
7.00E-05
40.05528
7972
0.132867
0.918919
0.99995
5.01E-05
25008.55
0.844488
6.00E-05
42.19095
8396
0.139933
0.933934
0.999959
4.14E-05
25913.07
0.843602
5.00E-05
44.65327
8901
0.14835
0.930931
0.999967
3.33E-05
27440.44
0.840413
4.00E-05
47.28141
9409
0.156817
0.935936
0.999976
2.44E-05
29798.31
0.832471
3.00E-05
51.21106
10206
0.1701
0.937938
0.999984
1.57E-05
32360.2
0.830001
2.00E-05
55.69849
11084
0.184733
0.914915
0.999992
8.29E-06
40696.32
0.828668
1.00E-05
68.1407
13560
0.226
0.916917
0.999993
7.35E-06
40860.31
0.820514
9.00E-06
69.38693
13808
0.230133
0.905906
0.999993
6.54E-06
43086.26
0.822501
8.00E-06
72.56784
14441
0.240683
0.90991
0.999994
5.59E-06
45369.46
0.817717
7.00E-06
76.98492
15352
0.255867
0.91992
0.999995
4.88E-06
48723.63
0.817586
6.00E-06
83.66834
16651
0.277517
0.908909
0.999996
4.11E-06
51922.08
0.810378
5.00E-06
86.40201
17210
0.286833
0.894895
0.999997
3.24E-06
57409.06
0.805051
4.00E-06
96.86935
19277
0.321283
0.886887
0.999998
2.37E-06
67244.38
0.787682
3.00E-06
113.3116
22549
0.375817
0.883884
0.999998
1.57E-06
87439.24
0.763432
2.00E-06
147.5025
29370
0.4895
0.873874
0.999999
7.30E-07
135984.2
0.724051
1.00E-06
228.2261
45417
0.75695
0.874875
训练集
0.5
c
0.5
a
测试集
0.1
c
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
0.499559
0.515984
182.5829
0.899814
0.5
2.18593
435
0.00725
0.923924
0.606113
0.398241
4202.759
0.960589
0.4
10.96482
2185
0.036417
0.984985
0.720845
0.28548
4989.995
0.928129
0.3
10.06533
2011
0.033517
0.968969
0.820824
0.185466
5666.899
0.908833
0.2
11.31658
2252
0.037533
0.962963
0.913407
0.088863
6468.276
0.890167
0.1
13.35176
2669
0.044483
0.943944
0.991783
0.008418
9085.854
0.858492
0.01
19.30653
3871
0.064517
0.944945
0.999179
8.34E-04
13474.49
0.849136
0.001
24.70854
4926
0.0821
0.915916
0.999917
8.33E-05
21834.68
0.847118
1.00E-04
39.28141
7832
0.130533
0.923924
0.999924
7.67E-05
22594.16
0.847712
9.00E-05
41.66834
8300
0.138333
0.912913
0.999936
6.48E-05
23324.59
0.84985
8.00E-05
41.62814
8286
0.1381
0.93994
0.999943
5.73E-05
24021.38
0.843215
7.00E-05
42.65829
8489
0.141483
0.928929
0.999951
4.97E-05
24449.18
0.843839
6.00E-05
43.9598
8756
0.145933
0.911912
0.99996
4.07E-05
25922.65
0.838819
5.00E-05
46.63317
9281
0.154683
0.913914
0.999967
3.31E-05
27721.9
0.84147
4.00E-05
49.60804
9880
0.164667
0.923924
0.999976
2.44E-05
29257.83
0.846424
3.00E-05
52.44221
10444
0.174067
0.921922
0.999984
1.65E-05
33433.06
0.838784
2.00E-05
59.96482
11933
0.198883
0.910911
0.999992
7.97E-06
40294.68
0.837451
1.00E-05
70.13568
13973
0.232883
0.916917
0.999993
7.14E-06
41447.01
0.829664
9.00E-06
72.50754
14429
0.240483
0.920921
0.999994
6.32E-06
43436.44
0.832969
8.00E-06
77.01005
15325
0.255417
0.912913
0.999994
5.67E-06
45515.5
0.832692
7.00E-06
79.71357
15863
0.264383
0.903904
0.999995
4.85E-06
48621.58
0.823265
6.00E-06
85.1809
16967
0.282783
0.908909
0.999996
4.06E-06
52096.44
0.824458
5.00E-06
90.80905
18087
0.30145
0.915916
0.999997
3.18E-06
56452.7
0.811706
4.00E-06
98.50754
19614
0.3269
0.906907
0.999998
2.42E-06
69077.34
0.810997
3.00E-06
124.4874
24781
0.413017
0.896897
0.999998
1.56E-06
89547.6
0.794402
2.00E-06
157.2312
31290
0.5215
0.886887
0.999999
7.79E-07
137262.4
0.754337
1.00E-06
239.2613
47613
0.79355
0.886887
训练集
0.5
c
0.5
a
测试集
0.2
c
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
0.499574
0.516447
180.3618
0.800947
0.5
2.105528
422
0.007033
0.825826
0.60819
0.396577
4089.965
0.947369
0.4
11.37688
2266
0.037767
0.974975
0.719785
0.284101
5063.613
0.920393
0.3
10.51256
2093
0.034883
0.95996
0.821756
0.183251
5692.839
0.902908
0.2
11.97487
2383
0.039717
0.941942
0.912977
0.089543
6441.352
0.884533
0.1
12.61307
2534
0.042233
0.940941
0.991778
0.008391
9114.005
0.861545
0.01
17.1608
3415
0.056917
0.927928
0.999187
8.34E-04
13574.83
0.853804
0.001
24.68844
4938
0.0823
0.92993
0.99992
8.14E-05
21810.96
0.853859
1.00E-04
38.47739
7673
0.127883
0.904905
0.999925
7.54E-05
22252.22
0.855489
9.00E-05
39.37186
7835
0.130583
0.91992
0.999935
6.59E-05
22982.17
0.850147
8.00E-05
40.54774
8069
0.134483
0.908909
0.999942
5.82E-05
24360.98
0.853391
7.00E-05
42.86432
8545
0.142417
0.925926
0.99995
5.06E-05
24677.08
0.850861
6.00E-05
43.68342
8712
0.1452
0.90991
0.99996
4.02E-05
25614.11
0.851419
5.00E-05
44.96482
8948
0.149133
0.918919
0.999967
3.31E-05
28052.94
0.850931
4.00E-05
50.21608
9993
0.16655
0.908909
0.999976
2.46E-05
29987.39
0.849905
3.00E-05
55.91457
11128
0.185467
0.910911
0.999984
1.66E-05
32564.42
0.848864
2.00E-05
57.42714
11428
0.190467
0.918919
0.999992
8.37E-06
40750.95
0.846173
1.00E-05
72.9397
14516
0.241933
0.906907
0.999993
7.16E-06
41606.82
0.845036
9.00E-06
73.82915
14699
0.244983
0.90991
0.999994
6.30E-06
43684.39
0.840952
8.00E-06
76.90452
15312
0.2552
0.914915
0.999994
5.66E-06
47273.61
0.8407
7.00E-06
84.40704
16798
0.279967
0.903904
0.999995
4.93E-06
49040.06
0.843416
6.00E-06
86.94472
17314
0.288567
0.908909
0.999996
4.00E-06
52722.31
0.835816
5.00E-06
93.31658
18570
0.3095
0.913914
0.999997
3.14E-06
58800.85
0.82903
4.00E-06
103.8492
20674
0.344567
0.905906
0.999998
2.40E-06
67739.11
0.822259
3.00E-06
120.4271
23965
0.399417
0.891892
0.999998
1.56E-06
89028.13
0.805494
2.00E-06
155.598
30964
0.516067
0.888889
0.999999
7.53E-07
133418.5
0.783356
1.00E-06
235.794
46923
0.78205
0.915916
训练集
0.5
c
0.5
a
测试集
0.3
c
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
0.499563
0.516764
181.0653
0.699901
0.5
2.326633
464
0.007733
0.735736
0.598109
0.405997
4276.201
0.93407
0.4
10.8191
2154
0.0359
0.965966
0.719868
0.285054
4898.859
0.913049
0.3
10.84422
2160
0.036
0.942943
0.819455
0.184858
5667.925
0.895871
0.2
11.59296
2324
0.038733
0.942943
0.91238
0.089912
6447.111
0.882762
0.1
12.84925
2558
0.042633
0.921922
0.991728
0.008504
9021.392
0.861354
0.01
18.29648
3642
0.0607
0.915916
0.999179
8.39E-04
13362.78
0.856238
0.001
25.00503
4976
0.082933
0.912913
0.999918
8.29E-05
21841.37
0.858099
1.00E-04
39.49246
7860
0.131
0.907908
0.999925
7.58E-05
22117.96
0.85398
9.00E-05
39.78894
7918
0.131967
0.910911
0.999935
6.59E-05
23374.83
0.857531
8.00E-05
42.23116
8410
0.140167
0.908909
0.999943
5.73E-05
24088.55
0.859095
7.00E-05
43.53769
8673
0.14455
0.918919
0.999951
4.99E-05
24232.19
0.858331
6.00E-05
42.79899
8518
0.141967
0.912913
0.999959
4.10E-05
25768.17
0.855725
5.00E-05
46.08543
9188
0.153133
0.913914
0.999968
3.27E-05
27126.85
0.856711
4.00E-05
50.10553
9972
0.1662
0.915916
0.999975
2.52E-05
29311.8
0.8591
3.00E-05
53.26131
10600
0.176667
0.907908
0.999984
1.64E-05
33121.41
0.856565
2.00E-05
58.23116
11588
0.193133
0.906907
0.999992
7.99E-06
40591.23
0.854623
1.00E-05
-683.337
-135968
-2.26613
0.915916
0.999992
7.57E-06
42242.42
0.855076
9.00E-06
74.38693
14813
0.246883
0.911912
0.999993
6.59E-06
43143.39
0.853502
8.00E-06
76.71357
15274
0.254567
0.901902
0.999994
5.78E-06
44787.41
0.853467
7.00E-06
79.02513
15734
0.262233
0.912913
0.999995
4.78E-06
48122.64
0.851927
6.00E-06
85.11558
16954
0.282567
0.908909
0.999996
4.01E-06
52609.96
0.846429
5.00E-06
92.59799
18427
0.307117
0.8999
0.999997
3.24E-06
55338.42
0.843195
4.00E-06
98.55779
19625
0.327083
0.904905
0.999998
2.34E-06
66951.1
0.836621
3.00E-06
117.8543
23454
0.3909
0.905906
0.999998
1.56E-06
89347.83
0.824241
2.00E-06
156.9146
31236
0.5206
0.904905
0.999999
7.51E-07
135976.4
0.806228
1.00E-06
240.4925
47867
0.797783
0.8999
训练集
0.5
c
0.5
a
测试集
0.4
c
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
0.49958
0.51682
179.0151
0.600656
0.5
2.261307
451
0.007517
0.637638
0.600314
0.403284
4147.814
0.919779
0.4
9.869347
1967
0.032783
0.955956
0.720156
0.283361
4858.015
0.904261
0.3
9.979899
1987
0.033117
0.936937
0.8191
0.184599
5643.131
0.890448
0.2
11.17085
2232
0.0372
0.937938
0.914373
0.088719
6489.02
0.879337
0.1
13.38693
2672
0.044533
0.910911
0.991696
0.008514
9133.513
0.860916
0.01
17.74372
3532
0.058867
0.901902
0.999189
8.28E-04
13622.89
0.857425
0.001
25.07538
5000
0.083333
0.893894
0.999918
8.26E-05
22021.09
0.862878
1.00E-04
40.02513
7973
0.132883
0.901902
0.999927
7.37E-05
22306.3
0.863869
9.00E-05
40.60302
8080
0.134667
0.90991
0.999935
6.60E-05
22824.64
0.862667
8.00E-05
41.44724
8256
0.1376
0.900901
0.999942
5.83E-05
23730.49
0.861877
7.00E-05
42.99497
8556
0.1426
0.906907
0.999951
4.98E-05
24605.66
0.864664
6.00E-05
43.64824
8699
0.144983
0.908909
0.999959
4.14E-05
25657.53
0.863321
5.00E-05
45.70352
9095
0.151583
0.903904
0.999968
3.28E-05
26791.41
0.864135
4.00E-05
47.70352
9501
0.15835
0.904905
0.999975
2.49E-05
29539.54
0.864261
3.00E-05
52.92462
10549
0.175817
0.911912
0.999984
1.63E-05
33273.41
0.864085
2.00E-05
59.76382
11893
0.198217
0.910911
0.999992
8.12E-06
39976.45
0.862948
1.00E-05
71.15075
14167
0.236117
0.903904
0.999993
7.32E-06
41968.86
0.864312
9.00E-06
74.44221
14814
0.2469
0.910911
0.999993
6.54E-06
43223.41
0.865111
8.00E-06
76.43719
15219
0.25365
0.917918
0.999994
5.61E-06
45152.72
0.859422
7.00E-06
80.22613
15973
0.266217
0.914915
0.999995
4.88E-06
48118.48
0.865222
6.00E-06
84.86432
16888
0.281467
0.911912
0.999996
4.01E-06
52712.63
0.860388
5.00E-06
92.54774
18424
0.307067
0.903904
0.999997
3.13E-06
58584.19
0.857068
4.00E-06
103.7337
20659
0.344317
0.907908
0.999998
2.38E-06
68493.08
0.855494
3.00E-06
121.3367
24154
0.402567
0.907908
0.999998
1.53E-06
90109.07
0.845353
2.00E-06
158.4422
31530
0.5255
0.920921
0.999999
7.59E-07
138970.9
0.835685
1.00E-06
244.603
48679
0.811317
0.902903
训练集
0.5
c
0.5
a
测试集
0.5
c
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
0.499569
0.516438
180.5829
0.500707
0.5
2.130653
440
0.007333
0.62963
0.603337
0.401501
4137.648
0.907405
0.4
8.809045
1785
0.02975
0.950951
0.719646
0.285824
4989.814
0.895856
0.3
10.12563
2015
0.033583
0.926927
0.820554
0.184931
5547.925
0.88653
0.2
11.10553
2223
0.03705
0.920921
0.91342
0.090681
6274.035
0.874789
0.1
12.07035
2418
0.0403
0.906907
0.991693
0.008545
9126.075
0.863416
0.01
19.40201
3885
0.06475
0.896897
0.999186
8.31E-04
13454.45
0.862772
0.001
24.82915
4952
0.082533
0.897898
0.999919
8.20E-05
21767.77
0.868139
1.00E-04
38.61809
7686
0.1281
0.901902
0.999927
7.38E-05
22688.33
0.867088
9.00E-05
40.18593
8005
0.133417
0.900901
0.999934
6.63E-05
23244.96
0.869241
8.00E-05
41.75377
8309
0.138483
0.908909
0.999942
5.85E-05
24121.77
0.869422
7.00E-05
43.72864
8720
0.145333
0.904905
0.999951
4.99E-05
24631.73
0.869216
6.00E-05
43.59799
8678
0.144633
0.903904
0.999959
4.08E-05
26436.8
0.868698
5.00E-05
47.10553
9375
0.15625
0.903904
0.999967
3.28E-05
27315.33
0.869442
4.00E-05
48.9397
9748
0.162467
0.906907
0.999975
2.47E-05
29643.38
0.869251
3.00E-05
52.57286
10464
0.1744
0.902903
0.999984
1.62E-05
32292.43
0.871007
2.00E-05
57.10553
11374
0.189567
0.90991
0.999992
8.19E-06
41034.54
0.872123
1.00E-05
72.9799
14523
0.24205
0.904905
0.999993
7.21E-06
41606.16
0.871153
9.00E-06
72.73367
14476
0.241267
0.910911
0.999994
6.43E-06
44033.12
0.870458
8.00E-06
78.95477
15721
0.262017
0.922923
0.999994
5.58E-06
46175.42
0.872355
7.00E-06
80.93467
16114
0.268567
0.916917
0.999995
4.84E-06
48331.53
0.872234
6.00E-06
85.45226
17005
0.283417
0.910911
0.999996
3.92E-06
51810.29
0.869437
5.00E-06
90.86935
18086
0.301433
0.907908
0.999997
3.20E-06
57527.16
0.871218
4.00E-06
100.7487
20049
0.33415
0.906907
0.999998
2.36E-06
66458.26
0.87069
3.00E-06
116.598
23204
0.386733
0.910911
0.999998
1.53E-06
90787.31
0.865046
2.00E-06
159.1608
31681
0.528017
0.917918
0.999999
7.22E-07
135588.1
0.852511
1.00E-06
237.0905
47182
0.786367
0.918919
训练集
0.5
c
0.5
a
测试集
0.6
c
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
0.49954
0.51599
179.8894
0.401804
0.5
2.316583
461
0.007683
0.556557
0.59933
0.406381
4250.412
0.89401
0.4
10.37186
2066
0.034433
0.937938
0.719687
0.286905
4938.744
0.889005
0.3
10.90452
2174
0.036233
0.924925
0.819933
0.184967
5618.357
0.88076
0.2
12.12563
2429
0.040483
0.928929
0.913921
0.088409
6496.568
0.87154
0.1
12.70352
2539
0.042317
0.908909
0.991857
0.00838
9006.839
0.86322
0.01
18.86432
3756
0.0626
0.906907
0.99916
8.59E-04
13661.67
0.865655
0.001
25.45729
5066
0.084433
0.893894
0.999918
8.28E-05
21811.38
0.872898
1.00E-04
38.63819
7692
0.1282
0.902903
0.999926
7.49E-05
22495.78
0.874558
9.00E-05
40.0603
7977
0.13295
0.906907
0.999934
6.62E-05
23493.12
0.872938
8.00E-05
42.41206
8448
0.1408
0.905906
0.999942
5.86E-05
23754.33
0.874804
7.00E-05
43.43216
8670
0.1445
0.910911
0.999951
4.93E-05
24638.54
0.874382
6.00E-05
44.87437
8940
0.149
0.902903
0.999959
4.10E-05
25591.86
0.876173
5.00E-05
46.9598
9354
0.1559
0.907908
0.999967
3.31E-05
27544.34
0.876485
4.00E-05
49.32663
9840
0.164
0.901902
0.999976
2.44E-05
29500.22
0.877858
3.00E-05
53.20101
10596
0.1766
0.90991
0.999984
1.64E-05
32686.55
0.877843
2.00E-05
59.0201
11752
0.195867
0.90991
0.999992
8.09E-06
40127.56
0.881937
1.00E-05
72.15075
14386
0.239767
0.916917
0.999993
7.20E-06
41876.3
0.882174
9.00E-06
76.05025
15140
0.252333
0.915916
0.999994
6.38E-06
42827.65
0.882782
8.00E-06
75.92965
15113
0.251883
0.90991
0.999994
5.67E-06
45319.83
0.883391
7.00E-06
80.84422
16103
0.268383
0.920921
0.999995
4.74E-06
48401.74
0.882063
6.00E-06
86.63317
17247
0.28745
0.915916
0.999996
4.03E-06
52139.26
0.884145
5.00E-06
93.47236
18609
0.31015
0.921922
0.999997
3.26E-06
57559.79
0.884181
4.00E-06
102.2261
20343
0.33905
0.923924
0.999998
2.38E-06
68157.14
0.888698
3.00E-06
122.0854
24310
0.405167
0.922923
0.999998
1.58E-06
90160.77
0.883803
2.00E-06
162.1357
32267
0.537783
0.920921
0.999999
7.54E-07
136942.9
0.884135
1.00E-06
246.6382
49081
0.818017
0.935936
训练集
0.5
c
0.5
a
测试集
0.7
c
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
0.499574
0.516069
181.598
0.299521
0.5
2.351759
468
0.0078
0.431431
0.603308
0.403244
4161.523
0.8816
0.4
10.51256
2109
0.03515
0.938939
0.720446
0.285064
5090.678
0.876962
0.3
10.66332
2122
0.035367
0.927928
0.819105
0.183854
5576.99
0.874825
0.2
10.99497
2188
0.036467
0.921922
0.914607
0.088364
6373.834
0.865982
0.1
11.83417
2355
0.03925
0.906907
0.991859
0.008432
9179.874
0.863959
0.01
17.33668
3465
0.05775
0.906907
0.999178
8.35E-04
13354.09
0.868713
0.001
24.28141
4832
0.080533
0.8999
0.999917
8.33E-05
21958.43
0.877858
1.00E-04
38.9799
7757
0.129283
0.90991
0.999925
7.56E-05
22522.38
0.878406
9.00E-05
40.44221
8049
0.13415
0.904905
0.999935
6.60E-05
23078.95
0.879306
8.00E-05
40.58291
8091
0.13485
0.90991
0.999943
5.76E-05
24030.42
0.880363
7.00E-05
42.77889
8514
0.1419
0.908909
0.999951
4.94E-05
24621.14
0.87991
6.00E-05
43.65327
8687
0.144783
0.910911
0.999959
4.11E-05
26012.24
0.88232
5.00E-05
45.75377
9106
0.151767
0.910911
0.999967
3.28E-05
27212.47
0.882244
4.00E-05
48.61809
9676
0.161267
0.91992
0.999975
2.47E-05
29759
0.884784
3.00E-05
52.51256
10450
0.174167
0.921922
0.999984
1.62E-05
32716.42
0.8857
2.00E-05
57.48744
11440
0.190667
0.915916
0.999992
8.01E-06
40474.56
0.890539
1.00E-05
69.8794
13906
0.231767
0.922923
0.999993
7.26E-06
41655.76
0.892093
9.00E-06
73.37688
14602
0.243367
0.91992
0.999994
6.35E-06
43837.8
0.892833
8.00E-06
75.29648
14984
0.249733
0.917918
0.999994
5.72E-06
45508.2
0.892898
7.00E-06
78.9196
15720
0.262
0.921922
0.999995
4.84E-06
48046.68
0.894608
6.00E-06
83.58794
16635
0.27725
0.92993
0.999996
4.09E-06
52130.5
0.89579
5.00E-06
92.37186
18382
0.306367
0.925926
0.999997
3.08E-06
58917.21
0.897878
4.00E-06
103.402
20577
0.34295
0.932933
0.999998
2.39E-06
69297.22
0.902184
3.00E-06
119.9246
23865
0.39775
0.935936
0.999998
1.58E-06
88388.22
0.905192
2.00E-06
153.0553
30458
0.507633
0.942943
0.999999
7.63E-07
134608.2
0.911686
1.00E-06
229.7487
45721
0.762017
0.948949
训练集
0.5
c
0.5
a
测试集
0.8
c
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
0.49957
0.517276
178.6432
0.201734
0.5
2.276382
453
0.00755
0.571572
0.596191
0.409275
4308.121
0.863985
0.4
9.879397
1966
0.032767
0.935936
0.721564
0.284414
5039.603
0.871897
0.3
10.20101
2031
0.03385
0.934935
0.819136
0.18578
5468.508
0.868648
0.2
10.49246
2089
0.034817
0.914915
0.913967
0.088
6519.111
0.862174
0.1
12.50251
2488
0.041467
0.912913
0.991692
0.008503
9109.593
0.863783
0.01
16.60804
3321
0.05535
0.903904
0.999177
8.44E-04
13365.6
0.873124
0.001
23.06533
4590
0.0765
0.914915
0.999918
8.28E-05
21980.47
0.884749
1.00E-04
37.50251
7463
0.124383
0.920921
0.999926
7.47E-05
22913.42
0.883321
9.00E-05
39.23618
7808
0.130133
0.918919
0.999934
6.65E-05
23212.56
0.885262
8.00E-05
39.67839
7929
0.13215
0.916917
0.999942
5.91E-05
23828.58
0.88493
7.00E-05
41.80402
8335
0.138917
0.921922
0.999952
4.88E-05
24650.27
0.886756
6.00E-05
43.17588
8593
0.143217
0.916917
0.999959
4.12E-05
26077.28
0.887304
5.00E-05
45.24121
9005
0.150083
0.917918
0.999967
3.29E-05
27834.55
0.886736
4.00E-05
48.25126
9603
0.16005
0.917918
0.999975
2.49E-05
29445.53
0.889709
3.00E-05
49.30151
9812
0.163533
0.924925
0.999984
1.65E-05
33110.89
0.892244
2.00E-05
56.48744
11242
0.187367
0.930931
0.999992
8.01E-06
40939.64
0.89988
1.00E-05
69.28141
13787
0.229783
0.931932
0.999993
7.23E-06
41251.03
0.901535
9.00E-06
70.90955
14127
0.23545
0.926927
0.999993
6.56E-06
43254.39
0.902294
8.00E-06
72.59799
14464
0.241067
0.933934
0.999994
5.67E-06
45273.62
0.90414
7.00E-06
76.40201
15235
0.253917
0.931932
0.999995
4.91E-06
48802.05
0.904774
6.00E-06
82.31658
16381
0.273017
0.931932
0.999996
4.08E-06
52308.5
0.90825
5.00E-06
89.9598
17917
0.298617
0.946947
0.999997
3.22E-06
56429.72
0.909331
4.00E-06
95.1206
18929
0.315483
0.946947
0.999998
2.28E-06
69000.35
0.917566
3.00E-06
115.2161
22928
0.382133
0.952953
0.999998
1.55E-06
90282.26
0.927214
2.00E-06
151.8141
30227
0.503783
0.951952
0.999999
7.41E-07
136573
0.934241
1.00E-06
227.6432
45301
0.755017
0.961962
训练集
0.5
c
0.5
a
测试集
0.9
c
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
0.49955
0.516249
182.2814
0.102912
0.5
2.045226
422
0.007033
0.608609
0.59648
0.408358
4231.94
0.855051
0.4
9.879397
1967
0.032783
0.937938
0.717152
0.286692
4963.658
0.864402
0.3
9.698492
1930
0.032167
0.928929
0.818898
0.185611
5492.337
0.860725
0.2
10.47236
2085
0.03475
0.918919
0.913352
0.08792
6436.628
0.856802
0.1
11.82412
2353
0.039217
0.91992
0.991814
0.008432
9097.111
0.865841
0.01
17.57789
3498
0.0583
0.923924
0.999183
8.30E-04
13446.7
0.877003
0.001
23.85427
4748
0.079133
0.927928
0.999917
8.34E-05
22125.06
0.886022
1.00E-04
37.55779
7475
0.124583
0.942943
0.999927
7.38E-05
22538.11
0.887737
9.00E-05
37.86935
7537
0.125617
0.93994
0.999934
6.61E-05
23101.24
0.888532
8.00E-05
39.25628
7812
0.1302
0.937938
0.999943
5.80E-05
23967.55
0.887994
7.00E-05
40.37186
8050
0.134167
0.941942
0.999951
4.97E-05
25325.18
0.889281
6.00E-05
42.64824
8487
0.14145
0.926927
0.999959
4.15E-05
26357.19
0.890514
5.00E-05
45.08543
8972
0.149533
0.948949
0.999967
3.31E-05
27632.76
0.896303
4.00E-05
46.47236
9248
0.154133
0.944945
0.999975
2.50E-05
30036.84
0.896992
3.00E-05
51.74874
10298
0.171633
0.943944
0.999984
1.61E-05
32758.53
0.903004
2.00E-05
55.60302
11081
0.184683
0.937938
0.999992
7.96E-06
40773.38
0.909452
1.00E-05
69.0201
13751
0.229183
0.955956
0.999993
7.05E-06
42092.54
0.910735
9.00E-06
71.1005
14149
0.235817
0.957958
0.999994
6.49E-06
43774.2
0.910327
8.00E-06
73.9196
14726
0.245433
0.94995
0.999994
5.67E-06
45774.17
0.91329
7.00E-06
78.1005
15542
0.259033
0.948949
0.999995
4.75E-06
48704.56
0.915996
6.00E-06
81.55779
16230
0.2705
0.953954
0.999996
4.01E-06
52890.92
0.920347
5.00E-06
88.80402
17689
0.294817
0.952953
0.999997
3.15E-06
57608.75
0.926902
4.00E-06
96.25126
19154
0.319233
0.962963
0.999998
2.40E-06
67016.85
0.93419
3.00E-06
114.1357
22729
0.378817
0.968969
0.999998
1.57E-06
90306.9
0.946389
2.00E-06
152.6432
30376
0.506267
0.980981
0.999999
7.32E-07
136105.9
0.959251
1.00E-06
227.8442
45342
0.7557
0.983984
训练集
0.5
c
0.5
a
测试集
1
c
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
0.499518
0.516981
177.9095
6.09E-04
0.5
2.045226
407
0.006783
0.109109
0.601165
0.403701
4208.09
0.84492
0.4
9.824121
1955
0.032583
0.930931
0.719308
0.2858
5071.151
0.856746
0.3
9.809045
1952
0.032533
0.931932
0.819674
0.184769
5579.241
0.857264
0.2
10.64824
2119
0.035317
0.952953
0.912692
0.089165
6338.638
0.857128
0.1
11.61809
2312
0.038533
0.938939
0.99186
0.008414
9149.03
0.866193
0.01
17.22613
3429
0.05715
0.928929
0.999175
8.41E-04
13378.6
0.876867
0.001
23.0201
4582
0.076367
0.951952
0.999919
8.20E-05
21734.3
0.893491
1.00E-04
36.8392
7331
0.122183
0.957958
0.999926
7.48E-05
22469.31
0.896248
9.00E-05
38.23116
7624
0.127067
0.971972
0.999935
6.54E-05
23319.39
0.893949
8.00E-05
40.77387
8114
0.135233
0.948949
0.999942
5.81E-05
23928.25
0.895368
7.00E-05
40.89447
8139
0.13565
0.948949
0.999951
5.00E-05
24340.47
0.897646
6.00E-05
42
8374
0.139567
0.955956
0.999959
4.17E-05
25587.68
0.897335
5.00E-05
44.0402
8765
0.146083
0.955956
0.999967
3.30E-05
27485.81
0.901082
4.00E-05
47.1608
9401
0.156683
0.957958
0.999975
2.47E-05
29732.46
0.901761
3.00E-05
51.43216
10268
0.171133
0.958959
0.999984
1.60E-05
33218.88
0.909412
2.00E-05
56.33668
11211
0.18685
0.973974
0.999992
8.20E-06
40049.86
0.918768
1.00E-05
68.94472
13720
0.228667
0.960961
0.999993
7.40E-06
41860.41
0.919377
9.00E-06
71
14131
0.235517
0.975976
0.999993
6.53E-06
43537.62
0.922888
8.00E-06
73.67337
14661
0.24435
0.980981
0.999994
5.57E-06
45599.75
0.925745
7.00E-06
76.27638
15181
0.253017
0.972973
0.999995
4.79E-06
47753.81
0.931062
6.00E-06
79.81407
15900
0.265
0.983984
0.999996
4.08E-06
51002.03
0.933843
5.00E-06
87.39196
17391
0.28985
0.983984
0.999997
3.18E-06
57933.75
0.938179
4.00E-06
98.76884
19656
0.3276
0.981982
0.999998
2.34E-06
66033.08
0.95077
3.00E-06
112.191
22341
0.37235
0.997998
0.999998
1.59E-06
91024.78
0.965694
2.00E-06
154.3015
30739
0.512317
1
0.999999
7.53E-07
132599.5
0.985453
1.00E-06
221.402
44059
0.734317
1
