测试集的构成比例对网络分类性能的影响cp -一个渣渣

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c和p，训练集c：p=1：1

测试集c和p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，观察分类准确率。

实验过程

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；向右侧网络输入排斥子让右侧向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26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
p	2	判断是否达到收敛
梯度下降
c	3	判断是否达到收敛
p	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，p）-4-4-2-（2*k）,k∈{0,1}

得到的数据

	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	147.1156	149.3769	149.4774	148.6834	147.7387	143.608	144.0101	147.1859	147.6382	146.9146	148.8543
0.4	6276.191	6215.095	6098.075	6083.377	6290.528	6294.472	6176.106	6170.256	6484.824	6527.412	6306.643
0.3	10407.46	10095	10101.43	9926.829	10142.37	10298.52	10020.49	10198.22	10131.84	10246.46	10420.39
0.2	15413.02	15648.27	15381.82	15081.81	15518.25	15136.69	15237	15533.43	15523	15367.05	15286.05
0.1	24893	24143.51	23415.88	25458.15	23850.98	24598.91	24297.87	24151.85	24210.84	23795.91	24414.25
0.01	59839.55	57398.58	57319.32	60125.44	59417.48	58611.96	59634.89	58560.76	59353.21	58745.15	59492.45
0.001	115431.3	111056.3	114920.8	114448.1	116262.8	110127.1	113627.5	107564.9	112414.9	112087.3	110956.2
1.00E-04	201610.7	209420.7	208738.6	201745.8	205856.7	210147.9	208380.2	209740.5	202288.7	201300.8	206577
9.00E-05	214043.2	216748.9	221076.9	218610	211840.9	217360.5	216305.4	216424.9	210094	216170.4	225841.7
8.00E-05	221814.5	214543.9	216763.3	217172.3	221671.8	216707.4	214417.1	221045.8	215015	222313.3	214257.2
7.00E-05	227647.3	225663	224240.8	225940.6	220378.4	231569.8	230410.2	228854.8	224129	221232.3	227719.7
6.00E-05	227186.4	236632.7	235333.2	241970.3	237669.7	240362.2	240610.8	236488.6	238662.6	232360.4	237107
5.00E-05	245535.8	246959.7	250549.8	245810.6	248245.3	247742.3	251281.2	258579.6	253117.2	260700.7	245581.8
4.00E-05	269459.6	267081.9	268540.2	271148.9	262207.7	265902.8	257265.3	261405.7	265370	264792.1	266492.5
3.00E-05	288701.1	288744.6	278070.8	282794.2	280333.4	275920.6	284119.7	282179.7	288700.1	282872.8	282278.7
2.00E-05	317002.9	307150	309043	319202.7	308706.8	315850.9	321754.6	307254.4	313644	327235	323231
1.00E-05	365424.3	364250.3	365575.7	363737.2	366601.9	376979.7	380203	370808.6	383179.2	356469.7	368836.5
9.00E-06	392786	372373.8	377167.3	379308.8	367669	381299.2	381642.5	393012.9	372760.6	396853.3	377565.8
8.00E-06	388277.8	399035.5	383537.9	392360	385873.2	391096.8	373728.2	415835.2	394640.4	394890.9	387094.8
7.00E-06	400704.8	396170.6	414295.7	401929.8	417806.1	385739.1	394560.6	406917.3	415909.8	401407	404034.1
6.00E-06	418156.4	409632.3	408388.9	416268.6	410188.3	427177.7	405782.9	431013.1	408631.4	421662	415028.5
5.00E-06	429071.5	420660.3	439646.4	444445	451018.4	438041.1	428783.3	438260.3	444052.7	428640.5	449725.6
4.00E-06	475880.1	462762.7	471101	464173.3	459961.9	498834.6	447440.4	458926.4	477415.8	465639.4	463744.1
3.00E-06	494180.2	470073.5	492323.9	491751.2	482770.3	497042.6	492925.9	488579	498906.6	483048.2	492239.1
2.00E-06	523083	540346.3	547355	522686.8	539823.1	553016.1	536623	540344.3	534051.6	528076.3	560042.5
1.00E-06	626599.7	642248.7	645703.1	624471.1	620645.9	606547.8	623932.2	620509.7	625129.6	656599.3	634245.9

迭代次数曲线是高度重合的，因为11组测试的进样样本比例构成是一致的都是

c：p=1：1。

平均准确率的数据

p		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	0.899794	0.800464	0.700469	0.602029	0.4987	0.398861	0.300788	0.199763	0.100714	0
	0.4	1	0.993124	0.985533	0.978667	0.971685	0.962249	0.959407	0.949055	0.94661	0.936273	0.931228
	0.3	1	0.997123	0.995252	0.992324	0.989115	0.987324	0.983868	0.981942	0.980594	0.976283	0.974985
	0.2	0.999713	0.998627	0.998053	0.996444	0.996157	0.995392	0.993994	0.99243	0.991534	0.991353	0.991293
	0.1	0.991378	0.992334	0.994185	0.993013	0.994723	0.994879	0.996278	0.996056	0.9967	0.997656	0.998174
	0.01	0.805469	0.843628	0.86149	0.860428	0.885031	0.910534	0.925835	0.952118	0.962002	0.981564	1
	0.001	0.502568	0.56382	0.590958	0.662426	0.693804	0.756037	0.801309	0.862481	0.906087	0.952505	1
	1.00E-04	0.362101	0.418197	0.468861	0.559645	0.622512	0.672788	0.742682	0.796052	0.871822	0.938179	1
	9.00E-05	0.350089	0.408308	0.453423	0.533176	0.609474	0.667084	0.735776	0.801334	0.871459	0.933049	1
	8.00E-05	0.332579	0.414143	0.482166	0.543906	0.601149	0.680223	0.739448	0.795997	0.873336	0.933466	1
	7.00E-05	0.341975	0.409706	0.469022	0.5237	0.609937	0.664579	0.73143	0.796233	0.866324	0.934477	1
	6.00E-05	0.349933	0.408665	0.464646	0.51877	0.602925	0.661707	0.732909	0.797421	0.86648	0.934653	1
	5.00E-05	0.329244	0.404998	0.443464	0.522492	0.587029	0.668332	0.728895	0.789584	0.864352	0.930564	1
	4.00E-05	0.322468	0.3719	0.436165	0.495812	0.592864	0.653654	0.728517	0.799468	0.861314	0.930936	1
	3.00E-05	0.29949	0.362709	0.448821	0.513066	0.588508	0.661762	0.724141	0.787416	0.857496	0.930287	1
	2.00E-05	0.290507	0.379259	0.427387	0.507653	0.574444	0.64132	0.713523	0.793784	0.859105	0.928552	1
	1.00E-05	0.280139	0.335858	0.423363	0.495184	0.561788	0.634821	0.707185	0.785363	0.85316	0.929085	1
	9.00E-06	0.24607	0.361719	0.424575	0.491004	0.571461	0.633312	0.707416	0.776882	0.855896	0.925171	1
	8.00E-06	0.260743	0.334485	0.423625	0.483141	0.565254	0.633513	0.716933	0.770695	0.851756	0.925921	1
	7.00E-06	0.25597	0.334219	0.396879	0.49047	0.558227	0.642869	0.70886	0.776983	0.851007	0.928184	1
	6.00E-06	0.254732	0.342428	0.415078	0.475375	0.55957	0.631305	0.711853	0.777089	0.853175	0.927163	1
	5.00E-06	0.266246	0.343142	0.40537	0.471854	0.543297	0.634383	0.704619	0.77977	0.851384	0.925946	1
	4.00E-06	0.248595	0.335647	0.39871	0.483348	0.547648	0.615535	0.703905	0.781052	0.848346	0.924407	1
	3.00E-06	0.251186	0.336422	0.397111	0.473856	0.5554	0.620419	0.701405	0.774382	0.847229	0.926535	1
	2.00E-06	0.254184	0.317901	0.398499	0.473735	0.553287	0.617628	0.700017	0.772491	0.851651	0.925337	1
	1.00E-06	0.238138	0.313062	0.38693	0.479238	0.542749	0.622884	0.696319	0.771445	0.84736	0.923129	1

从图中可以看到平均准确率的变化是高度规则的。

p	平均准确率p-ave	1	0.9	0.8	0.7	0.6	0.5	0.4	0.3	0.2	0.1	0
c	δ	0	0.1	0.2	0.3	0.4	0.5	0.6	0.7	0.8	0.9	1
		0.2556	0.335488	0.407014	0.48172	0.555868	0.628667	0.705851	0.776615	0.851096	0.926088	1

将1e-6<=δ<=1e-5的平均准确率取平均值。数据清晰的表明当c与p的进样比例是1：1的情况下，网络

d2（c，p）-4-4-2-（2*k）,k∈{0,1}

可以识别测试集中的25.56%的p，和100%的c。如果用pc表示测试集中c占的比例

，用pp表示测试集中p占的比例。则总的平均准确率为

pp	平均准确率p-ave	1	0.9	0.8	0.7	0.6	0.5	0.4	0.3	0.2	0.1	0
pc		0	0.1	0.2	0.3	0.4	0.5	0.6	0.7	0.8	0.9	1
实测值		0.2556	0.335488	0.407014	0.48172	0.555868	0.628667	0.705851	0.776615	0.851096	0.926088	1
计算值	0.2556*pp+pc	0.2556	0.33004	0.40448	0.47892	0.55336	0.6278	0.70224	0.77668	0.85112	0.92556	1

从图上看很直观，这两条线高度重合。表明这种算法的精度是有保证的。

表明网络d2（c，p）-4-4-2-（2*k）,k∈{0,1}的准确率是一种线性变量与测试集的比例高度相关。可以用表达式

精确的估算。

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	p
测试集	0	c
	1	p
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
0.499498	0.530138	147.1156	1	0.5	2.040201	406	0.006767	1
0.602299	0.401437	6276.191	1	0.4	12.72864	2533	0.042217	1
0.710881	0.29236	10407.46	1	0.3	19.04523	3790	0.063167	1
0.810562	0.192418	15413.02	0.999713	0.2	26.62814	5299	0.088317	1
0.906917	0.094307	24893	0.991378	0.1	42.68342	8494	0.141567	1
0.991307	0.008775	59839.55	0.805469	0.01	100.5779	20015	0.333583	1
0.999158	8.48E-04	115431.3	0.502568	0.001	194.2714	38660	0.644333	0.966967
0.999916	8.40E-05	201610.7	0.362101	1.00E-04	338.3216	67326	1.1221	0.777778
0.999926	7.37E-05	214043.2	0.350089	9.00E-05	358.5427	71350	1.189167	0.833834
0.999935	6.49E-05	221814.5	0.332579	8.00E-05	370.799	73789	1.229817	0.698699
0.999943	5.72E-05	227647.3	0.341975	7.00E-05	381.6131	75941	1.265683	0.900901
0.999951	4.94E-05	227186.4	0.349933	6.00E-05	379.6181	75544	1.259067	0.734735
0.99996	4.05E-05	245535.8	0.329244	5.00E-05	420.2513	83631	1.39385	0.702703
0.999967	3.36E-05	269459.6	0.322468	4.00E-05	454.1005	90383	1.506383	0.828829
0.999976	2.37E-05	288701.1	0.29949	3.00E-05	491.3367	97777	1.629617	0.698699
0.999984	1.62E-05	317002.9	0.290507	2.00E-05	527.2915	104931	1.74885	0.696697
0.999992	8.14E-06	365424.3	0.280139	1.00E-05	605.2513	120445	2.007417	0.638639
0.999993	7.12E-06	392786	0.24607	9.00E-06	648.4322	129038	2.150633	0.677678
0.999994	6.43E-06	388277.8	0.260743	8.00E-06	642.8593	127929	2.13215	0.775776
0.999994	5.58E-06	400704.8	0.25597	7.00E-06	663.794	132095	2.201583	0.653654
0.999995	4.78E-06	418156.4	0.254732	6.00E-06	691.4472	137598	2.2933	0.682683
0.999996	3.99E-06	429071.5	0.266246	5.00E-06	709.5226	141210	2.3535	0.590591
0.999997	3.20E-06	475880.1	0.248595	4.00E-06	787.201	156669	2.61115	0.720721
0.999998	2.38E-06	494180.2	0.251186	3.00E-06	819.9196	163164	2.7194	0.638639
0.999998	1.58E-06	523083	0.254184	2.00E-06	865.2613	172187	2.869783	0.586587
0.999999	7.97E-07	626599.7	0.238138	1.00E-06	1039.332	206827	3.447117	0.581582

训练集	0.5	c
	0.5	p
测试集	0.1	c
	0.9	p
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
0.499428	0.53068	149.3769	0.899794	0.5	1.959799	406	0.006767	0.922923
0.596002	0.408051	6215.095	0.993124	0.4	12.91457	2570	0.042833	1
0.710895	0.293288	10095	0.997123	0.3	17.60804	3504	0.0584	1
0.810447	0.192132	15648.27	0.998627	0.2	26.44221	5262	0.0877	1
0.907157	0.093659	24143.51	0.992334	0.1	39.98492	7957	0.132617	1
0.991354	0.008732	57398.58	0.843628	0.01	94.23618	18753	0.31255	1
0.99918	8.24E-04	111056.3	0.56382	0.001	184.2513	36667	0.611117	0.957958
0.999918	8.27E-05	209420.7	0.418197	1.00E-04	343.8794	68449	1.140817	0.866867
0.999926	7.46E-05	216748.9	0.408308	9.00E-05	359.593	71559	1.19265	0.830831
0.999936	6.41E-05	214543.9	0.414143	8.00E-05	358.3467	71326	1.188767	0.784785
0.999943	5.69E-05	225663	0.409706	7.00E-05	375.5528	74751	1.24585	0.83984
0.999952	4.79E-05	236632.7	0.408665	6.00E-05	390.2764	77665	1.294417	0.778779
0.99996	4.04E-05	246959.7	0.404998	5.00E-05	410.397	81669	1.36115	0.875876
0.999968	3.20E-05	267081.9	0.3719	4.00E-05	445.3568	88626	1.4771	0.815816
0.999975	2.47E-05	288744.6	0.362709	3.00E-05	483.0905	96150	1.6025	0.813814
0.999984	1.63E-05	307150	0.379259	2.00E-05	514.9648	102478	1.707967	0.776777
0.999992	8.03E-06	364250.3	0.335858	1.00E-05	608.608	121114	2.018567	0.668669
0.999993	7.28E-06	372373.8	0.361719	9.00E-06	624.5075	124277	2.071283	0.6997
0.999994	6.15E-06	399035.5	0.334485	8.00E-06	665.2915	132394	2.206567	0.68969
0.999994	5.67E-06	396170.6	0.334219	7.00E-06	661.593	131657	2.194283	0.711712
0.999995	4.77E-06	409632.3	0.342428	6.00E-06	684.9598	136307	2.271783	0.751752
0.999996	4.05E-06	420660.3	0.343142	5.00E-06	707.5678	140806	2.346767	0.734735
0.999997	3.15E-06	462762.7	0.335647	4.00E-06	776.7387	154571	2.576183	0.80981
0.999998	2.36E-06	470073.5	0.336422	3.00E-06	784.6734	156166	2.602767	0.636637
0.999998	1.59E-06	540346.3	0.317901	2.00E-06	899.7085	179072	2.984533	0.652653
0.999999	7.79E-07	642248.7	0.313062	1.00E-06	426.0804	84790	1.413167	0.742743

训练集	0.5	c
	0.5	p
测试集	0.2	c
	0.8	p
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
0.49953	0.530377	149.4774	0.800464	0.5	2.120603	422	0.007033	0.841842
0.607443	0.398335	6098.075	0.985533	0.4	13.91457	2784	0.0464	0.997998
0.712454	0.292196	10101.43	0.995252	0.3	18.44221	3670	0.061167	1
0.809695	0.193394	15381.82	0.998053	0.2	26.38693	5267	0.087783	1
0.907243	0.094008	23415.88	0.994185	0.1	40.15075	7990	0.133167	1
0.99133	0.008756	57319.32	0.86149	0.01	95.19095	18943	0.315717	1
0.999157	8.49E-04	114920.8	0.590958	0.001	190.4171	37893	0.63155	0.963964
0.999919	8.16E-05	208738.6	0.468861	1.00E-04	346.8392	69021	1.15035	0.831832
0.999927	7.35E-05	221076.9	0.453423	9.00E-05	367.2412	73112	1.218533	0.81982
0.999936	6.43E-05	216763.3	0.482166	8.00E-05	362.4171	72121	1.202017	0.875876
0.999943	5.72E-05	224240.8	0.469022	7.00E-05	372.5879	74145	1.23575	0.825826
0.999951	4.89E-05	235333.2	0.464646	6.00E-05	391.4623	77916	1.2986	0.80981
0.99996	4.02E-05	250549.8	0.443464	5.00E-05	422.7387	84125	1.402083	0.761762
0.999968	3.22E-05	268540.2	0.436165	4.00E-05	450.4271	89636	1.493933	0.798799
0.999976	2.38E-05	278070.8	0.448821	3.00E-05	460.7688	91708	1.528467	0.767768
0.999984	1.63E-05	309043	0.427387	2.00E-05	513.7337	102249	1.70415	0.761762
0.999992	8.05E-06	365575.7	0.423363	1.00E-05	604.7487	120345	2.00575	0.681682
0.999993	7.09E-06	377167.3	0.424575	9.00E-06	622.6181	123901	2.065017	0.732733
0.999994	6.42E-06	383537.9	0.423625	8.00E-06	635.7839	126521	2.108683	0.796797
0.999994	5.67E-06	414295.7	0.396879	7.00E-06	686.9296	136699	2.278317	0.711712
0.999995	4.79E-06	408388.9	0.415078	6.00E-06	684.5226	136236	2.2706	0.700701
0.999996	4.06E-06	439646.4	0.40537	5.00E-06	733.1256	145892	2.431533	0.80981
0.999997	3.23E-06	471101	0.39871	4.00E-06	778.3116	154884	2.5814	0.768769
0.999998	2.45E-06	492323.9	0.397111	3.00E-06	815.7739	162339	2.70565	0.743744
0.999998	1.63E-06	547355	0.398499	2.00E-06	906.5729	180408	3.0068	0.640641
0.999999	8.19E-07	645703.1	0.38693	1.00E-06	1067.342	212417	3.540283	0.620621

训练集	0.5	c
	0.5	p
测试集	0.3	c
	0.7	p
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
0.499448	0.529456	148.6834	0.700469	0.5	2.201005	438	0.0073	0.736737
0.603439	0.401989	6083.377	0.978667	0.4	13.36683	2661	0.04435	0.995996
0.71033	0.293779	9926.829	0.992324	0.3	18.25628	3633	0.06055	1
0.809611	0.193233	15081.81	0.996444	0.2	26.66834	5307	0.08845	1
0.9075	0.09338	25458.15	0.993013	0.1	43.74372	8705	0.145083	1
0.991245	0.008823	60125.44	0.860428	0.01	101.3719	20189	0.336483	1
0.999153	8.51E-04	114448.1	0.662426	0.001	191.6533	38172	0.6362	0.971972
0.999916	8.41E-05	201745.8	0.559645	1.00E-04	341.3116	67921	1.132017	0.897898
0.999926	7.45E-05	218610	0.533176	9.00E-05	363.2161	72281	1.204683	0.901902
0.999933	6.72E-05	217172.3	0.543906	8.00E-05	361.7136	71982	1.1997	0.895896
0.999943	5.73E-05	225940.6	0.5237	7.00E-05	376.5176	74930	1.248833	0.815816
0.999951	4.87E-05	241970.3	0.51877	6.00E-05	397.3116	79066	1.317767	0.887888
0.99996	4.05E-05	245810.6	0.522492	5.00E-05	402.1156	80038	1.333967	0.902903
0.999968	3.23E-05	271148.9	0.495812	4.00E-05	444.1005	88376	1.472933	0.771772
0.999975	2.48E-05	282794.2	0.513066	3.00E-05	462.1106	91962	1.5327	0.82983
0.999984	1.59E-05	319202.7	0.507653	2.00E-05	520.8191	103644	1.7274	0.90991
0.999992	7.89E-06	363737.2	0.495184	1.00E-05	595.1859	118443	1.97405	0.761762
0.999993	7.25E-06	379308.8	0.491004	9.00E-06	621.1256	123604	2.060067	0.760761
0.999994	6.42E-06	392360	0.483141	8.00E-06	641.5477	127668	2.1278	0.673674
0.999994	5.60E-06	401929.8	0.49047	7.00E-06	657.3417	130826	2.180433	0.786787
0.999995	4.74E-06	416268.6	0.475375	6.00E-06	679.8894	135298	2.254967	0.842843
0.999996	4.03E-06	444445	0.471854	5.00E-06	728.1457	144903	2.41505	0.747748
0.999997	3.17E-06	464173.3	0.483348	4.00E-06	759.5075	151146	2.5191	0.756757
0.999998	2.37E-06	491751.2	0.473856	3.00E-06	804.3116	160059	2.66765	0.726727
0.999998	1.59E-06	522686.8	0.473735	2.00E-06	857.2161	170586	2.8431	0.717718
0.999999	8.19E-07	624471.1	0.479238	1.00E-06	1021.563	203291	3.388183	0.725726

训练集	0.5	c
	0.5	p
测试集	0.4	c
	0.6	p
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
0.499522	0.529873	147.7387	0.602029	0.5	2.035176	421	0.007017	0.642643
0.602132	0.40252	6290.528	0.971685	0.4	12.28643	2461	0.041017	0.993994
0.711184	0.293026	10142.37	0.989115	0.3	18.28141	3638	0.060633	1
0.810135	0.192014	15518.25	0.996157	0.2	26.8995	5353	0.089217	1
0.906733	0.094114	23850.98	0.994723	0.1	39.80402	7921	0.132017	1
0.991253	0.008819	59417.48	0.885031	0.01	98.28141	19558	0.325967	1
0.999184	8.21E-04	116262.8	0.693804	0.001	191.9648	38217	0.63695	0.967968
0.999918	8.20E-05	205856.7	0.622512	1.00E-04	338.7236	67406	1.123433	0.880881
0.999928	7.23E-05	211840.9	0.609474	9.00E-05	349.191	69489	1.15815	0.882883
0.999935	6.55E-05	221671.8	0.601149	8.00E-05	286.5327	57036	0.9506	0.925926
0.999942	5.84E-05	220378.4	0.609937	7.00E-05	361.201	71894	1.198233	0.901902
0.99995	4.99E-05	237669.7	0.602925	6.00E-05	391.6633	77957	1.299283	0.932933
0.999959	4.08E-05	248245.3	0.587029	5.00E-05	408.2915	81250	1.354167	0.857858
0.999967	3.28E-05	262207.7	0.592864	4.00E-05	433.9497	86356	1.439267	0.838839
0.999976	2.44E-05	280333.4	0.588508	3.00E-05	463.9799	92332	1.538867	0.877878
0.999984	1.61E-05	308706.8	0.574444	2.00E-05	515.9598	102692	1.711533	0.812813
0.999992	8.08E-06	366601.9	0.561788	1.00E-05	612.4724	121883	2.031383	0.872873
0.999993	7.34E-06	367669	0.571461	9.00E-06	608.0704	121038	2.0173	0.864865
0.999994	6.39E-06	385873.2	0.565254	8.00E-06	638.8894	127139	2.118983	0.843844
0.999994	5.68E-06	417806.1	0.558227	7.00E-06	690.3668	137383	2.289717	0.817818
0.999995	4.79E-06	410188.3	0.55957	6.00E-06	677.7839	134879	2.247983	0.788789
0.999996	4.03E-06	451018.4	0.543297	5.00E-06	761.2261	151487	2.524783	0.753754
0.999997	3.19E-06	459961.9	0.547648	4.00E-06	771.8342	153599	2.559983	0.775776
0.999998	2.38E-06	482770.3	0.5554	3.00E-06	897.1809	178546	2.975767	0.731732
0.999998	1.57E-06	539823.1	0.553287	2.00E-06	964.196	191880	3.198	0.803804
0.999999	7.95E-07	620645.9	0.542749	1.00E-06	1049.256	208809	3.48015	0.756757

训练集	0.5	c
	0.5	p
测试集	0.5	c
	0.5	p
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
0.499497	0.531202	143.608	0.4987	0.5	1.979899	394	0.006567	0.538539
0.599318	0.405333	6294.472	0.962249	0.4	13.19095	2641	0.044017	0.98999
0.711283	0.292612	10298.52	0.987324	0.3	18.76884	3735	0.06225	1
0.810641	0.192034	15136.69	0.995392	0.2	27.0804	5389	0.089817	1
0.906923	0.094133	24598.91	0.994879	0.1	42.76382	8522	0.142033	1
0.991266	0.008806	58611.96	0.910534	0.01	100.3216	19968	0.3328	1
0.999164	8.40E-04	110127.1	0.756037	0.001	187.5477	37332	0.6222	0.980981
0.999918	8.23E-05	210147.9	0.672788	1.00E-04	358.0905	71268	1.1878	0.93994
0.999926	7.38E-05	217360.5	0.667084	9.00E-05	370.4623	73722	1.2287	0.924925
0.999934	6.59E-05	216707.4	0.680223	8.00E-05	369.4623	73523	1.225383	0.925926
0.999942	5.80E-05	231569.8	0.664579	7.00E-05	314.9799	62689	1.044817	0.8999
0.999951	4.95E-05	240362.2	0.661707	6.00E-05	412.5226	82100	1.368333	0.944945
0.999959	4.07E-05	247742.3	0.668332	5.00E-05	424.9497	84572	1.409533	0.904905
0.999968	3.23E-05	265902.8	0.653654	4.00E-05	455.9698	90738	1.5123	0.908909
0.999976	2.45E-05	275920.6	0.661762	3.00E-05	472.6935	94066	1.567767	0.87988
0.999984	1.58E-05	315850.9	0.64132	2.00E-05	541.3568	107754	1.7959	0.86987
0.999992	8.06E-06	376979.7	0.634821	1.00E-05	646.196	128593	2.143217	0.82983
0.999993	7.18E-06	381299.2	0.633312	9.00E-06	652.8342	129923	2.165383	0.881882
0.999994	6.49E-06	391096.8	0.633513	8.00E-06	682.9548	135909	2.26515	0.834835
0.999994	5.68E-06	385739.1	0.642869	7.00E-06	653.7236	130091	2.168183	0.915916
0.999995	4.76E-06	427177.7	0.631305	6.00E-06	713.3166	141951	2.36585	0.821822
0.999996	4.00E-06	438041.1	0.634383	5.00E-06	727.5578	144801	2.41335	0.831832
0.999997	3.17E-06	498834.6	0.615535	4.00E-06	846.8342	168541	2.809017	0.808809
0.999998	2.37E-06	497042.6	0.620419	3.00E-06	849.6583	169083	2.81805	0.800801
0.999998	1.63E-06	553016.1	0.617628	2.00E-06	976.3819	194306	3.238433	0.875876
0.999999	7.82E-07	606547.8	0.622884	1.00E-06	1051.598	209269	3.487817	0.80981

训练集	0.5	c
	0.5	p
测试集	0.6	c
	0.4	p
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
0.499516	0.532101	144.0101	0.398861	0.5	2.628141	524	0.008733	0.443443
0.604804	0.400418	6176.106	0.959407	0.4	16.16583	3237	0.05395	0.994995
0.711513	0.292618	10020.49	0.983868	0.3	20.40704	4078	0.067967	1
0.810709	0.192582	15237	0.993994	0.2	30.39196	6052	0.100867	1
0.906985	0.094419	24297.87	0.996278	0.1	46.24121	9207	0.15345	1
0.991297	0.008775	59634.89	0.925835	0.01	96.71357	19255	0.320917	1
0.999153	8.54E-04	113627.5	0.801309	0.001	183.7035	36562	0.609367	0.983984
0.999918	8.22E-05	208380.2	0.742682	1.00E-04	336.995	67067	1.117783	0.947948
0.999928	7.23E-05	216305.4	0.735776	9.00E-05	349.6884	69591	1.15985	0.975976
0.999934	6.63E-05	214417.1	0.739448	8.00E-05	346.5226	68963	1.149383	0.952953
0.999943	5.68E-05	230410.2	0.73143	7.00E-05	372.1106	74053	1.234217	0.931932
0.999952	4.79E-05	240610.8	0.732909	6.00E-05	389.7136	77557	1.292617	0.946947
0.999959	4.10E-05	251281.2	0.728895	5.00E-05	406.4322	80882	1.348033	0.885886
0.999968	3.24E-05	257265.3	0.728517	4.00E-05	415.6131	82713	1.37855	0.852853
0.999976	2.38E-05	284119.7	0.724141	3.00E-05	459.196	91387	1.523117	0.923924
0.999984	1.63E-05	321754.6	0.713523	2.00E-05	519.5528	103395	1.72325	0.871872
0.999992	8.02E-06	380203	0.707185	1.00E-05	613.5427	122102	2.035033	0.928929
0.999993	7.32E-06	381642.5	0.707416	9.00E-06	616.2814	122643	2.04405	0.843844
0.999994	6.38E-06	373728.2	0.716933	8.00E-06	606.4724	120691	2.011517	0.906907
0.999994	5.62E-06	394560.6	0.70886	7.00E-06	9.135678	1828	0.030467	0.876877
0.999995	4.78E-06	405782.9	0.711853	6.00E-06	682.593	135842	2.264033	0.877878
0.999996	4.02E-06	428783.3	0.704619	5.00E-06	721.8342	143647	2.394117	0.891892
0.999997	3.17E-06	447440.4	0.703905	4.00E-06	754.5377	150157	2.502617	0.860861
0.999998	2.44E-06	492925.9	0.701405	3.00E-06	831.3065	165432	2.7572	0.865866
0.999998	1.61E-06	536623	0.700017	2.00E-06	907.191	180540	3.009	0.881882
0.999999	7.89E-07	623932.2	0.696319	1.00E-06	1054.201	209786	3.496433	0.834835

训练集	0.5	c
	0.5	p
测试集	0.7	c
	0.3	p
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
0.499421	0.530705	147.1859	0.300788	0.5	2.19598	437	0.007283	0.352352
0.607273	0.398441	6170.256	0.949055	0.4	16.31156	3249	0.05415	0.98999
0.711602	0.292775	10198.22	0.981942	0.3	18.39698	3668	0.061133	1
0.809789	0.192777	15533.43	0.99243	0.2	26.59296	5299	0.088317	1
0.906926	0.094385	24151.85	0.996056	0.1	40.11055	7990	0.133167	1
0.991251	0.008819	58560.76	0.952118	0.01	95.53769	19019	0.316983	0.998999
0.999154	8.51E-04	107564.9	0.862481	0.001	174.8693	34806	0.5801	0.987988
0.999918	8.23E-05	209740.5	0.796052	1.00E-04	341.7839	68017	1.133617	0.96997
0.999926	7.44E-05	216424.9	0.801334	9.00E-05	354.0402	70464	1.1744	0.940941
0.999934	6.58E-05	221045.8	0.795997	8.00E-05	361.3869	71926	1.198767	0.922923
0.999944	5.61E-05	228854.8	0.796233	7.00E-05	386.6884	76956	1.2826	0.925926
0.999951	4.92E-05	236488.6	0.797421	6.00E-05	420.2613	83635	1.393917	0.933934
0.999959	4.12E-05	258579.6	0.789584	5.00E-05	460.6985	91679	1.527983	0.940941
0.999967	3.28E-05	261405.7	0.799468	4.00E-05	468.794	93299	1.554983	0.960961
0.999976	2.43E-05	282179.7	0.787416	3.00E-05	503.0754	100122	1.6687	0.898899
0.999984	1.64E-05	307254.4	0.793784	2.00E-05	511.9598	101885	1.698083	0.925926
0.999992	8.01E-06	370808.6	0.785363	1.00E-05	607.5779	120920	2.015333	0.922923
0.999993	7.27E-06	393012.9	0.776882	9.00E-06	643.8442	128130	2.1355	0.922923
0.999994	6.36E-06	415835.2	0.770695	8.00E-06	681.8693	135697	2.261617	0.941942
0.999994	5.54E-06	406917.3	0.776983	7.00E-06	666.9045	132716	2.211933	0.924925
0.999995	4.86E-06	431013.1	0.777089	6.00E-06	706.4221	140599	2.343317	0.937938
0.999996	3.98E-06	438260.3	0.77977	5.00E-06	717.9296	142869	2.38115	0.862863
0.999997	3.21E-06	458926.4	0.781052	4.00E-06	752.5427	149760	2.496	0.903904
0.999998	2.36E-06	488579	0.774382	3.00E-06	798.2261	158850	2.6475	0.863864
0.999998	1.60E-06	540344.3	0.772491	2.00E-06	884.0804	175938	2.9323	0.910911
0.999999	7.86E-07	620509.7	0.771445	1.00E-06	1014.653	201922	3.365367	0.902903

训练集	0.5	c
	0.5	p
测试集	0.8	c
	0.2	p
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
0.499471	0.530313	147.6382	0.199763	0.5	2.376884	473	0.007883	0.238238
0.607675	0.396889	6484.824	0.94661	0.4	15.98492	3181	0.053017	0.983984
0.711471	0.291903	10131.84	0.980594	0.3	18.45729	3705	0.06175	1
0.81034	0.192206	15523	0.991534	0.2	26.90452	5355	0.08925	1
0.906968	0.094075	24210.84	0.9967	0.1	40.17085	8010	0.1335	1
0.991286	0.008796	59353.21	0.962002	0.01	97.15075	19333	0.322217	1
0.999151	8.54E-04	112414.9	0.906087	0.001	184.1055	36669	0.61115	0.995996
0.999918	8.26E-05	202288.7	0.871822	1.00E-04	334.5427	66574	1.109567	0.968969
0.999925	7.53E-05	210094	0.871459	9.00E-05	346.2462	68919	1.14865	0.984985
0.999934	6.64E-05	215015	0.873336	8.00E-05	355.1508	70675	1.177917	0.961962
0.999942	5.78E-05	224129	0.866324	7.00E-05	384.7337	76562	1.276033	0.967968
0.999951	4.87E-05	238662.6	0.86648	6.00E-05	405.9497	80804	1.346733	0.961962
0.99996	4.02E-05	253117.2	0.864352	5.00E-05	435.7085	86707	1.445117	0.96997
0.999968	3.19E-05	265370	0.861314	4.00E-05	450.799	89725	1.495417	0.976977
0.999976	2.45E-05	288700.1	0.857496	3.00E-05	492.3266	97974	1.6329	0.932933
0.999984	1.62E-05	313644	0.859105	2.00E-05	537.2915	106955	1.782583	0.950951
0.999992	7.99E-06	383179.2	0.85316	1.00E-05	654.6432	130275	2.17125	0.933934
0.999993	7.34E-06	372760.6	0.855896	9.00E-06	623.3518	124063	2.067717	0.935936
0.999994	6.44E-06	394640.4	0.851756	8.00E-06	656.5628	130656	2.1776	0.937938
0.999994	5.63E-06	415909.8	0.851007	7.00E-06	697.9497	138892	2.314867	0.930931
0.999995	4.67E-06	408631.4	0.853175	6.00E-06	683.7839	136089	2.26815	0.933934
0.999996	4.00E-06	444052.7	0.851384	5.00E-06	742.7236	147802	2.463367	0.916917
0.999997	3.24E-06	477415.8	0.848346	4.00E-06	809.5075	161092	2.684867	0.937938
0.999998	2.37E-06	498906.6	0.847229	3.00E-06	827.4975	164672	2.744533	0.936937
0.999998	1.55E-06	534051.6	0.851651	2.00E-06	904.5025	179997	2.99995	0.926927
0.999999	7.72E-07	625129.6	0.84736	1.00E-06	1062.995	211551	3.52585	0.913914

训练集	0.5	c
	0.5	p
测试集	0.9	c
	0.1	p
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
0.499505	0.530991	146.9146	0.100714	0.5	2.276382	453	0.00755	0.129129
0.604423	0.400491	6527.412	0.936273	0.4	13.9196	2770	0.046167	0.983984
0.711739	0.292659	10246.46	0.976283	0.3	18.13065	3624	0.0604	1
0.809975	0.192286	15367.05	0.991353	0.2	26.42714	5259	0.08765	1
0.906997	0.094119	23795.91	0.997656	0.1	39.42211	7845	0.13075	1
0.991311	0.008757	58745.15	0.981564	0.01	97.52764	19424	0.323733	1
0.999153	8.52E-04	112087.3	0.952505	0.001	184.1457	36645	0.61075	1
0.999918	8.26E-05	201300.8	0.938179	1.00E-04	337.9849	67261	1.121017	0.990991
0.999928	7.26E-05	216170.4	0.933049	9.00E-05	370.8894	73808	1.230133	0.986987
0.999935	6.48E-05	222313.3	0.933466	8.00E-05	381.5879	75937	1.265617	0.991992
0.999943	5.75E-05	221232.3	0.934477	7.00E-05	374.7588	74592	1.2432	0.976977
0.999951	4.95E-05	232360.4	0.934653	6.00E-05	394.2714	78476	1.307933	0.980981
0.999959	4.11E-05	260700.7	0.930564	5.00E-05	441.7538	87925	1.465417	0.983984
0.999968	3.20E-05	264792.1	0.930936	4.00E-05	366.3216	72929	1.215483	0.983984
0.999976	2.45E-05	282872.8	0.930287	3.00E-05	476.1809	94776	1.5796	0.970971
0.999984	1.61E-05	327235	0.928552	2.00E-05	553.196	110117	1.835283	0.967968
0.999992	8.01E-06	356469.7	0.929085	1.00E-05	601.4774	119726	1.995433	0.972973
0.999993	7.13E-06	396853.3	0.925171	9.00E-06	670.5075	133431	2.22385	0.966967
0.999994	6.21E-06	394890.9	0.925921	8.00E-06	664.7236	132280	2.204667	0.967968
0.999994	5.65E-06	401407	0.928184	7.00E-06	676.4724	134618	2.243633	0.97998
0.999995	4.87E-06	421662	0.927163	6.00E-06	710.4121	141372	2.3562	0.978979
0.999996	4.16E-06	428640.5	0.925946	5.00E-06	723.0402	143885	2.398083	0.971972
0.999997	3.21E-06	465639.4	0.924407	4.00E-06	784.3417	156084	2.6014	0.968969
0.999998	2.35E-06	483048.2	0.926535	3.00E-06	813.9196	161970	2.6995	0.95996
0.999998	1.54E-06	528076.3	0.925337	2.00E-06	901.191	179338	2.988967	0.961962
0.999999	8.11E-07	656599.3	0.923129	1.00E-06	1109.442	220779	3.67965	0.960961

训练集	0.5	c
	0.5	p
测试集	1	c
	0	p
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
0.499512	0.530374	148.8543	0	0.5	2.155779	429	0.00715	0
0.605797	0.399978	6306.643	0.931228	0.4	13.97487	2781	0.04635	0.984985
0.711183	0.29228	10420.39	0.974985	0.3	18.71357	3724	0.062067	1
0.810918	0.191953	15286.05	0.991293	0.2	26.24121	5238	0.0873	1
0.906977	0.093858	24414.25	0.998174	0.1	42.22111	8403	0.14005	1
0.991248	0.008814	59492.45	1	0.01	102.4221	20382	0.3397	1
0.999154	8.51E-04	110956.2	1	0.001	188.6834	37548	0.6258	1
0.999919	8.12E-05	206577	1	1.00E-04	345.6482	68784	1.1464	1
0.999926	7.45E-05	225841.7	1	9.00E-05	380.3417	75692	1.261533	1
0.999936	6.46E-05	214257.2	1	8.00E-05	365.0302	72643	1.210717	1
0.999943	5.75E-05	227719.7	1	7.00E-05	380.2864	75677	1.261283	1
0.999952	4.83E-05	237107	1	6.00E-05	398.6633	79334	1.322233	1
0.99996	4.04E-05	245581.8	1	5.00E-05	414.191	82424	1.373733	1
0.999967	3.30E-05	266492.5	1	4.00E-05	449.3116	89413	1.490217	1
0.999976	2.40E-05	282278.7	1	3.00E-05	474.9598	94517	1.575283	1
0.999984	1.63E-05	323231	1	2.00E-05	543.8794	108232	1.803867	1
0.999992	7.97E-06	368836.5	1	1.00E-05	621.2161	123638	2.060633	1
0.999993	7.13E-06	377565.8	1	9.00E-06	635.4623	126457	2.107617	1
0.999994	6.41E-06	387094.8	1	8.00E-06	653.1658	130044	2.1674	1
0.999994	5.60E-06	404034.1	1	7.00E-06	680.8241	135499	2.258317	1
0.999995	4.73E-06	415028.5	1	6.00E-06	700.2714	139385	2.323083	1
0.999996	3.96E-06	449725.6	1	5.00E-06	679.5126	135255	2.25425	1
0.999997	3.13E-06	463744.1	1	4.00E-06	781.4372	155522	2.592033	1
0.999998	2.39E-06	492239.1	1	3.00E-06	831.1809	165420	2.757	1
0.999998	1.60E-06	560042.5	1	2.00E-06	941.6432	187389	3.12315	1
0.999999	8.12E-07	634245.9	1	1.00E-06	1067.819	212512	3.541867	1

转载自原文链接, 如需删除请联系管理员。

原文链接：测试集的构成比例对网络分类性能的影响cp，转载请注明来源！