取无理数的前3万个有效数字,每10个数字组成一张图片
前2500张图片做训练集,后500张图片做测试集,让x分别等于
3 | 5 | 6 | 7 | 8 | 10 | 11 | 12 | 13 | 14 | 15 |
17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 26 | 27 | 28 |
29 | 30 | 31 | 32 | 33 | 34 | 35 | 37 | 38 | 39 | 40 |
41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 50 | 51 | 52 |
53 | 54 | 55 | 56 | 57 | 58 | 59 | 60 | 61 | 62 | 63 |
65 | 66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | 74 | 75 |
76 | 77 | 78 | 79 | 80 | 82 | 83 | 84 | 85 | 86 | 87 |
88 | 89 | 90 | 91 | 92 | 93 | 94 | 95 | 96 | 97 | 98 |
99 |
让网络的收敛标准δ等于0.5到1e-4的16个值,每个值收敛199次取平均值,
统计当δ=1e-4时的迭代次数和分类准确率与x的关系。
实验共收敛了89*16*199次
比较迭代次数
根号x | 迭代次数 | 根号x | 迭代次数 | 根号x | 迭代次数 |
8 | 783088.3 | 28 | 702341.7 | 60 | 672742 |
62 | 774492.5 | 50 | 701950.8 | 98 | 671972.6 |
86 | 764687.5 | 51 | 700513.9 | 32 | 671520.3 |
88 | 754498.3 | 12 | 696393.1 | 53 | 671064.6 |
83 | 751454.1 | 78 | 696308.6 | 48 | 668988.9 |
55 | 749043.1 | 10 | 694794.8 | 19 | 666679 |
52 | 746851.5 | 15 | 694355.3 | 65 | 666596 |
47 | 746827.6 | 44 | 692737 | 41 | 666001.6 |
34 | 740813.1 | 31 | 692346.9 | 69 | 665235 |
84 | 740406.4 | 61 | 692247.9 | 54 | 662053.5 |
11 | 740245.7 | 46 | 692196.6 | 18 | 660403.3 |
29 | 735647.9 | 90 | 691722.7 | 6 | 659161.1 |
67 | 732879.5 | 33 | 690411.6 | 22 | 659072.1 |
59 | 732409 | 80 | 689793.3 | 72 | 658334.7 |
42 | 730669.6 | 77 | 689593.9 | 82 | 654149.6 |
93 | 729768.8 | 40 | 688906.4 | 99 | 652916.4 |
68 | 729676.9 | 89 | 684409.9 | 20 | 649598.9 |
63 | 727313.1 | 97 | 684279.4 | 23 | 646539.7 |
87 | 726341.6 | 24 | 682478.4 | 5 | 643888.1 |
26 | 722887.4 | 58 | 682452.2 | 66 | 643457.9 |
38 | 715226.9 | 71 | 681690.5 | 14 | 642825.4 |
76 | 711597.1 | 70 | 681167.3 | 45 | 629977.4 |
56 | 708782.5 | 96 | 681095.3 | 75 | 629553.8 |
17 | 708562.8 | 35 | 679045.2 | 3 | 625240.1 |
74 | 706798.2 | 27 | 678591.1 | 30 | 625240.1 |
13 | 705637.6 | 85 | 677746.7 | 39 | 622623.4 |
7 | 705298.8 | 73 | 676184.8 | 91 | 621037 |
43 | 704949.2 | 57 | 675862 | 92 | 619325.5 |
21 | 704103.6 | 95 | 674251.1 | 94 | 614488.7 |
37 | 702753.7 | 79 | 673434.8 |
迭代次数最大的是根号8,迭代次数最小的是根号94.
假设1:完全相同的两个对象无法被分成两类,与之对应的分类迭代次数为无穷大,分类准确率是50%,50%。相等收敛标准下迭代次数越大表明二者差异越小。
按照假设1,这个结果表明根号2的结构与根号94差异最大,而与根号8差异最小。
再比较分类准确率
根号x | 分类准确率 | 根号x | 分类准确率 | 根号x | 分类准确率 |
99 | 0.5161005 | 35 | 0.50381407 | 15 | 0.49998492 |
74 | 0.51311558 | 14 | 0.50380905 | 32 | 0.49952764 |
79 | 0.51069849 | 37 | 0.50368342 | 58 | 0.49948241 |
69 | 0.51055779 | 31 | 0.50361809 | 43 | 0.49939698 |
34 | 0.51042714 | 91 | 0.50271859 | 88 | 0.49869849 |
23 | 0.51020101 | 87 | 0.50252261 | 77 | 0.49860804 |
96 | 0.50925126 | 20 | 0.50240201 | 24 | 0.49809045 |
5 | 0.50867337 | 19 | 0.50236181 | 75 | 0.49795477 |
66 | 0.50859799 | 40 | 0.50234673 | 46 | 0.49762312 |
44 | 0.50835176 | 94 | 0.50205025 | 29 | 0.49750754 |
3 | 0.50807035 | 12 | 0.50203518 | 93 | 0.49747236 |
30 | 0.50807035 | 82 | 0.50203015 | 63 | 0.49735678 |
17 | 0.50775377 | 89 | 0.50193467 | 55 | 0.49731156 |
28 | 0.50769347 | 92 | 0.50179899 | 78 | 0.4970804 |
21 | 0.50766834 | 73 | 0.50162312 | 33 | 0.49652261 |
52 | 0.50729648 | 50 | 0.50157286 | 56 | 0.49651759 |
85 | 0.50729648 | 76 | 0.50153266 | 68 | 0.49651759 |
13 | 0.50723618 | 26 | 0.50148241 | 61 | 0.49647739 |
42 | 0.50712563 | 53 | 0.50142211 | 83 | 0.49605025 |
72 | 0.50709548 | 84 | 0.50133166 | 8 | 0.49590452 |
39 | 0.50667839 | 6 | 0.50131156 | 98 | 0.49525628 |
27 | 0.50647739 | 65 | 0.50085427 | 97 | 0.49509548 |
60 | 0.50603015 | 62 | 0.50074874 | 67 | 0.49441206 |
7 | 0.50596985 | 54 | 0.50054271 | 71 | 0.49424623 |
90 | 0.5058191 | 48 | 0.50050754 | 70 | 0.49405025 |
10 | 0.50578392 | 45 | 0.50046734 | 47 | 0.49370352 |
57 | 0.50446734 | 95 | 0.50039196 | 11 | 0.49312563 |
59 | 0.50443216 | 41 | 0.50027136 | 51 | 0.49240201 |
38 | 0.50405025 | 22 | 0.50018593 | 18 | 0.48945729 |
86 | 0.5039397 | 80 | 0.50008543 |
有60个数的分类准确率的平均值是大于50%的,其中最大的是根号99,可以达到51.61%。
N[sqr(2),30000]
比如根号2和根号3的数据是
g2g3 | ||||||||||
训练集 | 0-2500 | |||||||||
测试集 | 2500-3000 | |||||||||
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 | 最大值p-max | 迭代次数标准差 | pave标准差 |
0.500338192 | 0.499239469 | 20.55778894 | 0.500080402 | 0.5 | 5.256281407 | 1052 | 0.017533333 | 0.524 | 25.22970705 | 0.003709768 |
0.467823775 | 0.5322995 | 11919.66834 | 0.516839196 | 0.4 | 133.9045226 | 26653 | 0.444216667 | 0.531 | 969.2051142 | 0.007147707 |
0.555233687 | 0.444791463 | 57628.05528 | 0.519824121 | 0.3 | 628.3266332 | 125043 | 2.08405 | 0.536 | 23085.578 | 0.004646739 |
0.616973437 | 0.382998604 | 135998.9497 | 0.512130653 | 0.2 | 1476.110553 | 293751 | 4.89585 | 0.539 | 19865.73736 | 0.009454095 |
0.733029551 | 0.266970767 | 191893.0201 | 0.513376884 | 0.1 | 2068.492462 | 411637 | 6.860616667 | 0.554 | 25796.91008 | 0.01233183 |
0.694724672 | 0.305275328 | 303482.5276 | 0.507361809 | 0.01 | 3274.658291 | 651664 | 10.86106667 | 0.556 | 21025.07752 | 0.014651041 |
0.642949758 | 0.357050292 | 418999.3568 | 0.506165829 | 0.001 | 4491.703518 | 893852 | 14.89753333 | 0.545 | 47223.97265 | 0.015214963 |
0.637957293 | 0.362042703 | 420436.4372 | 0.507 | 9.00E-04 | 4506.653266 | 896830 | 14.94716667 | 0.543 | 42325.38175 | 0.014135738 |
0.562718115 | 0.437281893 | 431819.0553 | 0.50618593 | 8.00E-04 | 4635.075377 | 922391 | 15.37318333 | 0.557 | 50420.4394 | 0.01460779 |
0.597865141 | 0.402134839 | 440818.8392 | 0.502422111 | 7.00E-04 | 4669.361809 | 929210 | 15.48683333 | 0.551 | 55378.81957 | 0.016427478 |
0.562741749 | 0.437258266 | 458893.4372 | 0.505748744 | 6.00E-04 | 4868.668342 | 968868 | 16.1478 | 0.554 | 61587.45673 | 0.016143427 |
0.592875786 | 0.407124209 | 462545.7789 | 0.506160804 | 5.00E-04 | 4901.522613 | 975407 | 16.25678333 | 0.54 | 55633.63555 | 0.014304452 |
0.612977949 | 0.387022063 | 490165.3065 | 0.50620603 | 4.00E-04 | 5200.180905 | 1034847 | 17.24745 | 0.546 | 75881.90078 | 0.013377877 |
0.567800493 | 0.432199477 | 515554.4271 | 0.505447236 | 3.00E-04 | 5476.834171 | 1089896 | 18.16493333 | 0.556 | 77536.28047 | 0.015290684 |
0.607998131 | 0.392001886 | 576056.5377 | 0.505371859 | 2.00E-04 | 6097.01005 | 1213314 | 20.2219 | 0.54 | 95671.37734 | 0.014181608 |
0.638164184 | 0.361835814 | 625240.1055 | 0.508070352 | 1.00E-04 | 6616.929648 | 1316774 | 21.94623333 | 0.549 | 115278.666 | 0.015353296 |