在上一次实验中已经由数据证实了可以用特征迭代次数去区分minst数据集中的0和1时有可能的,因为50张0和50张1的分类正确率可以打到99%。
本文将迭代次数按大小排列,比较了图片和迭代次数的关系
a | b | 迭代次数 | 时间ms | 时间min | ||
![]() |
0*34 | 0.504092 | 0.496892 | 31808.4 | 76592 | 1.276533 |
![]() |
0*26 | 0.504145 | 0.496745 | 34094.86 | 83691 | 1.39485 |
![]() |
0*41 | 0.504069 | 0.496969 | 34775.88 | 84630 | 1.4105 |
![]() |
0*36 | 0.50385 | 0.49755 | 41926.72 | 99082 | 1.651367 |
![]() |
0*27 | 0.503765 | 0.498165 | 47928.64 | 111719 | 1.861983 |
![]() |
0*38 | 0.503758 | 0.498158 | 48533.77 | 110892 | 1.8482 |
![]() |
0*37 | 0.503619 | 0.498219 | 50003.07 | 114741 | 1.91235 |
![]() |
0*3 | 0.503581 | 0.498881 | 50215.99 | 111575 | 1.859583 |
![]() |
0*18 | 0.50375 | 0.49815 | 52884.1 | 120598 | 2.009967 |
![]() |
0*30 | 0.503511 | 0.498611 | 52966.01 | 123203 | 2.053383 |
![]() |
0*6 | 0.503368 | 0.498868 | 55467.06 | 125845 | 2.097417 |
![]() |
0*8 | 0.503476 | 0.499077 | 56096.01 | 125986 | 2.099767 |
![]() |
0*28 | 0.503291 | 0.499491 | 60411.63 | 135295 | 2.254917 |
![]() |
0*7 | 0.503256 | 0.499456 | 64569.89 | 142200 | 2.37 |
![]() |
0*47 | 0.503086 | 0.499886 | 67707.4 | 147675 | 2.46125 |
![]() |
0*4 | 0.503037 | 0.50084 | 70608.87 | 155951 | 2.599183 |
![]() |
0*17 | 0.503267 | 0.501109 | 72017.99 | 156489 | 2.60815 |
![]() |
0*44 | 0.503364 | 0.500174 | 72019.38 | 156790 | 2.613167 |
![]() |
0*45 | 0.502973 | 0.500373 | 72615.19 | 157305 | 2.62175 |
![]() |
0*9 | 0.50308 | 0.50048 | 72920.47 | 157379 | 2.622983 |
![]() |
0*50 | 0.503098 | 0.500834 | 73901.24 | 162317 | 2.705283 |
![]() |
0*24 | 0.502874 | 0.501075 | 74839.92 | 162557 | 2.709283 |
![]() |
0*23 | 0.502903 | 0.501203 | 74969.03 | 162938 | 2.715633 |
![]() |
0*19 | 0.502897 | 0.50074 | 75722.1 | 165457 | 2.757617 |
![]() |
0*13 | 0.502945 | 0.501067 | 76511.02 | 170373 | 2.83955 |
![]() |
0*20 | 0.502779 | 0.500913 | 76613.24 | 165682 | 2.761367 |
![]() |
0*21 | 0.502754 | 0.501154 | 78554.36 | 170378 | 2.839633 |
![]() |
0*16 | 0.502863 | 0.500763 | 79751.51 | 186547 | 3.109117 |
![]() |
0*46 | 0.502553 | 0.501353 | 80005.62 | 171478 | 2.857967 |
![]() |
0*32 | 0.50252 | 0.50182 | 83654.15 | 180615 | 3.01025 |
![]() |
0*2 | 0.502641 | 0.501541 | 84550.08 | 180167 | 3.002783 |
![]() |
0*22 | 0.502473 | 0.501673 | 86017.23 | 185109 | 3.08515 |
![]() |
0*11 | 0.502387 | 0.501988 | 87989.04 | 187844 | 3.130733 |
![]() |
0*42 | 0.50239 | 0.50209 | 90076.06 | 193220 | 3.220333 |
![]() |
0*43 | 0.502098 | 0.502498 | 93610.47 | 201571 | 3.359517 |
![]() |
0*10 | 0.502299 | 0.502511 | 93919.15 | 213039 | 3.55065 |
![]() |
0*12 | 0.502466 | 0.502768 | 95686.91 | 203531 | 3.392183 |
![]() |
0*48 | 0.502223 | 0.502623 | 95741.22 | 201800 | 3.363333 |
![]() |
0*1 | 0.502221 | 0.503021 | 97086.1 | 204735 | 3.41225 |
![]() |
0*49 | 0.502118 | 0.503218 | 97346.76 | 204799 | 3.413317 |
![]() |
0*40 | 0.502179 | 0.503389 | 97427.12 | 204526 | 3.408767 |
![]() |
0*15 | 0.502229 | 0.502829 | 98331.42 | 216606 | 3.6101 |
![]() |
0*39 | 0.502076 | 0.503176 | 100795.5 | 211763 | 3.529383 |
![]() |
0*14 | 0.502212 | 0.503448 | 103378.9 | 224424 | 3.7404 |
![]() |
0*25 | 0.501949 | 0.503349 | 103918.8 | 218470 | 3.641167 |
![]() |
0*31 | 0.501894 | 0.50371 | 104757.8 | 227289 | 3.78815 |
![]() |
0*29 | 0.502027 | 0.503527 | 112442.9 | 234996 | 3.9166 |
![]() |
0*5 | 0.501786 | 0.503786 | 112862.4 | 239529 | 3.99215 |
![]() |
0*35 | 0.502023 | 0.504843 | 116743.7 | 243684 | 4.0614 |
![]() |
0*33 | 0.501833 | 0.504933 | 119631 | 248238 | 4.1373 |
a | b | 迭代次数 | 时间ms | 时间min | ||
![]() |
1*17 | 0.501511 | 0.505356 | 113085.8 | 235103 | 3.918383 |
![]() |
1*30 | 0.508745 | 0.512645 | 120761.4 | 243498 | 4.0583 |
![]() |
1*14 | 0.501444 | 0.504744 | 122373 | 255182 | 4.253033 |
![]() |
1*6 | 0.504215 | 0.508821 | 123439.5 | 261633 | 4.36055 |
![]() |
1*12 | 0.503918 | 0.509033 | 129424.3 | 267415 | 4.456917 |
![]() |
1*35 | 0.501017 | 0.506317 | 132802 | 266470 | 4.441167 |
![]() |
1*5 | 0.503435 | 0.508935 | 134629.3 | 282495 | 4.70825 |
![]() |
1*46 | 0.500885 | 0.506685 | 135243.7 | 271229 | 4.520483 |
![]() |
1*1 | 0.505469 | 0.512169 | 135295.5 | 302633 | 5.043883 |
![]() |
1*11 | 0.503724 | 0.508224 | 135336.4 | 281378 | 4.689633 |
![]() |
1*10 | 0.503584 | 0.5092 | 135825.9 | 282623 | 4.710383 |
![]() |
1*9 | 0.503415 | 0.508715 | 136958.3 | 281435 | 4.690583 |
![]() |
1*22 | 0.508187 | 0.514087 | 136999 | 279906 | 4.6651 |
![]() |
1*32 | 0.503518 | 0.508822 | 138253.6 | 277344 | 4.6224 |
![]() |
1*2 | 0.503662 | 0.508362 | 138619.9 | 174259 | 2.904317 |
![]() |
1*50 | 0.50053 | 0.50723 | 138767.7 | 277808 | 4.630133 |
![]() |
1*3 | 0.500866 | 0.506566 | 139057.3 | 317648 | 5.294133 |
![]() |
1*38 | 0.505616 | 0.511516 | 139506.7 | 279306 | 4.6551 |
![]() |
1*44 | 0.505531 | 0.512431 | 139982.5 | 280296 | 4.6716 |
![]() |
1*8 | 0.50095 | 0.50655 | 140646.4 | 287694 | 4.7949 |
![]() |
1*47 | 0.500705 | 0.507105 | 140681 | 281496 | 4.6916 |
![]() |
1*15 | 0.505747 | 0.512147 | 142043.2 | 293122 | 4.885367 |
![]() |
1*7 | 0.50086 | 0.50666 | 142070.6 | 295809 | 4.93015 |
![]() |
1*18 | 0.505487 | 0.512087 | 142772 | 292768 | 4.879467 |
![]() |
1*43 | 0.500936 | 0.506336 | 143081.6 | 287021 | 4.783683 |
![]() |
1*23 | 0.507865 | 0.515165 | 144089.6 | 294738 | 4.9123 |
![]() |
1*36 | 0.508275 | 0.514884 | 144472.9 | 289326 | 4.8221 |
![]() |
1*4 | 0.500699 | 0.506899 | 144642.3 | 316810 | 5.280167 |
![]() |
1*26 | 0.50094 | 0.50664 | 144746.8 | 295797 | 4.92995 |
![]() |
1*34 | 0.500629 | 0.507429 | 145520.8 | 291038 | 4.850633 |
![]() |
1*13 | 0.503092 | 0.509792 | 145810.5 | 299617 | 4.993617 |
![]() |
1*27 | 0.500582 | 0.507782 | 145828.5 | 296504 | 4.941733 |
![]() |
1*37 | 0.500709 | 0.507109 | 146633.2 | 292617 | 4.87695 |
![]() |
1*16 | 0.50313 | 0.51003 | 147017 | 304313 | 5.071883 |
![]() |
1*39 | 0.500637 | 0.507737 | 147505.9 | 300222 | 5.0037 |
![]() |
1*48 | 0.503003 | 0.510203 | 147777.1 | 295941 | 4.93235 |
![]() |
1*33 | 0.500784 | 0.507084 | 148811.9 | 297957 | 4.96595 |
![]() |
1*25 | 0.502923 | 0.510123 | 149811 | 304366 | 5.072767 |
![]() |
1*45 | 0.502981 | 0.51058 | 149999.9 | 298816 | 4.980267 |
![]() |
1*19 | 0.500294 | 0.508294 | 150626.6 | 306709 | 5.111817 |
![]() |
1*28 | 0.502889 | 0.510289 | 151403.7 | 306664 | 5.111067 |
![]() |
1*49 | 0.500599 | 0.507599 | 152002.6 | 304685 | 5.078083 |
![]() |
1*24 | 0.500305 | 0.508405 | 153926.4 | 312512 | 5.208533 |
![]() |
1*41 | 0.500378 | 0.508078 | 154345 | 307673 | 5.127883 |
![]() |
1*29 | 0.500624 | 0.507624 | 155807.4 | 313365 | 5.22275 |
![]() |
1*40 | 0.505116 | 0.513416 | 156111.3 | 311209 | 5.186817 |
![]() |
1*42 | 0.500343 | 0.508143 | 156573.8 | 311144 | 5.185733 |
![]() |
1*20 | 0.500362 | 0.508262 | 159081.9 | 323487 | 5.39145 |
![]() |
1*31 | 0.502814 | 0.510814 | 159971.4 | 317844 | 5.2974 |
![]() |
1*21 | 0.500206 | 0.508706 | 162120.7 | 242122 | 4.035367 |
0的迭代次数的范围是31808-119631
1的迭代次数的范围是113085-162120
1只有第17号图片是小于119631的,所以这100张图片的分类正确率可能达到99%.另外很明显的可以发现外观比较接近的图片确实迭代次数也比较接近
比如这几张长的比较饱满的0,都被分在了一起,
这几张长的比较像的也被分到了一起,
不过也有失手的时候,这张夹在44和47中的8就明显的有问题。
这种图就是神来之笔了,很细微的差别都找到了。
虽然统计样本的数量还是太少,不过貌似迭代次数还可以作为比较图片匹配度的一个很有力的工具。
本文所用数据同《用共振频率去进行图片分类的尝试》中的数据