验算神经网络谐振子模型的第二组数据
驗算用二分類神經網絡估算多分類神經網絡迭代次數n和收斂誤差δ的經驗關系式
這次同樣統計了11個網絡，包括6個二分類網絡(0,1),(0,2),(0,3) ,(1,2) ,(1,3) ,(2,3),4個三分類網絡(0,1,2), (0,1,3), (0,2,3), (1,2,3)和1個4分類網絡(0,1,2,3).所不同的是這次用的是11*11的圖片。隱藏層節點數是30個，網絡結構為
統計了0.5到1e-5共25個收斂誤差，每個收斂誤差測量199次，計算迭代次數的平均值，參數設置方案
static double ret=0.1; //學習率
static double weix=1000;
在收斂誤差一致的前提下得到的數據為
| 0*1 | 0*2 | 0*3 | 1*2 | 1*3 | 2*3 | 0*1*2 | 0*1*3 | 0*2*3 | 1*2*3 | 0*1*2*3 | |
| δ | 迭代次數n | 迭代次數n | 迭代次數n | 迭代次數n | 迭代次數n | 迭代次數n | 迭代次數n | 迭代次數n | 迭代次數n | 迭代次數n | 迭代次數n |
| 0.5 | 7.658291 | 7.180905 | 7.98995 | 9.145729 | 7.854271 | 9.065327 | 100.407 | 106.196 | 145.1307 | 162.2462 | 326.7638 |
| 0.4 | 107.5628 | 149.9598 | 153.0804 | 144.1508 | 141.005 | 170.6985 | 331.3668 | 335.9347 | 421.0704 | 440.2412 | 721.9246 |
| 0.3 | 137.3467 | 191.2613 | 198.5276 | 181.5829 | 187.804 | 223.0302 | 414.7839 | 409.191 | 492.196 | 543.7186 | 823.9698 |
| 0.2 | 175.3015 | 221.4774 | 235.8794 | 219.7186 | 242.5176 | 272.9347 | 555.6633 | 597.0603 | 588.7236 | 629.2362 | 1083.603 |
| 0.1 | 221.8141 | 274.1859 | 298.402 | 300.6935 | 301.6834 | 347.3467 | 842.0503 | 807.1106 | 741.4422 | 795.804 | 1551.894 |
| 0.01 | 516.5276 | 549.3367 | 574.8442 | 577.8894 | 512.593 | 708.6683 | 1739.437 | 1718.91 | 1876.834 | 1948.417 | 3845.523 |
| 0.001 | 1715.186 | 1396.06 | 1371.965 | 1237.156 | 1470.477 | 1611.97 | 5517.618 | 5596.447 | 6646.884 | 6018.06 | 14272.66 |
| 9.00E-04 | 1793.854 | 1476.01 | 1392.015 | 1257.367 | 1490.608 | 1649.337 | 6249.704 | 6054.251 | 7209.432 | 6864.397 | 15653.62 |
| 8.00E-04 | 1849.462 | 1584.503 | 1424.432 | 1293.899 | 1560.729 | 1670.714 | 6943.226 | 6525.894 | 7979.347 | 8011.693 | 17548.21 |
| 7.00E-04 | 1992.854 | 1821.241 | 1476.99 | 1335.055 | 1666.693 | 1699.688 | 7463.02 | 7105.452 | 8773.754 | 9691.387 | 20475.94 |
| 6.00E-04 | 2199.874 | 2032.91 | 1550.196 | 1481.719 | 1922.759 | 1730.899 | 8305.452 | 8406.362 | 9861.97 | 11980.4 | 23457.43 |
| 5.00E-04 | 2465.241 | 2252.844 | 1706.925 | 1690.965 | 2230.563 | 1816.281 | 10042.53 | 10061.64 | 11079.37 | 14170.71 | 27144.43 |
| 4.00E-04 | 2700.749 | 2398.352 | 1956.995 | 1899.246 | 2450.467 | 1984.548 | 12510.08 | 12423.29 | 13147.52 | 18091.9 | 30729.05 |
| 3.00E-04 | 3409.794 | 2595.955 | 2258.196 | 2176.151 | 2828.312 | 2249.307 | 15432.62 | 14539.76 | 15630.1 | 22257.08 | 36344.84 |
| 2.00E-04 | 4697.849 | 3333.241 | 3280.136 | 3429.779 | 3950.879 | 2811.397 | 20695.39 | 17663.18 | 19632.18 | 29997.2 | 47360.86 |
| 1.00E-04 | 5720.523 | 5329.407 | 4995.824 | 5190.538 | 6355.276 | 8262.528 | 32783.53 | 31727.12 | 31181.01 | 42632.25 | 71855.67 |
| 9.00E-05 | 6267.256 | 5674.975 | 5439.764 | 5227.673 | 7115.095 | 9929.201 | 34601.95 | 33933.1 | 32990.93 | 46021.2 | 75395.2 |
| 8.00E-05 | 7108.06 | 6059.955 | 5798.221 | 5408.719 | 8040.196 | 12154.21 | 37459.96 | 37377.31 | 33980.8 | 48427.32 | 80703.34 |
| 7.00E-05 | 7949.99 | 6618.236 | 6356.955 | 5627.955 | 9393.97 | 14009.3 | 40344.16 | 40377.47 | 35928.17 | 51737.93 | 87493.71 |
| 6.00E-05 | 8649.397 | 7186.98 | 7000.035 | 6146.859 | 9737.96 | 15939.2 | 43516.6 | 43889.91 | 38389.49 | 55812.1 | 94452.1 |
| 5.00E-05 | 10822.58 | 7928.839 | 7895.312 | 8069.362 | 10185.26 | 17929.13 | 46624.09 | 46996.8 | 42460.77 | 60497.78 | 107191.2 |
| 4.00E-05 | 13707.5 | 9506.08 | 9301.462 | 10232.92 | 11185.84 | 20232.03 | 51985.67 | 50822.04 | 46892.28 | 67810.1 | 123258.8 |
| 3.00E-05 | 14857.76 | 12761.77 | 10899.54 | 10913 | 14982.67 | 22280.96 | 60292.16 | 58752.41 | 51561.68 | 76704.28 | 138033.5 |
| 2.00E-05 | 15723.82 | 17539.03 | 13536.82 | 12054.83 | 20858.32 | 24163.66 | 70154.87 | 69790.08 | 58804.32 | 93506.7 | 172970.7 |
| 1.00E-05 | 25435.2 | 23741.71 | 24120.64 | 14911.19 | 32801.64 | 29417.9 | 89878.83 | 89167.63 | 74352.3 | 126825.9 | 237208.6 |
經擬合得到
| 11*11 | 0*1 | 36.916898197096046*δ**-0.5628401452873982 | |||||||||
| 0.9833906458591285?? ******? 決定系數 r**2 | |||||||||||
| 0*2 | 45.77503673337746*δ**-0.5220542727902084 | ||||||||||
| 0.9353104932989644?? ******? 決定系數 r**2 | |||||||||||
| 0*3 | 48.298055478642326*δ**-0.5057114058999935 | ||||||||||
| 0.9019987405496206?? ******? 決定系數 r**2 | |||||||||||
| 1*2 | 49.00995388262112*δ**-0.4961181916717476 | ||||||||||
| 0.9553825211292994?? ******? 決定系數 r**2 | |||||||||||
| 1*3 | 43.12892612498791*δ**-0.5454667033313928 | ||||||||||
| 0.9063025601235393?? ******? 決定系數 r**2 | |||||||||||
| 2*3 | 46.548201619862056*δ**-0.5558858453616066 | ||||||||||
| 0.8703614211691612?? ******? 決定系數 r**2 | |||||||||||
| a | 計算值/實驗值 | b | 計算值/實驗值 | ||||||||
| 11*11 | 0*1*2 | 151.73096129839246*δ**-0.567287416362122 | 130 | 0.860927 | 0.56 | 0.987654 | |||||
| 0.9717737321593617?? ******? 決定系數 r**2 | |||||||||||
| 0*1*3 | 153.92419785188702*δ**-0.5632487115074262 | 127 | 0.830065 | 0.56 | 0.994671 | ||||||
| 0.9725305629591284?? ******? 決定系數 r**2 | |||||||||||
| 0*2*3 | 194.14698809810508*δ**-0.5358986879050944 | 139 | 0.716495 | 0.55 | 1.028037 | ||||||
| 0.9572303851112073?? ******? 決定系數 r**2 | |||||||||||
| 1*2*3 | 185.57963663285054*δ**-0.5744580187691357 | 138 | 0.745946 | 0.55 | 0.958188 | ||||||
| 0.9822146188650052?? ******? 決定系數 r**2 | |||||||||||
| 0*1*2*3 | 343.3471676374182*δ**-0.5728684593396957 | 267 | 0.778426 | 0.56 | 0.979021 | ||||||
| 0.9942519845625751?? ******? 決定系數 r**2 | |||||||||||
計算組合參數a和b
計算表明a的計算值是實驗值的約78%，b的計算值約為實驗值的98%。
對比9*9的數據，a的計算值是實驗值的約90%，b的計算值是實驗值的約96%。
總結
以上是生活随笔為你收集整理的验算神经网络谐振子模型的第二组数据的全部內容，希望文章能夠幫你解決所遇到的問題。
- 上一篇： 一个训练集未知的神经网络
- 下一篇： 计算Be原子基态能级