What does the training accuracy plot of my convolution neural network (CNN) show?

1 次查看(过去 30 天)
Hello everybody
the result of my CNN is shown in the picture attached. I'm wondering about the accuracy why it goes down and up during the training? is it normal or it should grow gradually? and what is the possible error that may I have on my net (or parameters)!! Additionally, whatever I change the training options; the test accuracy does not exceed 42% !!!
if true
Training on single CPU.
|=========================================================================================|
| Epoch | Iteration | Time Elapsed | Mini-batch | Mini-batch | Base Learning|
| | | (seconds) | Loss | Accuracy | Rate |
|=========================================================================================|
| 1 | 1 | 6.67 | 1.6277 | 0.00% | 1.00e-04 |
| 1 | 20 | 128.30 | 1.6677 | 25.00% | 1.00e-04 |
| 1 | 40 | 253.77 | 1.6505 | 50.00% | 1.00e-04 |
| 1 | 60 | 381.41 | 0.9331 | 87.50% | 1.00e-04 |
| 1 | 80 | 505.31 | 1.0754 | 25.00% | 1.00e-04 |
| 2 | 100 | 629.83 | 1.6579 | 12.50% | 1.00e-04 |
| 2 | 120 | 758.84 | 1.3724 | 62.50% | 1.00e-04 |
| 2 | 140 | 884.09 | 1.1539 | 50.00% | 1.00e-04 |
| 2 | 160 | 1028.53 | 1.1311 | 37.50% | 1.00e-04 |
| 2 | 180 | 1154.14 | 1.4353 | 37.50% | 1.00e-04 |
| 3 | 200 | 1277.55 | 0.9360 | 50.00% | 1.00e-04 |
| 3 | 220 | 1401.44 | 0.9559 | 50.00% | 1.00e-04 |
| 3 | 240 | 1525.49 | 1.6097 | 25.00% | 1.00e-04 |
| 3 | 260 | 1649.96 | 0.9116 | 62.50% | 1.00e-04 |
| 3 | 280 | 1774.19 | 1.0897 | 37.50% | 1.00e-04 |
| 4 | 300 | 1898.34 | 1.4818 | 12.50% | 1.00e-04 |
| 4 | 320 | 2022.42 | 1.1853 | 50.00% | 1.00e-04 |
| 4 | 340 | 2146.87 | 0.9665 | 62.50% | 1.00e-04 |
| 4 | 360 | 2272.24 | 1.1143 | 37.50% | 1.00e-04 |
| 4 | 380 | 2396.43 | 1.1264 | 37.50% | 1.00e-04 |
| 5 | 400 | 2522.21 | 1.5471 | 50.00% | 1.00e-04 |
| 5 | 420 | 2646.45 | 1.3815 | 50.00% | 1.00e-04 |
| 5 | 440 | 2776.98 | 0.7213 | 87.50% | 1.00e-04 |
| 5 | 460 | 2906.50 | 0.8455 | 87.50% | 1.00e-04 |
| 6 | 480 | 3033.40 | 1.7557 | 12.50% | 1.00e-04 |
| 6 | 500 | 3159.12 | 1.1510 | 50.00% | 1.00e-04 |
| 6 | 520 | 3290.33 | 1.0716 | 62.50% | 1.00e-04 |
| 6 | 540 | 3419.24 | 1.2187 | 37.50% | 1.00e-04 |
| 6 | 560 | 3545.82 | 1.3443 | 37.50% | 1.00e-04 |
| 7 | 580 | 3671.92 | 0.9136 | 50.00% | 1.00e-04 |
| 7 | 600 | 3796.45 | 0.8985 | 62.50% | 1.00e-04 |
| 7 | 620 | 3920.45 | 1.4416 | 37.50% | 1.00e-04 |
| 7 | 640 | 4051.54 | 0.9950 | 75.00% | 1.00e-04 |
| 7 | 660 | 4191.68 | 0.8132 | 75.00% | 1.00e-04 |
| 8 | 680 | 4328.36 | 1.3569 | 25.00% | 1.00e-04 |
| 8 | 700 | 4463.55 | 1.1009 | 50.00% | 1.00e-04 |
| 8 | 720 | 4593.56 | 1.0073 | 62.50% | 1.00e-04 |
| 8 | 740 | 4718.89 | 1.0589 | 50.00% | 1.00e-04 |
| 8 | 760 | 4843.50 | 0.9829 | 50.00% | 1.00e-04 |
| 9 | 780 | 4965.23 | 1.2858 | 62.50% | 1.00e-04 |
| 9 | 800 | 5086.95 | 1.4522 | 50.00% | 1.00e-04 |
| 9 | 820 | 5207.89 | 0.4955 | 100.00% | 1.00e-04 |
| 9 | 840 | 5328.95 | 0.7283 | 100.00% | 1.00e-04 |
| 10 | 860 | 5450.18 | 1.6487 | 37.50% | 1.00e-04 |
| 10 | 880 | 5570.79 | 0.8402 | 75.00% | 1.00e-04 |
| 10 | 900 | 5692.05 | 0.8969 | 62.50% | 1.00e-04 |
| 10 | 920 | 5812.29 | 1.1199 | 37.50% | 1.00e-04 |
| 10 | 940 | 5932.70 | 1.0859 | 50.00% | 1.00e-04 |
| 11 | 960 | 6053.34 | 0.7106 | 62.50% | 1.00e-04 |
| 11 | 980 | 6173.80 | 0.8470 | 50.00% | 1.00e-04 |
| 11 | 1000 | 6295.36 | 1.3543 | 25.00% | 1.00e-04 |
| 11 | 1020 | 6415.40 | 1.0594 | 50.00% | 1.00e-04 |
| 11 | 1040 | 6537.31 | 0.4968 | 75.00% | 1.00e-04 |
| 12 | 1060 | 6659.25 | 1.0452 | 50.00% | 1.00e-04 |
| 12 | 1080 | 6780.46 | 0.8746 | 62.50% | 1.00e-04 |
| 12 | 1100 | 6900.97 | 1.1169 | 50.00% | 1.00e-04 |
| 12 | 1120 | 7022.03 | 0.9600 | 50.00% | 1.00e-04 |
| 12 | 1140 | 7144.63 | 0.8063 | 50.00% | 1.00e-04 |
| 13 | 1160 | 7266.01 | 1.0481 | 75.00% | 1.00e-04 |
| 13 | 1180 | 7385.75 | 1.3504 | 50.00% | 1.00e-04 |
| 13 | 1200 | 7505.62 | 0.3157 | 100.00% | 1.00e-04 |
| 13 | 1220 | 7627.16 | 0.6529 | 87.50% | 1.00e-04 |
| 14 | 1240 | 7749.26 | 1.1844 | 62.50% | 1.00e-04 |
| 14 | 1260 | 7874.78 | 0.6447 | 75.00% | 1.00e-04 |
| 14 | 1280 | 7994.68 | 0.7824 | 62.50% | 1.00e-04 |
| 14 | 1300 | 8114.98 | 0.9300 | 62.50% | 1.00e-04 |
| 14 | 1320 | 8237.20 | 0.8984 | 62.50% | 1.00e-04 |
| 15 | 1340 | 8359.44 | 0.4070 | 75.00% | 1.00e-04 |
| 15 | 1360 | 8481.32 | 1.0424 | 62.50% | 1.00e-04 |
| 15 | 1380 | 8601.45 | 0.8956 | 50.00% | 1.00e-04 |
| 15 | 1400 | 8722.41 | 0.9647 | 62.50% | 1.00e-04 |
| 15 | 1420 | 8844.57 | 0.2415 | 100.00% | 1.00e-04 |
| 15 | 1425 | 8874.94 | 0.6794 | 62.50% | 1.00e-04 |
|=========================================================================================|
accuracy =
0.3787
end

回答(0 个)

类别

Help CenterFile Exchange 中查找有关 Image Data Workflows 的更多信息

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by