-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathReadme2
97 lines (69 loc) · 4.43 KB
/
Readme2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
Train on 60000 samples, validate on 10000 samples
Epoch 1/20
Epoch 00001: LearningRateScheduler setting learning rate to 0.003.
60000/60000 [==============================] - 9s 155us/step - loss: 0.3423 - acc: 0.9357 - val_loss: 0.0864 - val_acc: 0.9830
Epoch 2/20
Epoch 00002: LearningRateScheduler setting learning rate to 0.0022744503.
60000/60000 [==============================] - 6s 94us/step - loss: 0.1037 - acc: 0.9788 - val_loss: 0.0508 - val_acc: 0.9883
Epoch 3/20
Epoch 00003: LearningRateScheduler setting learning rate to 0.0018315018.
60000/60000 [==============================] - 5s 91us/step - loss: 0.0771 - acc: 0.9820 - val_loss: 0.0489 - val_acc: 0.9885
Epoch 4/20
Epoch 00004: LearningRateScheduler setting learning rate to 0.0015329586.
60000/60000 [==============================] - 6s 93us/step - loss: 0.0604 - acc: 0.9855 - val_loss: 0.0438 - val_acc: 0.9902
Epoch 5/20
Epoch 00005: LearningRateScheduler setting learning rate to 0.0013181019.
60000/60000 [==============================] - 6s 94us/step - loss: 0.0532 - acc: 0.9871 - val_loss: 0.0290 - val_acc: 0.9927
Epoch 6/20
Epoch 00006: LearningRateScheduler setting learning rate to 0.0011560694.
60000/60000 [==============================] - 6s 92us/step - loss: 0.0466 - acc: 0.9882 - val_loss: 0.0303 - val_acc: 0.9924
Epoch 7/20
Epoch 00007: LearningRateScheduler setting learning rate to 0.0010295127.
60000/60000 [==============================] - 6s 92us/step - loss: 0.0423 - acc: 0.9889 - val_loss: 0.0327 - val_acc: 0.9907
Epoch 8/20
Epoch 00008: LearningRateScheduler setting learning rate to 0.0009279307.
60000/60000 [==============================] - 6s 93us/step - loss: 0.0392 - acc: 0.9900 - val_loss: 0.0272 - val_acc: 0.9928
Epoch 9/20
Epoch 00009: LearningRateScheduler setting learning rate to 0.0008445946.
60000/60000 [==============================] - 6s 92us/step - loss: 0.0347 - acc: 0.9913 - val_loss: 0.0272 - val_acc: 0.9939
Epoch 10/20
Epoch 00010: LearningRateScheduler setting learning rate to 0.0007749935.
60000/60000 [==============================] - 6s 93us/step - loss: 0.0339 - acc: 0.9915 - val_loss: 0.0233 - val_acc: 0.9942
Epoch 11/20
Epoch 00011: LearningRateScheduler setting learning rate to 0.0007159905.
60000/60000 [==============================] - 5s 92us/step - loss: 0.0317 - acc: 0.9922 - val_loss: 0.0339 - val_acc: 0.9904
Epoch 12/20
Epoch 00012: LearningRateScheduler setting learning rate to 0.000665336.
60000/60000 [==============================] - 6s 93us/step - loss: 0.0284 - acc: 0.9929 - val_loss: 0.0222 - val_acc: 0.9939
Epoch 13/20
Epoch 00013: LearningRateScheduler setting learning rate to 0.0006213753.
60000/60000 [==============================] - 6s 92us/step - loss: 0.0274 - acc: 0.9937 - val_loss: 0.0223 - val_acc: 0.9946
Epoch 14/20
Epoch 00014: LearningRateScheduler setting learning rate to 0.0005828638.
60000/60000 [==============================] - 6s 92us/step - loss: 0.0256 - acc: 0.9938 - val_loss: 0.0230 - val_acc: 0.9936
Epoch 15/20
Epoch 00015: LearningRateScheduler setting learning rate to 0.0005488474.
60000/60000 [==============================] - 5s 92us/step - loss: 0.0258 - acc: 0.9936 - val_loss: 0.0222 - val_acc: 0.9940
Epoch 16/20
Epoch 00016: LearningRateScheduler setting learning rate to 0.0005185825.
60000/60000 [==============================] - 6s 93us/step - loss: 0.0239 - acc: 0.9948 - val_loss: 0.0212 - val_acc: 0.9940
Epoch 17/20
Epoch 00017: LearningRateScheduler setting learning rate to 0.000491481.
60000/60000 [==============================] - 5s 91us/step - loss: 0.0225 - acc: 0.9952 - val_loss: 0.0206 - val_acc: 0.9948
Epoch 18/20
Epoch 00018: LearningRateScheduler setting learning rate to 0.0004670715.
60000/60000 [==============================] - 6s 94us/step - loss: 0.0202 - acc: 0.9954 - val_loss: 0.0203 - val_acc: 0.9948
Epoch 19/20
Epoch 00019: LearningRateScheduler setting learning rate to 0.0004449718.
60000/60000 [==============================] - 6s 93us/step - loss: 0.0194 - acc: 0.9957 - val_loss: 0.0218 - val_acc: 0.9944
Epoch 20/20
Epoch 00020: LearningRateScheduler setting learning rate to 0.000424869.
60000/60000 [==============================] - 5s 92us/step - loss: 0.0194 - acc: 0.9956 - val_loss: 0.0205 - val_acc: 0.9949
<keras.callbacks.History at 0x7f655c617630>
Model evaluate
[0.0205034158016555, 0.9949]
Strategy Chosen
a) Reduce the 32 channels to 16
b) Explicitly not used biases
c) Reduce drop out ratio
These were sufficient to get me the results