Created
July 19, 2018 17:53
-
-
Save parcmepperman/51a953bc5555a089dea38eb0b1d4178c to your computer and use it in GitHub Desktop.
Results DL Lab #6
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
First run of CNNmodel.py (nothing changed) | |
step 0, training accuracy 0.18 | |
step 100, training accuracy 0.88 | |
step 200, training accuracy 0.94 | |
step 300, training accuracy 0.96 | |
step 400, training accuracy 0.92 | |
test accuracy 0.9433 | |
Time for building convnet: | |
103588 | |
First run of CNmodel.py (dataset changed to fashion-mnist-master/data/fashion)] | |
step 0, training accuracy 0.08 | |
step 100, training accuracy 0.62 | |
step 200, training accuracy 0.7 | |
step 300, training accuracy 0.86 | |
step 400, training accuracy 0.84 | |
test accuracy 0.8282 | |
Time for building convnet: | |
94340 | |
Fifth run of CNmodel.py (AdaGrad Optimizer, no change)] | |
step 0, training accuracy 0.14 | |
step 100, training accuracy 0.72 | |
step 200, training accuracy 0.9 | |
step 300, training accuracy 0.8 | |
step 400, training accuracy 0.82 | |
test accuracy 0.8171 | |
Time for building convnet: | |
111625 | |
Seventeenth run of CNmodel.py (GradientDescent - 16,36,128 in/out channels) | |
step 0, training accuracy 0.22 | |
step 100, training accuracy 0.78 | |
step 200, training accuracy 0.8 | |
step 300, training accuracy 0.78 | |
step 400, training accuracy 0.8 | |
test accuracy 0.8268 | |
Time for building convnet: | |
46309 | |
Process finished with exit code 0 | |
Eighteenth run of CNmodel.py (GradientDescent - 8,45,128 in/out channels) | |
step 0, training accuracy 0.26 | |
step 100, training accuracy 0.74 | |
step 200, training accuracy 0.92 | |
step 300, training accuracy 0.8 | |
step 400, training accuracy 0.88 | |
test accuracy 0.8124 | |
Time for building convnet: | |
34380 | |
Nineteenth run of CNmodel.py (GradientDescent - 8,45,200 in/out channels) | |
step 0, training accuracy 0.14 | |
step 100, training accuracy 0.7 | |
step 200, training accuracy 0.88 | |
step 300, training accuracy 0.8 | |
step 400, training accuracy 0.9 | |
test accuracy 0.8379 | |
Time for building convnet: | |
35690 | |
Run # 20 (Final run) (GradientDescent -- 8,45,500 -- steps = 900 (from 500)) | |
step 0, training accuracy 0.34 | |
step 100, training accuracy 0.9 | |
step 200, training accuracy 0.88 | |
step 300, training accuracy 0.86 | |
step 400, training accuracy 0.9 | |
step 500, training accuracy 0.86 | |
step 600, training accuracy 0.86 | |
step 700, training accuracy 0.76 | |
step 800, training accuracy 0.9 | |
test accuracy 0.8584 | |
Time for building convnet: | |
64197 | |
From my observations the GitHub ReadMe for A MNIST-like fashion product database was completely accurate in that it is easy | |
to get a 93% accuracy from the minst dataset, once the data was changed it hovered around approx 82%, as you can see it wasn't | |
until I changed the step count to almost 1000, the in/out channels to 8,45,500 and applied the gradient descent optimizer | |
that I was able to achieve an 86% accuracy. I strongly believe that if I kept recording data and the separation of channels | |
and increased the amount of steps significantly that I might achieve a 90%, but 94% as in the MINST dataset is not achievable | |
at my level of knowledge... |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment