December 2018
Intermediate to advanced
764 pages
18h 18m
English
In the previous sections, we learned how to build a simple MLP architecture using TensorFLow and its high-level libraries. We got an accuracy of about 0.93-0.94 with pure TensorFlow, 0.96-0.98 with Keras, and 0.96-0.97 with TFLearn. Even though all the examples of our code use TensorFlow underneath, the difference in accuracy for the same architecture and parameters can be attributed to the fact that although we initialized some important hyper-parameters, the high-level libraries and TensorFlow abstract away many other hyper-parameters that we did not modify from their default values.
We observe that the code in TensorFlow is very detailed and lengthy as compared to Keras and TFLearn. The ...
Read now
Unlock full access