Exercises attempted by Toan Luong 2/4/2017 and followed Martin Görner's tutorials https://cloud.google.com/blog/big-data/2017/01/learn-tensorflow-and-deep-learning-without-a-phd
- 5-layer neural network using softmax, LR decay, Adam optimizer
- Similar to (1) but added drop-out with pkeep=0.75
- Convolutional Neural Network. 5 layers according to this structure
