1vn/mnist-tf-tutorial

mnist-tf-tutorial's Issues

Todos

Good job man! I'm impressed with how far you got. Good code design.

Todo:

• remove data reading from your trainer
• add a text file with the performance comparisons between different depths, dropout vs. no dropout, etc.
• should use a loop and a 'depth' hyperparam instead of manually defining the depth
• consider using tf.nn.dense instead of the manual stuff (but I'm glad you did this since you'll actually know what a dense layer is doing.)
• some comments to clean up
• LSTM isn't done, talk to Navie about his implementation and team up to get something that trains