Hands-on programming tutorial for TensorFlow and Keras
Presentation chart here: http://bit.ly/2Flgnnp
TensorFlow programming exercises:
-
TF-00-ReadData.py
Start your python basic program and load the data. -
TF-01-Placeholders.py
Define Placeholder for input: image and label. -
TF-02-VariableFunctions.py
Define Variables for model. -
TF-03-ModelLossOptimizer.py
Define Loss function and Optimizer function. -
TF-04-DefineAccuracy.py
Add Accuracy calculation. -
TF-05-RunGraphWithError.py
Connect to runtime and run graph. -
TF-06-WorkingBasic.py
Fix error: initialize variables. -
TF-07-IncreasedBatch.py
Try larger batch of images. -
TF-08-OneCNN.py
Create the first Convolution layer in the neural net. Add name scope to make the layers easier to read. -
TF-09-TwoCNN.py
Create the second Convolution layer in the neural net. -
TF-10-FullConnect.py
Create the fully connected layer in the neural net. -
TF-11-DropOut.py
Add the dropout layer in the neural net to control overfitting. -
TF-12-TensorBoard.py
Add FileWriter to visualize with TensorBoard. Add names and name scope to make it easier to read the graph. Add viewing images in TensorBoard. Add line graphs and histograms of variables. Add visualization for embedding.
Keras programming exercises:
-
Keras-00-ReadData.py
Load MNIST data from archive. -
Keras-01-DenseInput.py
From #1, flatten image into a vector to prepare for the Dense layer. -
Keras-02-DenseLayer.py
From #2, build a dense layer, compile and train. -
Keras-03-CNNInput.py
From #1, reshape image to prepare for the convolution layer. -
Keras-04-OneCNNLayer.py
From #4, build one CNN layer. -
Keras-05-TwoCNNLayer.py
From #5, add second CNN layer and a dropout layer.