This is a blog that contains jupyter notebooks that I have been practicing when I read the book Probabilistic Deep Learning With Python, Keras and TensorFlow Probability by Oliver Dürr, Beate Sick, and Elvis Murina. The table of content to the book’s jupyter notebook files can be found from here.

Index	Description	Link to Jupiter notebook	Content
1	In a CNN the order of the data does matter. The CNN performs better on the original data than the shuffled data (the shuffling here is the shuffling for pixels in images This was not the case when we used fully connected Neural Networks fcNN - here the performance did not change if the data was shuffled.	MNIST data using fcNN and CNN with pixels shuffling	Data visualization, shuffle image pixels, fully connected NN, fcNN summary, draw NN history function, evaluate categorical prediction function, CNN, …