> Chapter 3:
In Chapter 3, we caught up with years of research in machine learning, introducing convolutional neural networks (CNNs). In the following notebooks, we illustrate the operations forming the building blocks of CNNs, test a historical neural network, and experiment with modern tools to train more robust models.
(Reminder: Notebooks are better visualized with nbviewer: click here to continue on nbviewer.jupyter.org.)
- 3.1 - Discovering CNNs' Basic Operations
- Learn to use convolutional, average-pooling, and max-pooling layers with TensorFlow 2.
- 3.2 - Building and Training our First CNN with TensorFlow 2 and Keras
- Implement a LeNet-5 model and test it on hand-written digit recognition.
- 3.3 - Experimenting with Advanced Optimizers
- Use and compare some of the optimizers offered by TensorFlow 2, to improve the training of new models.
- 3.4 - Applying Regularization Methods to CNNs
- Avoid over-fitting and train more robust neural networks by applying regularizers, dropout, batch normalization, etc.