These are the materials I have prepared as a TA to teach to students who have enrolled in the Deep Learning course. This tutorial is divided into three sessions.
-
In the first session, you will learn about Tensorflow basics. You will learn about Computational Graphs, Sessions, Visualizing the graphs in TensorFlow, Variable Scopes, Name Scopes, and some of the apparent mistakes that beginners usually make. Linear Regression problem is also implemented in three different ways to illustrate what is the main idea of libraries like TensorFlow.
-
In the second session, you will learn how to create a Tensorflow model in an objective manner. You will also implement a simple Data Loader without the concerns of RAM management. You will see how the train set and validation set can be appropriately used to train the model and how to monitor the training process in the Tensorboard. You will also learn about different optimizers and learning rate decay mechanisms in TensorFlow.
-
In the third session, you will learn about saving and restoring the model. You will learn how to use TensorFlow's Dataset API to create a more effective Data Loader and how to parallelize I/O and computations. You will learn more visualizations that can be done using Tensorboard. You will also learn how Tensorflow's Dataset API can be used for data augmentation.