The purpose of this tutorial is to provide a practical guide in building Neural Networks rather than an in-depth explanation of the ever-changing universe of the Deep Learning. This means that tutorial assumes that a reader is familiar with all the concepts and nomenclature of the Deep Learning. However, for those who did spend hundreds of hours reading blogs and books dedicated to the Deep Learning and those who desire more theoretical/mathematical explanations, the tutorial will try to give one or more links to the websites, that will help you start.
This tutorial will use Python as the main language with, as the title suggests, TensorFlow package. Therefore, it is also assumed that the reader is familiar with at least basic Python syntax. At the time of writing the latest Python version is 3.6.2 and for TensorFlow it is 1.3.0. Explanation on how to set up Python environment and install all the necessary packages will be provided in the next chapter, so sit tight.
All code presented in this tutorial is available in the form of scripts. Each script is self-contained and has comments that supposed to guide the reader through the code (if it doesn't, please, let me know). Scripts were written with an average user in mind thus the scripts omit many advanced features that might speed-up the code but at the same time make code harder to understand. This code also can be used as a template for whatever reader might want to do with it, so feel free to play around.
This tutorial consists of the following chapters:
- Python environment set-up is a guide to a user on how to install all necessary tools in order to run the provided scripts (feel free to skip this if you know about conda and yml files).
- Introduction to TensorFlow provides a description of basic building blocks and concepts that are used in the model building in TensorFlow.
- Logistic Regression shows how to uses "real" data set in order to determine the type of the breast cancer.
- Linear Regression provides a simple example on how to fit a straight line to a data.
- Nonlinear Regression gives description on how to fit a nonlinear equation to a data
- Introduction to Recurrent Neural Network provides a very brief explanation why we might need to use the Recurrent Neural Networks and what options are available in TensorFlow out of the box. Also, we touch on issues of overfitting and underifting.
- Recurrent Neural Network and Sequences chapter shows how to deal with the situation when input sequences are variable length and also how to predict sequences.
- Concluding Remarks chapter contains a few random thoughts that may or may not be relevant to the tutorial.
So, if you still feel positive and want to continue, let us begin.