Deep Learning - Exercise 07

Learning goal: Getting started with TensorFlow

TensorFlow is an open-source software library for machine learning written in C++ and Python with a special focus on Deep Learning.

1. Install TensorFlow on your system

TensorFlow is currently supported only for Linux and Mac OS X. There is no direct support for Windows (not yet...).

If you want to install and run it, you can choose at least among the following approaches:

  • Install TensorFlow under Linux or Mac OS X using the Python package manager pip. Here is also a documentation by my own how to install TensorFlow under Ubuntu.
  • Use VMWarePlayer to install a Linux (Ubuntu) system in a virtual computer and then install TensorFlow using the Pyhton package manager pip. Here is also a documentation by my own how to install TensorFlow under Ubuntu.
  • TensorFlow is not directly supported for Windows. However, if you want to run it under Windows, you can use the new Bash shell that is integrated in Windows 10. Please note, that you will not be able to use visualizations.

    Read my documentation here wich provides a step-by-step documentation how to install and run TensorFlow under Windows.

2. Test your TensorFlow installation

After you have installed TensorFlow you should test if you really can run TensorFlow.

For this write a small Python program using TensorFlow that does a very basic machine learning task: linear regression.

3. Implement a MLP using TensorFlow

Implement a Multi-Layer Perceptron using TensorFlow and test how good it can classify the MNIST dataset.