Difficulty: intermediate
Estimated Time: 30 minutes

TensorFlow is the platform enabling building complex deep Neural Network architectures.

In this scenario you will learn how to use TensorFlow when building the network layer by layer. You will start with using simple dense type and then move to using more complex techniques like convolutional networks and max pooling and dropout.

You've completed TensorFlow Layers scenario!

Don’t stop now! The next scenario will only take about 10 minutes to complete.

Layers in Tensorflow

Step 1 of 6

MNIST Dataset

Artificial Neural Networks(ANN) are the state of the art tool when building Deep Learning solutions. They have proved to be effective in many tasks in the Computer Vision or Natural Language Processing fields.

The depth itself comes from forming layers of neurones. Deep architectures allow splitting the complex task into several phases, but can also introduce time-consuming computations, usually in the training stage.

TensorFlow offers few ways of building the ANN architectures. This tutorial will walk you though adding the layers process to the classification network. We will be working on the MNIST dataset. The task is to recognise the actual digit from its handwritten representation.

MNIST Classification

If you look at the help.py code, you can see the read_mnist function that uses built-in capabilities of TensorFlow to download the dataset locally and load it into the python variable. As a result (if not specified otherwise), the data will be downloaded into the MNIST_data/ folder.