In any real job working in an AI team, one of the primary goals will be to build regression models that can make predictions in non-linear datasets. Because of the complexity of the real world and the data that you'll be working with, simple linear regression models won't provide the predictive power you're seeking. That is why, in this chapter, we will discuss how to build world-class prediction models using MLP. More information can be found at http://www.deeplearningbook.org/contents/mlp.html, and an example of the MLP architecture is shown here:

We will implement a neural network with a simple architecture of only two layers, using TensorFlow, that will perform regression on the MNIST dataset (http://yann.lecun.com/exdb/mnist/) that we will provide. We...

Before we jump into building our awesome neural network, let's first have a look at the famous MNIST dataset. So let's visualize the MNIST dataset in this section.

Let's start therefore by downloading the data, using the following commands:

from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("/tmp/data/", one_hot=True)

If we examine the mnist variable content, we can see that it is structured in a specific format, with three major components—TRAIN, TEST...

Let's build our intuition around this project. What we need to do is build a deep learning technology that accurately assigns class labels to an input image. We're using a deep neural network, known as an MLP, to do this. The core of this technology is the mathematics of regression. The specific calculus proofs are outside the scope of this book, but in this section, we provide a foundational basis for your understanding. We also outline the structure of the project, so that it's easy to understand the primary steps needed to create our desired results.

Our first task is to define the model that will perform regression on the provided MNIST dataset. So, we will...

To code the implementation, we'll start by defining the hyperparameters, then we will define the model, followed by building and executing the training loop. We conclude by checking to see if our model is overfitting and build an inference code that loads the latest checkpoints and then makes predictions on the basis of learned parameters.

We will define all of the required hyperparameters in the hy_param.py file and then import it as a module in our other codes. This makes it easy in deployment, and is good practice to make your code as modular as possible. Let's look into the hyperparameter configurations that we have in our hy_param.py file:

#!/usr/bin...

Today's project was to build a classifier to solve the problem of identifying specific handwriting samples from a dataset of images. Our hypothetical use case was to apply deep learning to enable customers of a restaurant chain to write their phone numbers in a simple iPad application, so that they could get a text notification that their party was ready to be seated. Our specific task was to build the intelligence that would drive this application.

Our MLP model accuracy hit 87.42%! Not bad, given the depth of the model and the hyperparameters that we chose at the beginning. See if...

In the project in this chapter, we successfully built an MLP to produce a regression classification prediction, based on handwritten digits. We gained experience with the MNIST dataset and a deep neural network model architecture, which gave us the added opportunity to define some key hyperparameters. Finally, we looked at the model performance in testing and determined whether we succeeded in achieving our goals.