Implement Multilayer Perceptron (MLP) with Keras using Fashion MNIST

We walk, see, speak, and create. All of this activity originates in the brain. The study of it has resulted in our understanding of the interaction of electrical signals that result in ideas, thoughts, and decisions.

The atomic entity behind this activity is a single cell called the neuron (the orange, green, and blue circles in the figure above). A neuron contains a long extension called an axon, which can connect to other neurons enabling simple communication from one to another that it is connected with.

This all seems very simple but when millions of neurons are grouped in a single cluster, receiving signals from another such cluster, then fire off signals to yet another one, we see that complex processes can happen: walking, talking, thinking, and creating.

During the 1940s, this concept resulted in the notion of the Artificial Neural Network (ANN). Early on, the idea began to form that we could create machines that can think. However it was not practical to implement one to do any more of the simplest tasks. In the late 1990s, with the advent of more capable system and graphic processors the idea was revisited resulting in the study of Deep Learning and the notion of Machine Learning models trained and used in areas such as image and handwriting recognition, medical research, self-driving cars, or solving many probems requiring predictive analysis.

In this notebook we build the following:

• Multilayer Perceptron - an example of an Artificial Neural Network, as described in this article

• Implemented using Keras: a python library that offers MLP and other ANN examples as data models available for training and making predictions. Using Keras, we'll implement an ANN process images of clothing and predicting what a given image is, with high accuracy.

Get needed libraries and tools