Neural networks (NN) are fast becoming huge tool in the data science world. They can be used to analyze numerous types of data. This class will serve as an introductory lesson with the goal to teach the basics of neural networks, and how they are structured. We will use Google Colab for a live code along experience and build two NN during the class for text classification and image classification. Finally, we will introduce the SHAP tool which aids in visualizing what is going inside the NN under the hood.
At their core neural networks are just like any other machine learning algorithm, they are built to understand patterns. Neural networks have been around since the 1940s but have gained popularity as of late due to the increased availability of cheap computing power. This powerful class of machine learning algorithms can be used in multiple aspects of business. For example, LinkedIn and Facebook use a neural networks to detect spam and abusive content on their feeds. Neural networks have been developed to decrease bank fraud, forecast product production, and help with cancer detection.
In this class, we will focus building a basic understanding of how neural networks function, and what are the different components of a neural network. Furthermore, we will discuss how to train, fit, and visualize networks using the Keras/Tensorflow framework and SHAP (SHapley Additive exPlanations). We will then use this knowledge to live code two neural networks using the MNIST Fashion database (image recognition) and stack overflow data (text classification). Finally, we will discuss what is going on under the hood of the neural networks based on the SHAP visualization tool.