This liveProject puts you on the front lines at an indie bookstore, where poorly scrawled notes threaten to bring sales to a standstill. Learning classification algorithms from the ground up, you'll use four different and surprisingly simple classifiers to develop an AI solution that turns illegible scribbles into recognizable numbers.

In this liveProject, you'll implement particle swarm optimization and the Boids algorithm to develop computational models to track flocks of finches. You’ll start with a simple flock simulator, then modify it to adapt to obstacles, take a different approach with Gravity Boids, and finally use a swarm-inspired algorithm to find the highest point on a complex mathematical surface. You’ll soon have an algorithm that can predict the movements of everything from flocks of finches to swaying concertgoers!

In this liveProject, you'll take over administering a zoo automatically by crafting three different kinds of expert systems to help customers determine what animal they’re looking at. An expert system captures an expert's know-how by encoding it in a knowledge base. You’ll create a prototype 20-Questions-style game expert system, then expand it with logical rules and Boolean statements to build more powerful and realistic expert systems.

In this liveProject, you’ll build neural nets for a bank that are capable of classifying the numbers written on deposit slips. Customers often write out their important account details by hand, and you need to create a model capable of effectively reading their handwriting. You’ll use a feedforward network trained with backpropagation to identify hand-drawn digits with a high percentage of accuracy. Plus, this project contains an optional bonus milestone! Further your knowledge of neural networks by using the TensorFlow and Keras libraries to create an alternative and more flexible version of your first model. You’ll get hands-on experience with these tools' standard libraries, and with your new model you will be able to easily analyze the network's success rate and experiment with different activation functions.