FBNN

A new type of neural network that takes advantage of running in a simulated environment by using DSP tools.

Comment

Inspiration

If we want to be able to model the complexities of creatures, we need a better way of modeling how they behave. Before we toss hundreds of tires off the coast of Ft. Lauderdale to see if animals will turn them into homes, we can try and model the environment first.

What it does

This algorithm simulates agents with neural network brains in real time. Because of this environment however, we are able to utilize oscillators and filters in the network to give the networks emotion.

How we built it

This project was done entirely in one big python file from scratch, utilizing Numpy for math operations and OpenCV for visualizing.

Challenges we ran into

Immediately the scope was the biggest challenge, and many visualization tools we wanted to implement had to be left out.

Accomplishments that we're proud of

Developing the neural network from scratch was just a ton of fun and super fulfilling

What's next for FBNN

This algorithm feels promising, and we are going to try and extend the features to incorporate more filter types and visualization tools.

Built With

Share this project:

Updates