Neural Network Sandwich Recommender

Photo by Diego Duarte Cereceda on Unsplash

What if there was a computer program that could make sandwich recommendations powered using a neural network? This is what I have made.

Unfortunately, it doesn’t actually work that well. The algorithm is based on Daniel Shiffman’s Color Predictor perceptron, except instead of using quantifiable RGB values, I’m using sandwich components. It works by presenting the user with a series of randomly generated sandwiches, to which the user can respond with a positive reaction or a negative reaction. This will train the neural net to know what kinds of sandwiches the user likes and will then present the user with new sandwiches based on that information. After 10 sandwich offerings, it will begin to make sandwich recommendations based on the previous training set. Take it for a spin!

The ingredients those for both savory and sweet sandwiches. If you answer honestly, eventually you will be recommended a sandwich that you might like! Source code can be found here.



Daniel Shiffman’s Color “Predictor”

Neurogram on

The Nature of Code, Chapter 10: Neural Networks

Leave a Reply

Your email address will not be published. Required fields are marked *