Awhile back, I decided to combine my interest in machine learning and baking to see if I could build something that would take the ingredient amounts from a recipe and predict what type of baked good it was. I won’t get into the technical details here (click the link above if you’re into that). Here I’ll give you a quick overview of how I made this recipe and then get right to the yummy parts.
Baking is an exact science
The reason I thought this might work is because machine learning is all about finding patterns in data, and baking recipes in particular are full of patterns. For example, bread recipes commonly have a 5:3 ratio of flour:water. There are similar ratios for cake, cookies, and more. Because of this, I was able to train a machine learning model to learn these ratios with minimal recipe data. But you might be wondering: what’s the point of entering ingredient amounts for a recipe and having a computer tell you what you already know (“it’s a cookie”)? Good question! Read on, my friend.
This little project got much more fun when I started playing around with the ingredient amounts and seeing what the model predicted. After a lot of time spent trying this out, I landed on a combination of ingredient amounts that caused the model to predict the recipe was 50% cake, 50% cookie. The obvious next step was to try to make this and see what happened. All the model gave me was ingredient amounts, so I had to come up with the recipe instructions on my own. Sure enough, this fun concoction was DELICIOUS and had the crispiness of a cookie with the crumbly-ness of a cake. I present to you, the cakie.
I think it has the crispy exterior of a cookie with a perfect cake-like crumb inside. If you make this I’d love to hear what you think. Tag me on Instagram – I’m @sararob!