Computational Creativity

This project explores the intersection of computational creativity and evolutionary algorithms to generate novel cookie recipes. Developed as part of my Computational Creativity course with the help of my team mates-H.P. Pranaav and Trent Eriksen, Bitwise Bakeoff leverages a unique dataset of cookie recipes—organized from A to Z and numerically labeled—to fuel a genetic algorithm that produces imaginative, offbeat dessert ideas. The project not only highlights technical proficiency in data processing and algorithm design but also emphasizes a creative approach to problem-solving.

Technical Approach & Dataset Creation
To kick off the project, I implemented a one-time Python script to extract a comprehensive dataset from a curated GitHub repository. This script parsed recipe markdown files, retrieved titles and ingredients, and normalized the data into a structured JSON format. Although the script was intentionally non-optimized (taking around 30 minutes to run), it efficiently built the foundation needed to explore a genetic algorithm approach by including a broad range of cookie recipes with diverse ingredient lists.

Genetic Algorithm & Creativity Evaluation
The core of Bitwise Bakeoff is a genetic algorithm that treats each recipe as a binary-encoded string where every ingredient is assigned a bit. The algorithm evaluates recipes based on the length of their ingredient lists—favoring complexity and diversity—and employs custom crossover and mutation operations to create new recipe combinations. For example, crossover operations randomly merge ingredients from two recipes, while mutations may delete, alter, or even double the quantities of selected ingredients, mimicking real-world improvisations in the kitchen. This system not only generates quirky and unexpected recipes but also provides a platform for evaluating creativity through metrics such as novelty, coherence, and feasibility.