Teach Your Computer to Code FizzBuzz
summary
Workshop demonstrates evolving working FizzBuzz implementations using the JCLEC Java framework and experiments with DEAP-style tree GP in Python, seeding primitives like mod 3/5/15 and string terminals. Fitness comes from eight property tests; runs typically go thousands of generations (example config: 3000 generations). Results compile and pass the tests and are recorded in a Hall of Fame, but produced code is bloated, hard to read, and required manual parameter tuning. The same approach also improves existing code and hardware designs.
AI-written summary · about
chapters (9)
- 0:00Workshop introduction
- 3:28Optimization challenges
- 9:43Evolutionary computing basics
- 11:51Evolutionary algorithms
- 16:05Applying genetic algorithms
- 21:28Framework setup and tools
- 29:03Solving fizzbuzz
- 36:51Tree-based code generation
- 46:13Future of human-AI coding
related talks
09mOuvroir de Code Potentiel
59mOne kata, three languages
08mTen Ways To Fizz Buzz
45mKeynote: Writing Code? Pfft... Evolve it Instead!
33mLinear Genetic Programming in Python Bytecode
40mGenetic Programming in Clojure
37mType Hints: Putting more Buzz in your Fizz
30mConcept to Classroom: Creating fun differentiated programming projects