Machine Learning

Machine Learning experiments and prototypes

NN-Racer

Completed

Neural network evolution using genetic algorithms for autonomous racing car behavior.

Date
2024-05
Duration
2 months
Team
solo
Difficulty
hard

Project Story

NN-Racer is an experiment in evolutionary computation where neural networks learn to control racing cars through genetic algorithms. The cars improve their performance over generations through mutation and selection.

NN-Racer simulation

Neural network cars evolving racing behavior

The project demonstrates how genetic algorithms can train neural networks without traditional backpropagation, evolving behavior through fitness-based selection and random mutations of network weights and biases.

Technical Details

Tech Stack

Python TensorFlow Genetic Algorithms Neural Networks Simulation

Key Features

Neural network evolution
Genetic algorithm optimization
Autonomous behavior learning
Performance tracking
Visualization of evolution
Multi-objective optimization

Challenges Faced

Computational intensity of evolution
Hyperparameter tuning complexity
Local optima problems
Balancing exploration vs. exploitation

Key Learnings

💡 Genetic algorithms can train complex behaviors
💡 Neural network architecture affects learning
💡 Fitness function design is crucial
💡 Evolution can find unexpected solutions
💡 Performance improvements over generations

Explore More Machine Learning Projects

Adam Siwek

Independent AI Builder & Creator. Building practical tools and educational content for developers navigating the AI transition.

Always building, always learning

Let's Connect

"Building in public, learning in real-time."

© 2025 Adam Siwek. Crafted with passion and AI assistance.

Privacy-first • Open source • Always shipping