eriklindernoren/ML-From-Scratch ? reverse-engineered prompt

Build me a beginner friendly Python project that teaches machine learning by implementing the main algorithms from scratch, with clear NumPy code instead of black box libraries.

I want it to feel like a learning library, not a production tool. Include simple examples people can run to see polynomial regression, digit classification with a small neural network, clustering, handwritten digit generation, reinforcement learning for a basic game environment, image reconstruction, genetic algorithms, and association analysis. The code should be organized so someone can open each model and understand the math and training steps without digging through a huge framework.

Please add setup instructions, a requirements file, and runnable example scripts that print useful training progress and show plots or generated images where it makes sense. Keep the implementations accessible, commented, and focused on showing how the algorithms work rather than being the fastest possible version.