Want to get a real feel for machine learning in practice?
Here’s a simple but powerful exercise: classify thousands of digits (0 vs 1) from the MNIST dataset (12,666 train samples and 2,116 test samples) using logistic regression — trained with gradient descent. Just 5 epochs.
What you’ll get out of it:
- see how a model actually works with image data
- understand where linear models start to break down
- try a clean implementation in pure PHP: accuracy: 99.91%
- compare it with a more production-ready approach using RubixML: accuracy: 99.95%
If you’ve been meaning to move from theory to something hands-on, this is a great place to start:
https://aiwithphp.org/books/ai-for-php-developers/examples/part-3/logistic-regression/case-0/mnist-0-1
Top comments (0)