machine-learning-experiments

Why we included this project

Instead of reading about neural networks, you can run them. Each experiment in this collection pairs a Jupyter/Colab notebook with a browser demo, so you see both how a model was trained and what it does live. The supervised examples cover MLPs, CNNs, and RNNs on MNIST, QuickDraw, and ImageNet, and there's an unsupervised GAN that generates fashion images. The author is clear that this is a learning sandbox, not tuned production code, so it's a safe place to experiment with TensorFlow/Keras without expecting a polished library. Students and self-taught developers who want to compare algorithms side by side will get the most out of it.