DeepLearningProject

Why we included this project

This tutorial walks through an entire machine learning project the way you would actually run one, rather than a 30-minute demo on a clean dataset. You build your own dataset instead of using MNIST or CIFAR, then work through conventional algorithms before reaching deep learning, so the choices you make along the way are the ones a real project forces on you. It began as a project for a Harvard graduate data science course, and that teaching background shows in how carefully each step is explained, with a conda environment ready to get the notebook running. A PyTorch rewrite sits alongside the original TensorFlow version, which makes it easy to compare how the same pipeline looks in two frameworks. Anyone learning machine learning seriously, or an instructor looking for a substantive class project, will get real value out of it.