Housing regression and MNIST classification
California housing, MNIST and a spam filter: the classics, built end to end.
Pythonscikit-learnPyTorchJupyter
The rite-of-passage trio, California housing, MNIST and a spam filter, each built as a full pipeline instead of a tutorial to copy.
Things I implemented:
- pandas cleaning and exploration before any model
- A dumb baseline first, every time
- scikit-learn models with cross-validation for the housing and spam tasks
- A small PyTorch network for MNIST
- An error-analysis pass at the end of each notebook, reading what the model got wrong