sutskever-30-implementations

Why we included this project

This collection works through all 30 papers on Ilya Sutskever's well-known reading list, with each one implemented as a Jupyter notebook that relies only on NumPy. That constraint keeps the mechanics visible: attention heads, recurrent cells, and residual blocks show up as plain Python arrays rather than opaque framework calls. Every notebook includes synthetic data and plots, so you can watch activations and gradients build up line by line as you step through a transformer or a neural Turing machine. It suits engineers refreshing their fundamentals and educators looking for runnable examples to hand out. This is a study aid for deep learning's foundations, not production code.