Rishi Jain

Picking up the linear-algebra thread from the basics notes, where I left eigenvectors sitting on the “next” list. These follow Zachary Huang’s Give me 25 min, I will make Eigenvalue click forever, worked out with the small numerical examples and a NumPy snippet per section. Same format as last time,...

I wanted to go back and pin down the basics of linear algebra properly, since so much of what I want to learn next leans on them, so these are my notes from working back through the foundations, cleaned up enough to share. They follow Zachary Huang’s Give Me 30...

The first LSTM post was mostly a tour of the PyTorch API, the row-by-row reshape and the tuple-shaped hidden state and which line to keep for a many-to-one head, and it left the actual training algorithm sitting unexplained underneath the whole thing, so these are the notes that go after...

After two convolutional projects, MNIST and FER2013, the first video of a recurrent-network tutorial drops you straight back onto MNIST and asks you to classify the digits with an LSTM, which sounds like a step backwards until you see the trick in the framing. You take the 28×28 image and...

Before the MNIST classifier and the FER2013 classifier there were five Jupyter notebooks, which are homework rather than a project, though the project posts that came after do not make much sense without them sitting underneath. The notebooks are on GitHub, and they are my pass through Daniel Bourke’s Learn...