Books, papers & marginalia

Reading

A working library of ideas worth returning to—what I’m reading, what stayed with me, and the notes left in the margins.

On the desk

3 publications

Hover a spine to open the stack

Drag to rotateDouble-click to reset

Book2020

Mathematics for Machine Learning

by Deisenroth, Faisal & Ong

From my notesCurrently reading

The useful part is not memorizing another catalogue of formulas; it is learning to move between the geometry, the algebra, and the probabilistic story of the same model.

A focused bridge between undergraduate mathematics and modern machine learning, covering linear algebra, analytic geometry, matrix decompositions, probability, optimization, and the models built on top of them.

Open original

Drag to rotateDouble-click to reset

Book2025

Hands-On Machine Learning with Scikit-Learn and PyTorch

by Aurélien Géron

From my notesCurrently reading

Best treated as a workbench rather than a cover-to-cover read: build the examples, break the pipelines, and return to the theory when the behaviour stops being obvious.

A practical path from classical machine-learning workflows to deep neural networks, transformers, diffusion models, and reinforcement learning using Scikit-Learn, PyTorch, and the Hugging Face ecosystem.

Open original

Drag to rotateDouble-click to reset

NeurIPS 20172017

Attention Is All You Need

by Vaswani et al.

From my notesCurrently reading

The lasting idea is architectural subtraction. Once attention can connect every position directly, recurrence is no longer a necessary organising principle—and an entire new model family becomes possible.

The paper that introduced the Transformer: an encoder-decoder architecture built entirely from self-attention and feed-forward layers, removing recurrence and making sequence training substantially more parallel.

Open original