Excellent posts:
- Rules of ML from Google
- A Recipe for Training Neural Networks by Andrej Karpathy
I learn best from reading, not coursera-like videos (though who didn’t enjoy Andrew Ng’s course?). Papers are important to keep up with the latest, but to fill in the background usually left unexplained and to expose the bits I never knew I didn’t know: I turn to textbooks.
Starting out, I read a bunch of textbooks and as far as I know these are still good starting points:
Neural Networks and Deep Learning by Michael Nielsen: back prop, loss functions, problems when neural networks go deep
Machine Learning Yearning by Andrew Ng: how to train a model and ensure it works in the real world
Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow by Aurélien Géron: it’s all in the title, great book
On more specific topics, I can recommend:
Interpretable Machine Learning by Christoph Molnar: discusses the inherently explainable models and how to interpret black-box models.
Reinforcement Learning by Sutton and Barto: a slow building up of concepts to some of the most powerful RL algorithms and the language to understand the latest research.
Designing Data-Intensive Applications by Martin Kleppman: brilliant book not specifically about data science. Read this to understand the difficulties navigating big data.
Mining of Massive Datasets by Rajaraman and Ullman: a surprising deep and thorough book on scaling algorithms to big data, covering map-reduce, stream processing, PageRank and recommenders at scale.
Beyond purely technical topics, I also suggest:
- UX for beginners as a quick intro to an important role in product development that (can) interact with data scientists
- Weapons of Math Destruction on the damage of unacknowledged bias in models
- Gödel’s Proof because it’s linked to the “No free lunch” theorem and generally mind-blowing