Writing productivity
productivity 3 min read 11 January 2023

Learning Resources: Machine Learning, Product Analytics, and Career

A focused collection of learning resources for applied Machine Learning practitioners - from opportunity sizing frameworks to Machine Learning process courses and the embedding vs. dense layer distinction.

The Principle

Trying to learn multiple new things simultaneously produces shallow understanding across all of them. The more effective approach: pick one concept, skill, or domain, exhaust the best resources for it, then move on. This list is organized by topic so each item can be treated as a focused sprint.

Applied Machine Learning and Product

Machine Learning for Product Analytics by Ron Tidhar - how Machine Learning integrates with product analytics work in practice. Covers the organizational and prioritization questions that technical tutorials ignore: when is Machine Learning the right tool, how do you size the opportunity, how do you communicate results to stakeholders.

Intro to Opportunity Sizing - framing the “how much is this worth” question before building. The discipline of sizing opportunities before scoping technical work is one of the most underrated skills in applied Machine Learning.

Ken Jee’s Machine Learning Process Course - structured walkthrough of the end-to-end Machine Learning project lifecycle from a practitioner perspective. Covers problem framing, data exploration, modeling, and communication.

Neural Network Fundamentals

Embedding vs Dense Layers - a question that trips up many practitioners: what’s the architectural difference between using an Embedding layer and a Dense layer for the same task? The answer involves weight sharing, lookup-table efficiency, and the semantic expectations each creates.

The practical answer: Embedding layers are lookup tables. They map a discrete index to a dense vector without computing a dot product. Dense layers apply a weight matrix to a continuous input. Use embeddings for categorical features (especially vocabulary) where sparse one-hot encoding would be prohibitively wide. Use dense layers for continuous feature transformation.

Comprehensive References

Best of Machine Learning Python - ranked, curated list of the Python Machine Learning ecosystem organized by category. Better than searching PyPI when you need to know what the community consensus best-in-class library is for a specific task.

How to Use This List

Pick one resource. Finish it. Then pick another. The value is in depth, not breadth. Skimming five resources on the same topic produces less understanding than finishing one.

learning resources career machine-learning product-analytics
← All articles

Have a problem worth solving?

Whether you need a quantitative researcher, a Machine Learning systems builder, or a technical advisor, I take a small number of consulting engagements at a time.

Book a call →