Machine Learning Pills
Articles about Machine Learning Pills
RW #11 - Why Learn Machine Learning When AI Can Do the Work
💊 Pill of the Week
An Animated Guide to Vector Stores (and How Similarity Search Really Works)
How can a database search millions of vectors in just a few milliseconds?
Weekly Dose #22 - AI Is Getting Cheaper. Your Model Choices Just Got More Interesting
📰 The Weekly Dose
An Animated Guide to Choosing the Right ML Model
There are hundreds of machine learning models. So which one should you pick? The honest answer is: it depends. In this animated explainer, we turn “it depends” into a simple decision process. Fittingly, it looks like a decision tree.
An Animated Guide to Large Language Models
You type a question and a whole paragraph appears, as if the machine understood you.
Issue #142 - Overfitting and How to Detect It
💊 Pill of the Week
An Animated Guide to Linear Regression
Can you predict the price of a house just from its size?
An Animated Guide to Time Series
Some data comes with a clock attached, and that changes everything. Please subscribe to my YouTube channel to support the project. Thanks! Visit the YouTube channel In this short animated explainer, we look at what makes time series special, where they show up, and how to forecast them without fooling yourself.
Weekly Dose #21 - Route Better, Predict Faster, Sandbox Everything
📰 The Weekly Dose
An Animated Guide to ReAct Agents
Ask a plain language model a tricky question and it might confidently invent the answer. Give it tools and a loop, and it can actually work the answer out. In this animated explainer, we break down ReAct, the pattern behind many AI agents, and watch one solve a real question step by step.
An Animated Guide to Reading Box Plots (and Why the Median Wins)
When someone says “the average salary is 50k”, what does that really tell you?
Issue #142 - Data Leakage: When Historical Data Gives Your Model an Unfair Advantage
Why a clean historical dataset can give your ML model an unfair advantage Imagine you build a model to decide which subscription price to show a user. You train it on historical data, test it on a later period, and get a great result. The model seems ready to launch.
An Animated Guide to Decision Trees
Watch now (4 mins) | How does a machine learn to make decisions?
Weekly Dose #20 - The AI Stack Is Changing Faster Than the Models
📰 The Weekly Dose
An Animated Guide to Causal Inference
Did the tutoring program raise exam scores, or does it just look that way? In this short animated explainer, we start with a simple comparison that seems to show tutoring works. Then we find out why that comparison can fool us.
Issue #141 - SHAP Values: Useful, Misleading, or Both?
You have seen the plot. Someone opens a slide, a beeswarm chart appears, and the room nods along. Red dots on the right, blue on the left, features ranked top to bottom. Whatever the model was doing, it now looks understood. SHAP has become the default answer to “why did the model say that?” It is in every notebook template, every model card, every regulatory conversation. And it deserves a lot of that popularity. It is one of the few explanation methods with real mathematical grounding, and it works on models that are otherwise opaque.
Weekly Dose #19 - Context, Voice, Multimodality
📰 The Weekly Dose
Weekly Dose #18 - From Agent Sandboxes to 9 Billion DNA Predictions
📰 The Weekly Dose