← Back to QuantegyAI

Machine Learning Foundations

Eight interactive modules · about 3–4 hours · Courses 1–2 are the prerequisite. No coding required.

Track your progress. Sign in to save module completion and your mastery scores across devices. Your progress also saves on this device automatically. Open the portal →

This course constitutes the foundational core of machine learning. It builds on Introduction to AI (the concepts) and Math Foundations (the tools), translating them into working models. By the end you will understand the modeling workflow, fit regression and classification models, train naive Bayes, decision trees, k-means clustering, and PCA, and — most importantly — evaluate a model rigorously, so that you can distinguish a genuine result from a misleading one.

Every module is hands-on and runs entirely in your browser: you will adjust a regression line to reduce the error, observe k-means centroids converge, grow a decision tree split by split, and observe a model overfit. Each module also presents the corresponding scikit-learn code for later recognition. Each module concludes with a short mastery check; passing it marks the module complete.

Start here

Module 1

The Modeling Workflow

Data → features → model → loss → train → evaluate, and the train/test split that ensures honest evaluation. Supervised vs. unsupervised; regression vs. classification. Activity: run an example through a live pipeline, then classify real tasks by type.

Supervised learning

Module 2

Regression — Fitting a Line

Predicting a number. Residuals, squared error, and the line of best fit. Activity: drag a line over real points and watch the error fall — then let it auto-fit. AI anchor: price and demand prediction.

Module 3

Classification — Drawing Boundaries

Predicting a category. k-nearest-neighbors, decision boundaries, and probability outputs. Activity: drop a new point and watch its neighbors vote, with an adjustable k. AI anchor: spam, medical, and image classifiers.

Module 4

Naive Bayes

Turning Bayes’ rule into a classifier. The independence assumption and why it works anyway. Activity: a live spam filter — toggle words and watch the spam probability update. AI anchor: real spam and sentiment filters.

Module 5

Decision Trees

Binary tests that partition the data; impurity, greedy splits, and depth. Activity: grow a tree with a depth slider and observe it partition the plane — and begin to overfit. AI anchor: random forests and gradient boosting.

Unsupervised learning

Module 6

Clustering — k-means

Finding groups with no labels at all. Centroids, the assign-and-update loop, and choosing k. Activity: step through k-means and observe the clusters form; inertia decreases each round. AI anchor: customer segmentation.

Module 7

Dimensionality Reduction — PCA

Too many features; find the directions that carry the most signal. Variance, principal components, and projection. Activity: rotate a projection line and watch retained variance peak on the first component. AI anchor: visualizing embeddings.

Project

Module 8 · Project

Honest Evaluation — Overfitting, Cross-Validation & Bias–Variance

The methodology that distinguishes rigorous machine learning from self-deception. Observe a model overfit as its complexity increases — training error decreases while test error increases — and learn how cross-validation and the bias–variance trade-off support rigorous evaluation. Integrates every earlier module.

Capstone

Capstone

Build a Concept Manipulative

Put it all together: build a single-page interactive that teaches one ML Foundations concept, then submit it for grading and your certificate.

Why this matters next This is the trunk every branch grows from. The training loop, the loss, and honest evaluation you learn here are exactly what neural networks and deep learning (Course 4) scale up — a deep net is this same workflow with many layers and millions of weights. Master the trunk and the branches are far easier.
Next course → Neural Networks & Deep Learning — eight hands-on modules where this same workflow goes deep: a single neuron, activation functions, a network that solves XOR, gradient descent and backpropagation live, what depth buys you on a spiral, regularization, and an end-to-end capstone. Train every network in your browser. No coding required.

← Back to QuantegyAI