Machine Learning, Visualised
Linear regression to gradient descent — no calculus required
About this course
We build intuition for the math behind ML using interactive concepts and analogies. By the end you'll see what a model is actually doing.
What you'll learn
- Read and interpret loss curves
- Intuit gradient descent visually
- Choose between regression, classification, clustering
- Spot overfitting in the wild
Curriculum
End-to-End ML Project — From CSV to Web Demo
Pick any tabular dataset (Kaggle, OpenML, your own). Build and ship: 1. EDA notebook — at least 3 visualisations that revealed something. 2. A baseline model (linear/logistic regression). 3. A stronger model (random forest or gradient boosting). 4. Cross-validated metrics for both. Report at least 3 metrics. 5. One paragraph explaining which model you'd ship and why. 6. A simple Streamlit/Gradio demo where someone can type/select inputs and see a prediction. Submit notebook + demo link.
- 1EDA reveals at least one non-obvious insight
- 2Both models trained, hyperparameters explicit
- 3Metrics are cross-validated, not single-split
- 4Choice of model justified with trade-offs (latency, interpretability)
- 5Demo accepts user input and returns a prediction without crashing
What graduates shipped.
Once you ship, your project lives here — linked from your public verify page so recruiters can click through.