Activity: Create an SVM candidate review
Estimated active time: 45-75 minutes
Scenario
Use the synthetic learner-support dataset. The fictional platform wants to decide which model families deserve a place in an algorithm portfolio.
Task
Fit a linear SVM and an RBF SVM, tune C and gamma over a small grid, and report the sensitivity you find.
Required sections
- Problem and decision.
- Validation score for the linear and RBF kernels.
- A small grid over C and gamma, with the score at each point.
- How much the score moves across that grid — the sensitivity note.
- Keep, delay, or reject against the Module 1 baselines.
- Limitation note.
Rules
- Scale features first. An unscaled SVM is dominated by whichever column has the widest range.
- Report the grid, not only its best cell. The spread is the finding.
- Do not tune on the test set.
Output
Use the relevant learner-resource template from learner-resources/ or write a one-page memo with the same sections.