Interactive notebook
Predict, Experiment, Reflect
Predict first, then ask what the calculation leaves out. These public demonstrations complement, rather than replace, the courses.
Runs in your browser: nothing is saved and no account is needed. Optional AI features and Paper Q&A show their own notice and send nothing until you submit.
STOR 323IValues meet metrics
Who gets discovered?
A fictional recommendation platform has 100 established and 100 new creators. A reference evaluation labels 60 established and 40 new creators as relevant. Choose a policy before inspecting its outcomes.
All counts are synthetic, fixed teaching examples. These are not deployment recommendations, student assessments, or claims that a fairness metric identifies a unique right answer.
Inspect the fit, then explain it
| Strength | Training MSE | Reference MSE | Coefficient norm |
|---|
Objective, data, and what this cannot prove
Minimize mean((y-Xw)^2) + lambda sum(w^2) in the basis 1,x,x^2,...,x^5. All six coefficients, including the intercept, are regularized. The program solves the six-by-six normal equations with pivoting.
Training inputs are twelve equally spaced points on [-1,1]. The reference function is 0.7x-0.4x^2+0.2sin(3x); fixed noise is 0.28sin((i+1)2.399963). This deterministic example is chosen to illustrate both modest regularization and excessive shrinkage, not to estimate typical performance.
Reference MSE uses 101 noiseless grid points, not independent real-world test data. Basis scaling affects ridge regression. This one example cannot establish a universal generalization claim or justify tuning against a final test set.
03Question your explanation
An optional AI reflection coach
Ask for a conceptual hint or examine a value conflict. The program recomputes the chosen example; AI guidance is not a grade, a uniquely correct ethics verdict, or your instructor speaking personally.
The optional reason in the local policy form is not automatically included. Only this question and the selected challenge settings are sent when you click below.