Conditional Prediction

Due October 13, 2026 at 3:59 AM

Introduces the notion of making predictions using features, how to encode data into useful features, multiple regression, and classification via logistic regression.

Parts

Part 1: What Is Known Before Departure?

Fix the flight, target, prediction time, and feature boundary before measuring association.

Part 2: Loss Chooses the Baseline

Derive and inspect why the mean minimizes squared error while the median minimizes absolute error.

Part 3: From One Feature to a Matrix

Implement ordinary least squares and move from a fitted line to a multi-feature estimator.

Part 4: Features Create Geometry

Standardize numerical fields and encode categories without inventing numeric distances between names.

Part 5: Regularization and the Train–Test Gap

Trace how L2 strength changes training error, test error, and their separation under both MAE and MSE.

Part 6: Build an Image Representation

Derive shape measurements from an intensity matrix, implement them, and compare their logistic probabilities with raw pixels.

Part 7: Listen, Measure, Classify

Play NSynth notes, implement waveform descriptors, and test transfer to instruments absent during fitting.

Part 8: Probabilities Become Decisions

Move decision thresholds and preserve the false-positive and false-negative counts that accuracy collapses.