Core concepts

Train, Validation, and Test Sets

Generalization

Bias and Variance

Regularization

Data Leakage

Metrics

The Perceptron

From Bernoulli to Log Loss

Loss and Optimization

Batches, Epochs, and Early Stopping

An epoch is one complete pass through the training data; batches split that pass into successive weight updates.

Backpropagation

Convolution

Convolution slides a small filter across an image to produce a map showing where a local pattern appears.

Forecasting Baselines

A forecasting baseline is a simple rule that a more complex model must beat on the same future windows.

Rolling-Origin Backtesting

Rolling-origin backtesting evaluates a forecaster at several historical cutoff dates while preserving the direction of time.

Probabilistic Forecasting

Probabilistic forecasting describes a distribution of plausible future values instead of pretending that one path is certain.