Introduction
Many real datasets have lots of features: a 28 × 28 image has 784, a customer profile might have 500, a gene-expression sample 20,000. Each feature is a dimension. We're used to thinking in 2 or 3 dimensions, but high-dimensional spaces behave in ways that defy intuition — and these surprises cause real problems for ML. Together they're called the curse of dimensionality.
The good news: the next topics show how to reduce the number of dimensions while keeping most of the information, which speeds up training, helps visualisation, and sometimes improves accuracy.