EasyDeepLearn

What is kernel density estimation?

medium

Answer

  • Non-parametric density estimate: KDE(x) = (1/(nh)) Σ K((x    xi)/h)K((x\; - \;x_{i}) / h), where K is a kernel (usually Gaussian) and h is bandwidth.
  • Bandwidth choice is the key knob: too small → wiggly, too large → over-smoothed.
  • Silverman's rule: h ≈ 1.06 * σ * n^(-1/5).
  • Better: cross-validation.
  • Smoother alternative to histograms for continuous data; also foundation of some non-parametric classifiers (KDE-NB, Parzen windows).
Check yourself — multiple choice
  • Parametric only
  • Sum of scaled kernels around each data point; bandwidth h controls smoothness; Silverman / CV to pick h
  • Same as histogram
  • Requires normality

KDE: sum of kernels; bandwidth is the key hyperparameter (Silverman or CV).

#descriptive#eda#estimation

Practise Statistics Fundamentals

215 interview questions in this topic.

Related questions