Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of Graph Search.
This lecture covers Kernel Density Estimation (KDE) focusing on the bandwidth selection and the curse of dimensionality. It explains the smoothed plug-in estimator, the kernel density estimator, and the asymptotic risk of KDE. The lecture discusses the bias-variance tradeoff, optimal rates for KDE, and bandwidth selection methods like pilot estimator and least squares cross-validation. It also explores the estimation of integrated mean squared error and the leave-one-out cross-validation estimator. Additionally, it delves into the generalization of KDE to higher dimensions, emphasizing the importance of bandwidth selection. The lecture concludes with a comparison between parametric and nonparametric models, highlighting the tradeoff between flexibility and efficiency.