This exercise is divided in 3 parts:
-
Probability density estimation (PDF) of univariate data (kernel.csv) using Kernel (non-parametric) methods. I fit the Bandwidth hyperparameter of the Epanechnikov kernel using the mean integrated square error (MISE).
-
Gaussian Mixture Models hyperparameter tunning (covariance type and number of components) of multivariate data (diabetes.csv). I use K-fold Cross-Validation approach to calculate the maximum log-likelihood and therefore select the best model.
-
Same case as in point 1, this time I use K-fold Cross-Validation approach to tune the Bandwidth hyperparameter of the Epanechnikov kernel.