Section outline

  • Tasks for the lab on unsupervised learning. 

    TASK 1: Implement in Python a method for Parzen density estimation with Gaussian kernels (accepting a generic kernel function), and test it on 1d and 2d data. Use (cross) validation to set the lengthscale of the kernel. 

    TASK 2: Run clustering algorithms, exploring the scikit-learn library (kmeans, gaussian mixtures, hierarchical clustering, spectral clustering), on 2d density data and on any other dataset you like (for classification and regression datasets, ignore output features).

    TASK 3: Experiment with PCA from scikit-learn on the dataset(s) available for the course.