Homeworks
- H1: Decision trees and Evaluation
- statement release: September 26th
- deadline: Friday, October 7th
- statement release: September 26th
- H2: Bayesian learning and kNN
- statement release: October 7th
- deadline: Monday, October 17th
- statement release: October 7th
- H3: Neural networks and Regression
- statement release: October 17th
- deadline: Friday, October 28th
- statement release: October 17th
- H4: Clustering and PCA
- statement release: October 28th
- deadline: Monday, November 7th
- statement release: October 28th
FAQ
HW3
- 1. On question I.3: the MLP should be
learnt over the original univariate data space (i.e., please ignore the
basis function to answer this question).
- 1. On question I.2: Do we need to learn multiple classifiers for the different assumptions OR a single Bayesian classifier satisfying all the three assumptions? Answer: The later. A single Bayesian classifier considering that {y1,y2} follow a multivariate distribution, y3 is independent from the remaining variables, and y3 follows a Gaussian distribution.
- 2. On question I.1: weights according to the heuristic introduced in the lectures and labs, 1/d(x_i,x_j).
- 3. On question II.2: follow the principles in the Evaluation notebook. There is no need to assess whether the t-Test assumptions are satisfied.
- 4. On question II.3: only two reasons are sufficient, you can draw hypotheses based on the behavior of the classifiers, correction will accommodate plausible reasonings even if they are not strictly observed for the given dataset.
HW3
- 1. On question II.6: use two scatter plots to produce the visualization