ML_AI_HW_1
Machine Learning and Artificial Intelligence - Homework #1 - A.Y. 2018/2019 - Politecnico di Torino
In this homework you will dirty your hands on PCA applied on images. You have to show what happens if different principal components (PC) are chosen as basis for images representation and classification. Then, you have to choose and apply a classifier in order to..classify the images ;-) under different PC re-projection and you should comment the obtained results. In particular, here is the brief list of this homework sub-tasks:
-
Download and load the provided subset of PACS dataset; setup your programming environ- ment accordingly your needings.
-
Choose one image and shows what happens to the image when you re-project it with only first 60 PC, first 6 PC, first 2 PC, last 6 PC. Comment the results.
-
Using scatter-plot, visualize the dataset projected on first 2 PC. Repeat the exercise with only 3&4 PC, and with 10&11. What do you notice? Justify your answer from theoretical perspective behind PCA.
-
Classify the dataset (divided into training and test set) using a Naive Bayes Classifier in those cases: unmodified images, images projected into first 2PC, and on 3&4 PC. Show accuracy and compare results: what are your conclusions?
-
(Optional) Visualize decision boundaries of the classier in the first 2 PC case. Any consider- ation about those boundaries?