Python代写|COMS 4701 Artificial Intelligence Homework 5 – Conceptual


1. Perceptron versus Naive Bayes classifiers

Compare the perceptron and naive Bayes. How are they different and how are they alike? provide at least four
comparison statements.

2. Perceptron

Consider the two dimensional examples (0, 0), (0, 1), (1, 0), (1, 1) that are all positive examples, and the
examples, (2.1, 0), (0, 2.5), (1.6, 1.6) that are all negative examples.

(a) Are these examples linearly separable? Show your work.

(b) Find a perceptron that separates the examples. Use the step function that outputs 0 for negative examples
and 1 otherwise.

3. K-means Clustering

(a) In the K-means lecture, we initialized the center clusters µ1; · · · µk randomly. Discuss how effective it would
be to initializing all the center clusters µ1; · · · µk to a vector of zeros.

(b) Should the cost function J seen in lecture , decrease, increase or stay the same after each iteration of

(c) We know that K-means is sensitive to the initialization of the centers. The figures below show a dataset
with three clusters. Each figure has a different initialization of the four centers (the black points are the
centers). Which pair of initialization will lead to the same clustering? Mark all that apply.

4. Interpretability

You have a received a training data set and the best performing method to predict the label of interest is a
random forests.

You need to hand your classifier and discuss with collaborators the reason of the classification. Showing one
thousand trees is not satisfactory (even if each individual tree is an explanation by itself).

Propose an algorithm to extract explanations from a random forest model. Please be specific and clear about
the type of representation and interpretability method you propose. Feel free to research the topic and cite all
your sources.