Question

Question 4 1 pts Which of the following reasons is not the reason why the K-means algorithm will likely end up with sub-optimal clustering? (Select all that apply.) Bad choices for the initial cluster centers. Choosing a k that corresponds to the number of natural clusters in the dataset. Fast convergence of the K-means algorithm. Existence of closely located data samples in the dataset. Question 5 1 pts Which of the following is a step in K-means algorithm implementation? (Select all that apply.) ✓ Set initial cluster centers. Compute the distance between every data point and each cluster center. Assign each data point to its nearest cluster center. Initialize the cluster centers to random points at each iteration Question 6 1 pts Which of the following is true for the K-means algorithm? (Select all that apply.) The value of K needs to be set before running the algorithm On every iteration of the algorithm, the cost function should either stay the same or decrease (in particular, it should not increase) Once a data point has been assigned to a particular center, it will no longer be re-assigned to a different center in later iterations. The algorithm will produce the same set of centers regardless of the initialization.

Answer 1

Question 4:

correct options: 2,3

Choosing the correct value of k is essential for optimal clustering. There is no fixed value of k, we usually have to try out a range of values of k. A value of k that corresponds to natural number of clusters is a good place to start.

Fast convergence of K-means will not lead to sub-optimal clustering. It may be an indicator of K-means getting stuck in a local optima but it will not cause poor performance. If anything fast convergence can be computationally efficient.

Question 5:

correct options: 1,2,3

Setting initial cluster centers, Computing the distance between every point and cluster center and assigning that point to that cluster center are all steps in K-means implementation.

The cluster centers are modified on every iteration based on the cost function not by random initialization.

Question 6:

correct options: 1,2

The value of k needs to be selected before running K-means.

With every iteration, cost function must either stay the same or decrease or else our K-means algorithm is not working.

Data points can be assigned to any cluster center on any iteration based on the distance between the point and center.

Different initializations may lead to different clusters since kmeans algorithm may stuck in a local optimum and may not converge to global optimum. So it’s recommended to run the algorithm using different initializations of centroids and pick the one with better performance.