Clustering in Image Quantization

$0/1$ point $($graded$)$

If we use $K=3,$ which of the following will be the compressed image? K$-$Means Drawbacks

$\frac{0}{1}$ point $($graded$)$

Which of the following are drawbacks of the K$-$means algorithm with Euclidean distance $($as presented so far in

this lecture$)?$ Select all those apply.

Does not scale well to large datasets

Manual choice of $K$

Does not scale well with increasing number of dimensions Gaussian Mixture Model and the EM Algorithm

$0/1$ point $($graded$)$

Which of the following statements are true? Assume that we have a Gaussian mixture model with known $($or

estimated$)$ parameters $($means and variances of the Gaussians and the mixture weights$).$

A Gaussian mixture model can provide information about how likely it is that a given point belongs to

each cluster.

The EM algorithm converges to the same estimate of the parameters irrespective of the initialized

values.

An iteration of the EM algorithm is computationally more expensive $($in terms of order complexity$)$ when

compared to an iteration of the K$-$means algorithm for the same number of clusters.