how apply k-means on a n by m matrix based on rows

I want to apply a kmeans clustering with cosine distance on a n by m matrix and I want to define each row as a sample. how can I do that?

For example, I have a 30x50 matrix and I want to cluster them based on cosine similarity of each of the rows. Means that I have 50 samples. Consider each row as an image that I want to cluster them based on cosine similarity.

Answer 1

[Image Analyst]

So you have 30 images, and for each image there are 50 measurements/features/samples or whatever you want to call the data. So why can't you just call kmeans

numClusters = 4; % Whatever you want.
[indexes, clusterCentroids] = kmeans(data, numClusters, 'distance', 'cosine');

Why do you want that distance metric (cosine) instead of any of the other, more common, distance metrics?