Partition in this slide seems similar to the K-means algorithm in machine learning. However, is the efficiency of this algorithm more powerful than K-means even though it doesn't apply the gradient descent?