|
ref
Figure 1 In step 1, the data is dividing into cells. In step2, the centroid of each cell is calculated. In the third step new cell boundaries are calculated between the centroids. If a cell contains no points, the cell is removed. | ![]() |
Lloyd's method can be effective but relies on a good initial
codebook and can get stuck in local minima. To avoid prejudicing the solution
the initial model is often a random set of vectors. To solve the second
problem, several methods have been developed ().
These methods often rely on replacing empty cell by
splitting regions with high variance.