Next: Extension to N-D Up: R. Clapp: A modified Previous: R. Clapp: A modified

# Lloyd's Method

Lloyd's 1-D method attempts to identify the major features in a signal's histogram as accurately as possible, with as few points as possible. The selected points are referred to as the signal's ``codebook''. The methodology is iterative in nature. It starts from an initial code book, then repeats
1
Find for every point the closest codebook entry. These points form a cell.
2
Find the average of all points associated with a given codebook entry. The average becomes the new codebook entry.
3
Remove codebook entries that have no, or few, points associated with them.
Figure  demonstrates a single iteration of this approach. Steps 1-3 are repeated until the solution no longer changes significantly.

 ref Figure 1 In step 1, the data is dividing into cells. In step 2, the average 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 is attempting to solve a non-linear problem by an iterative scheme, so local minima can be problematic. To avoid prejudicing the solution the initial model is often a random set of vectors. To solve the problem, several methods have been developed Linde et al. (1980). These methods often rely on replacing an empty cell by splitting regions with high variance. Clapp (2002) followed a modified version of this approach for the velocity selection problem. Figure  shows a velocity model, the selected reference velocities, and the percentage error between the two using the methodology described in Clapp (2002). The methodology proved effective in selecting appropriate reference velocities with two noticeable drawbacks. At each depth step a new non-linear estimate is performed. At some depths steps the solution gets stuck in local-minima. Note the striping in the right panel of Figure . In addition, the methodology will sometimes pick too few reference velocities to adequately describe a layer by getting stuck in a local minima. Both of these problems can be minimized by using as a starting solution the reference velocities at the previous depth step and adding a random component to the cell splitting. Figure  shows the selected reference velocities and the percentage error in the selected reference velocities.

error1
Figure 2
The left panel shows a velocity model from a 2-D synthetic. The center panel shows the reference velocities selected using the Lloyd's approach described in Clapp (2002). The right panel shows the percent error between the selected reference velocity and the the true velocity. Note the striping due to local minima.

error2
Figure 3
The left panel shows the selected reference velocity using an improved selection scheme. The right panel shows the error percentage error between the selected reference and the true velocity. Note the reduced error compared to the right panel of Figure .

Next: Extension to N-D Up: R. Clapp: A modified Previous: R. Clapp: A modified
Stanford Exploration Project
4/5/2006