Consider using a K-means clustering pass over the leaf set to determine how to build Wagner trees in parallel. @Boarders has a solid vision of how to move forward on implementing this.
The K-mean pass should converge in linear time with a K constant factor. The hypothesis is that this will provide a better/faster tree construction, by using clustering to determine the leaf set permutation instead of a random permutation.
Consider using a K-means clustering pass over the leaf set to determine how to build Wagner trees in parallel. @Boarders has a solid vision of how to move forward on implementing this.
The K-mean pass should converge in linear time with a K constant factor. The hypothesis is that this will provide a better/faster tree construction, by using clustering to determine the leaf set permutation instead of a random permutation.