Relaxed Balance through Standard Rotations.
Kim S. Larsen, Eljas Soisalon-Soininen, and Peter Widmayer.
In 5th International Workshop on Algorithms and Data Structures (WADS), volume 1272 of Lecture Notes in Computer Science, pages 450-461. Springer, 1997.
In search trees with relaxed balance, rebalancing transformations need not be connected with updates, but may be delayed. For standard AVL tree rebalancing, we prove that even though the rebalancing operations are uncoupled from updates, their total number is bounded by O(M log (M+N)), where M is the number of updates to an AVL tree of initial size N. Hence, relaxed balancing of AVL trees comes at no extra cost asymptotically. Furthermore, our scheme differs from most other relaxed balancing schemes in an important aspect: No rebalancing transformation can be done in the wrong direction, i.e., no performed rotation can make the tree less balanced. Moreover, each performed rotation indeed corresponds to a real imbalance situation in the tree. Finally, and perhaps most importantly, our structure is capable of forgetting registered imbalance if later updates happen to improve the situation. Our results are of theoretical interest and have possible sequential and parallel applications.


publication
Link to the publication at the publisher's site - subscription may be required.
Text required by the publisher (if any): The final publication is available at link.springer.com.

full version
Link to the journal version containing all the material and proofs, some of which are usually omitted in the conference version due to space constraints.

other publications
Other publications by the author.