Self-Balancing, Memory Efficient, Dynamic Metric Space Data Maintenance, for Rapid Multi-Kernel Estimation

Aditya S Ellendula, Chandrajit Bajaj · 2025

Read Self-Balancing, Memory Efficient, Dynamic Metric Space Data Maintenance, for Rapid Multi-Kernel Estimation by Aditya S Ellendula, Chandrajit Bajaj (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-18003.txt.

Description

Aditya Ellendula and Chandrajit Bajaj built in 2025 a self-balancing tree that keeps neighborhood data fresh as points move. Many learning systems live in shifting spaces, and rebuilding the map from scratch each step is costly. Their two-knob octree folds space into nested boxes and rebalances as data drift. Updates and searches then run in logarithmic time.

Four uses showed the gains. Particle-based training ran more particles at less cost. Nearest-neighbor sorting updated in real time. Search for retrieval-augmented models stayed fast as the store grew. Joint tuning of inputs and hidden states used samples better. Across the four, speed rose steeply while accuracy held, above all in high dimensions.

Catalog record

CREATORAditya S Ellendula, Chandrajit Bajaj
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE81 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-18003

Browse the full library · About the Tower of Babel