DeDrift: Robust Similarity Search under Content Drift

TL;DR

DeDrift is a method that maintains high accuracy in large-scale similarity search over time by adaptively updating indexing structures, significantly reducing the need for costly reindexing despite content drift.

Contribution

We introduce DeDrift, a novel approach that dynamically updates embedding quantizers to counteract content drift in similarity search systems.

Findings

DeDrift reduces accuracy degradation caused by content drift.

DeDrift is up to 100 times faster than full index reconstruction.

We provide real-world datasets with long-term temporal information.

Abstract

The statistical distribution of content uploaded and searched on media sharing sites changes over time due to seasonal, sociological and technical factors. We investigate the impact of this "content drift" for large-scale similarity search tools, based on nearest neighbor search in embedding space. Unless a costly index reconstruction is performed frequently, content drift degrades the search accuracy and efficiency. The degradation is especially severe since, in general, both the query and database distributions change. We introduce and analyze real-world image and video datasets for which temporal information is available over a long time period. Based on the learnings, we devise DeDrift, a method that updates embedding quantizers to continuously adapt large-scale indexing structures on-the-fly. DeDrift almost eliminates the accuracy degradation due to the query and database content drift while being up to 100x faster than a full index reconstruction.