Robust Nearest Neighbour Retrieval Using Targeted Manifold Manipulation
Banibrata Ghosh, Haripriya Harikumar, Santu Rana · 2025
Read Robust Nearest Neighbour Retrieval Using Targeted Manifold Manipulation by Banibrata Ghosh, Haripriya Harikumar, Santu Rana (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-2511-06261.txt.
Description
Ghosh, Harikumar, and Rana redefine what it means for two images to be neighbors. Nearest-neighbour retrieval usually depends on hand-tuned feature layers and distance measures. Their method, TMM-NN, instead asks how easily a sample can be nudged into a chosen region of the model's feature space.
A small patch is added to the query image, and the network is lightly steered so patched inputs drift toward a dummy class. Images truly similar to the query need only a slight push and land there confidently, while unrelated ones resist. Ranking by that confidence retrieves the most meaningful neighbors, and tests showed the ranking stays robust under noise.
Catalog record
| CREATOR | Banibrata Ghosh, Haripriya Harikumar, Santu Rana |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 40 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-06261