From Nearest-Neighbour Classification to Attention

Chun Ma; Yanpeng Qu; Longzhi Yang

doi:10.1109/icaci65340.2025.11096289

From Nearest-Neighbour Classification to Attention

Chun Ma, Yanpeng Qu, Longzhi Yang

School of Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Multi-functional nearest-neighbour (MFNN) provides a unified framework that is capable of implementing multiple nearest-neighbour algorithms, such as k nearest-neighbour, fuzzy nearest-neighbour, fuzzy-rough nearest-neighbour algorithms. In this paper, the flexibility of the framework of MFNN is reviewed based on some interesting commonalities shared with MFNN and the attention mechanism, in terms of the way to represent the query, key and value. As a result, a new implementation of MFNN, MFNN-AT, is proposed based on the attention mechanism. Experimental results verify the effectiveness of this novel classification method.

Original language	English
Title of host publication	2025 17th International Conference on Advanced Computational Intelligence (ICACI)
Place of Publication	Piscataway, NJ
Publisher	IEEE
Pages	237-242
Number of pages	6
ISBN (Electronic)	9798331509798
ISBN (Print)	9798331509804
DOIs	https://doi.org/10.1109/icaci65340.2025.11096289
Publication status	Published - 7 Jul 2025
Event	2025 17th International Conference on Advanced Computational Intelligence - Bath, United Kingdom Duration: 7 Jul 2025 → 13 Jul 2025

Conference

Conference	2025 17th International Conference on Advanced Computational Intelligence
Abbreviated title	ICACI
Country/Territory	United Kingdom
City	Bath
Period	7/07/25 → 13/07/25

Keywords

Nearest-neighbour
Classification
Attention
Similarity relation

Access to Document

10.1109/icaci65340.2025.11096289

Cite this

@inproceedings{5c23dcaf37d54c4fa28408bbcef579e4,

title = "From Nearest-Neighbour Classification to Attention",

abstract = "Multi-functional nearest-neighbour (MFNN) provides a unified framework that is capable of implementing multiple nearest-neighbour algorithms, such as k nearest-neighbour, fuzzy nearest-neighbour, fuzzy-rough nearest-neighbour algorithms. In this paper, the flexibility of the framework of MFNN is reviewed based on some interesting commonalities shared with MFNN and the attention mechanism, in terms of the way to represent the query, key and value. As a result, a new implementation of MFNN, MFNN-AT, is proposed based on the attention mechanism. Experimental results verify the effectiveness of this novel classification method.",

keywords = "Nearest-neighbour, Classification, Attention, Similarity relation",

author = "Chun Ma and Yanpeng Qu and Longzhi Yang",

year = "2025",

month = jul,

day = "7",

doi = "10.1109/icaci65340.2025.11096289",

language = "English",

isbn = "9798331509804",

pages = "237--242",

booktitle = "2025 17th International Conference on Advanced Computational Intelligence (ICACI)",

publisher = "IEEE",

address = "United States",

note = "2025 17th International Conference on Advanced Computational Intelligence, ICACI ; Conference date: 07-07-2025 Through 13-07-2025",

}

TY - GEN

T1 - From Nearest-Neighbour Classification to Attention

AU - Ma, Chun

AU - Qu, Yanpeng

AU - Yang, Longzhi

PY - 2025/7/7

Y1 - 2025/7/7

N2 - Multi-functional nearest-neighbour (MFNN) provides a unified framework that is capable of implementing multiple nearest-neighbour algorithms, such as k nearest-neighbour, fuzzy nearest-neighbour, fuzzy-rough nearest-neighbour algorithms. In this paper, the flexibility of the framework of MFNN is reviewed based on some interesting commonalities shared with MFNN and the attention mechanism, in terms of the way to represent the query, key and value. As a result, a new implementation of MFNN, MFNN-AT, is proposed based on the attention mechanism. Experimental results verify the effectiveness of this novel classification method.

AB - Multi-functional nearest-neighbour (MFNN) provides a unified framework that is capable of implementing multiple nearest-neighbour algorithms, such as k nearest-neighbour, fuzzy nearest-neighbour, fuzzy-rough nearest-neighbour algorithms. In this paper, the flexibility of the framework of MFNN is reviewed based on some interesting commonalities shared with MFNN and the attention mechanism, in terms of the way to represent the query, key and value. As a result, a new implementation of MFNN, MFNN-AT, is proposed based on the attention mechanism. Experimental results verify the effectiveness of this novel classification method.

KW - Nearest-neighbour

KW - Classification

KW - Attention

KW - Similarity relation

UR - https://www.scopus.com/pages/publications/105013685991

U2 - 10.1109/icaci65340.2025.11096289

DO - 10.1109/icaci65340.2025.11096289

M3 - Conference contribution

SN - 9798331509804

SP - 237

EP - 242

BT - 2025 17th International Conference on Advanced Computational Intelligence (ICACI)

PB - IEEE

CY - Piscataway, NJ

T2 - 2025 17th International Conference on Advanced Computational Intelligence

Y2 - 7 July 2025 through 13 July 2025

ER -

From Nearest-Neighbour Classification to Attention

Abstract

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this