Detecting Scarce and Sparse Anomalous: Solving Dual Imbalance in Multi-Instance Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Jia, Lin-Han, Guo, Lan-Zhe, Zhou, Zhi, Han, Si-Ye, Li, Zi-Wen, Li, Yu-Feng
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914057771024384
author Jia, Lin-Han
Guo, Lan-Zhe
Zhou, Zhi
Han, Si-Ye
Li, Zi-Wen
Li, Yu-Feng
author_facet Jia, Lin-Han
Guo, Lan-Zhe
Zhou, Zhi
Han, Si-Ye
Li, Zi-Wen
Li, Yu-Feng
contents In real-world applications, it is highly challenging to detect anomalous samples with extremely sparse anomalies, as they are highly similar to and thus easily confused with normal samples. Moreover, the number of anomalous samples is inherently scarce. This results in a dual imbalance Multi-Instance Learning (MIL) problem, manifesting at both the macro and micro levels. To address this "needle-in-a-haystack problem", we find that MIL problem can be reformulated as a fine-grained PU learning problem. This allows us to address the imbalance issue in an unbiased manner using micro-level balancing mechanisms. To this end, we propose a novel framework, Balanced Fine-Grained Positive-Unlabeled (BFGPU)-based on rigorous theoretical foundations. Extensive experiments on both synthetic and real-world datasets demonstrate the effectiveness of BFGPU.
format Preprint
id arxiv_https___arxiv_org_abs_2503_13562
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Detecting Scarce and Sparse Anomalous: Solving Dual Imbalance in Multi-Instance Learning
Jia, Lin-Han
Guo, Lan-Zhe
Zhou, Zhi
Han, Si-Ye
Li, Zi-Wen
Li, Yu-Feng
Machine Learning
Artificial Intelligence
In real-world applications, it is highly challenging to detect anomalous samples with extremely sparse anomalies, as they are highly similar to and thus easily confused with normal samples. Moreover, the number of anomalous samples is inherently scarce. This results in a dual imbalance Multi-Instance Learning (MIL) problem, manifesting at both the macro and micro levels. To address this "needle-in-a-haystack problem", we find that MIL problem can be reformulated as a fine-grained PU learning problem. This allows us to address the imbalance issue in an unbiased manner using micro-level balancing mechanisms. To this end, we propose a novel framework, Balanced Fine-Grained Positive-Unlabeled (BFGPU)-based on rigorous theoretical foundations. Extensive experiments on both synthetic and real-world datasets demonstrate the effectiveness of BFGPU.
title Detecting Scarce and Sparse Anomalous: Solving Dual Imbalance in Multi-Instance Learning
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2503.13562