A Focus on Important Samples for Out-of-Distribution Detection
To ensure the reliability and security of machine learning classification models when deployed in the open world, it is crucial that these models can detect out-of-distribution (OOD) data that exhibits semantic shifts from the in-distribution (ID) data used during training. This necessity has spurre...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-06-01
|
| Series: | Mathematics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2227-7390/13/12/1998 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849432935610974208 |
|---|---|
| author | Jiaqi Wan Guoliang Wen Guangming Sun Yuntian Zhu Zhaohui Hu |
| author_facet | Jiaqi Wan Guoliang Wen Guangming Sun Yuntian Zhu Zhaohui Hu |
| author_sort | Jiaqi Wan |
| collection | DOAJ |
| description | To ensure the reliability and security of machine learning classification models when deployed in the open world, it is crucial that these models can detect out-of-distribution (OOD) data that exhibits semantic shifts from the in-distribution (ID) data used during training. This necessity has spurred extensive research on OOD detection. Previous methods required a large amount of finely labeled OOD data for model training, which is costly or performed poorly in open-world scenarios. To address these limitations, we propose a novel method named <i>focus on important samples</i> (FIS) in this paper. FIS leverages model-predicted OOD scores to identify and focus on important samples that are more beneficial for model training. By learning from these important samples, our method aims to achieve reliable OOD detection performance while reducing training costs and the risk of overfitting training data, thereby enabling the model to better distinguish between ID and OOD data. Extensive experiments across diverse OOD detection scenarios demonstrate that FIS achieves superior performance compared to existing approaches, highlighting its robust and efficient OOD detection performance in practical applications. |
| format | Article |
| id | doaj-art-3923dfcf442b4cdfaeafc9fdd8815574 |
| institution | Kabale University |
| issn | 2227-7390 |
| language | English |
| publishDate | 2025-06-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Mathematics |
| spelling | doaj-art-3923dfcf442b4cdfaeafc9fdd88155742025-08-20T03:27:14ZengMDPI AGMathematics2227-73902025-06-011312199810.3390/math13121998A Focus on Important Samples for Out-of-Distribution DetectionJiaqi Wan0Guoliang Wen1Guangming Sun2Yuntian Zhu3Zhaohui Hu4College of Electronic Engineering, National University of Defense Technology, Hefei 230037, ChinaCollege of Electronic Engineering, National University of Defense Technology, Hefei 230037, ChinaCollege of Electronic Engineering, National University of Defense Technology, Hefei 230037, ChinaCollege of Electronic Engineering, National University of Defense Technology, Hefei 230037, ChinaCollege of Computer Science and Technology, National University of Defense Technology, Changsha 410073, ChinaTo ensure the reliability and security of machine learning classification models when deployed in the open world, it is crucial that these models can detect out-of-distribution (OOD) data that exhibits semantic shifts from the in-distribution (ID) data used during training. This necessity has spurred extensive research on OOD detection. Previous methods required a large amount of finely labeled OOD data for model training, which is costly or performed poorly in open-world scenarios. To address these limitations, we propose a novel method named <i>focus on important samples</i> (FIS) in this paper. FIS leverages model-predicted OOD scores to identify and focus on important samples that are more beneficial for model training. By learning from these important samples, our method aims to achieve reliable OOD detection performance while reducing training costs and the risk of overfitting training data, thereby enabling the model to better distinguish between ID and OOD data. Extensive experiments across diverse OOD detection scenarios demonstrate that FIS achieves superior performance compared to existing approaches, highlighting its robust and efficient OOD detection performance in practical applications.https://www.mdpi.com/2227-7390/13/12/1998out-of-distribution detectionreliable machine learningactive learning |
| spellingShingle | Jiaqi Wan Guoliang Wen Guangming Sun Yuntian Zhu Zhaohui Hu A Focus on Important Samples for Out-of-Distribution Detection Mathematics out-of-distribution detection reliable machine learning active learning |
| title | A Focus on Important Samples for Out-of-Distribution Detection |
| title_full | A Focus on Important Samples for Out-of-Distribution Detection |
| title_fullStr | A Focus on Important Samples for Out-of-Distribution Detection |
| title_full_unstemmed | A Focus on Important Samples for Out-of-Distribution Detection |
| title_short | A Focus on Important Samples for Out-of-Distribution Detection |
| title_sort | focus on important samples for out of distribution detection |
| topic | out-of-distribution detection reliable machine learning active learning |
| url | https://www.mdpi.com/2227-7390/13/12/1998 |
| work_keys_str_mv | AT jiaqiwan afocusonimportantsamplesforoutofdistributiondetection AT guoliangwen afocusonimportantsamplesforoutofdistributiondetection AT guangmingsun afocusonimportantsamplesforoutofdistributiondetection AT yuntianzhu afocusonimportantsamplesforoutofdistributiondetection AT zhaohuihu afocusonimportantsamplesforoutofdistributiondetection AT jiaqiwan focusonimportantsamplesforoutofdistributiondetection AT guoliangwen focusonimportantsamplesforoutofdistributiondetection AT guangmingsun focusonimportantsamplesforoutofdistributiondetection AT yuntianzhu focusonimportantsamplesforoutofdistributiondetection AT zhaohuihu focusonimportantsamplesforoutofdistributiondetection |