Deep Metric Learning Using Negative Sampling Probability Annealing
Multiple studies have concluded that the selection of input samples is key for deep metric learning. For triplet networks, the selection of the anchor, positive, and negative pairs is referred to as triplet mining. The selection of the negatives is considered the be the most complicated task, due to...
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-10-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/22/19/7579 |
| _version_ | 1797476866956722176 |
|---|---|
| author | Gábor Kertész |
| author_facet | Gábor Kertész |
| author_sort | Gábor Kertész |
| collection | DOAJ |
| description | Multiple studies have concluded that the selection of input samples is key for deep metric learning. For triplet networks, the selection of the anchor, positive, and negative pairs is referred to as triplet mining. The selection of the negatives is considered the be the most complicated task, due to a large number of possibilities. The goal is to select a negative that results in a positive triplet loss; however, there are multiple approaches for this—semi-hard negative mining or hardest mining are well-known in addition to random selection. Since its introduction, semi-hard mining was proven to outperform other negative mining techniques; however, in recent years, the selection of the so-called hardest negative has shown promising results in different experiments. This paper introduces a novel negative sampling solution based on dynamic policy switching, referred to as negative sampling probability annealing, which aims to exploit the positives of all approaches. Results are validated on an experimental synthetic dataset using cluster-analysis methods; finally, the discriminative abilities of trained models are measured on real-life data. |
| first_indexed | 2024-03-09T21:09:51Z |
| format | Article |
| id | doaj.art-1cd94dae4e18470da3aee6274736591e |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-09T21:09:51Z |
| publishDate | 2022-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-1cd94dae4e18470da3aee6274736591e2023-11-23T21:51:27ZengMDPI AGSensors1424-82202022-10-012219757910.3390/s22197579Deep Metric Learning Using Negative Sampling Probability AnnealingGábor Kertész0John von Neumann Faculty of Informatics, Obuda University, 1034 Budapest, HungaryMultiple studies have concluded that the selection of input samples is key for deep metric learning. For triplet networks, the selection of the anchor, positive, and negative pairs is referred to as triplet mining. The selection of the negatives is considered the be the most complicated task, due to a large number of possibilities. The goal is to select a negative that results in a positive triplet loss; however, there are multiple approaches for this—semi-hard negative mining or hardest mining are well-known in addition to random selection. Since its introduction, semi-hard mining was proven to outperform other negative mining techniques; however, in recent years, the selection of the so-called hardest negative has shown promising results in different experiments. This paper introduces a novel negative sampling solution based on dynamic policy switching, referred to as negative sampling probability annealing, which aims to exploit the positives of all approaches. Results are validated on an experimental synthetic dataset using cluster-analysis methods; finally, the discriminative abilities of trained models are measured on real-life data.https://www.mdpi.com/1424-8220/22/19/7579negative sampling probability annealingtriplet miningdeep metric learningtriplet network |
| spellingShingle | Gábor Kertész Deep Metric Learning Using Negative Sampling Probability Annealing Sensors negative sampling probability annealing triplet mining deep metric learning triplet network |
| title | Deep Metric Learning Using Negative Sampling Probability Annealing |
| title_full | Deep Metric Learning Using Negative Sampling Probability Annealing |
| title_fullStr | Deep Metric Learning Using Negative Sampling Probability Annealing |
| title_full_unstemmed | Deep Metric Learning Using Negative Sampling Probability Annealing |
| title_short | Deep Metric Learning Using Negative Sampling Probability Annealing |
| title_sort | deep metric learning using negative sampling probability annealing |
| topic | negative sampling probability annealing triplet mining deep metric learning triplet network |
| url | https://www.mdpi.com/1424-8220/22/19/7579 |
| work_keys_str_mv | AT gaborkertesz deepmetriclearningusingnegativesamplingprobabilityannealing |