Optimizing Contrastive Learning with Semi-Online Triplet Mining

Contrastive learning is a machine learning technique in which models learn by contrasting similar and dissimilar data points. Its goal is to learn a representation of data in such a way that similar instances are close together in the representation space, while dissimilar instances are far apart. O...

Full description

Saved in:
Bibliographic Details
Main Authors: Przemysław Buczkowski, Marek Kozłowski, Piotr Brzeziński
Format: Article
Language:English
Published: MDPI AG 2025-07-01
Series:Applied Sciences
Subjects:
Online Access:https://www.mdpi.com/2076-3417/15/14/7865
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849246576066691072
author Przemysław Buczkowski
Marek Kozłowski
Piotr Brzeziński
author_facet Przemysław Buczkowski
Marek Kozłowski
Piotr Brzeziński
author_sort Przemysław Buczkowski
collection DOAJ
description Contrastive learning is a machine learning technique in which models learn by contrasting similar and dissimilar data points. Its goal is to learn a representation of data in such a way that similar instances are close together in the representation space, while dissimilar instances are far apart. Our industrial use case focuses on a special case of contrastive learning called triplet learning. Building triplets with adequate difficulty is crucial to effective training convergence in such a setup. By combining online and offline mining techniques, we propose a method of mining hard triplets that is both performant and memory-inexpensive. Our experiments demonstrate that the method leads to improved identity pairing (which is the specific case of clustering) both on a real-life industry shoe dataset and on a generated benchmark one.
format Article
id doaj-art-af74977fbcd3455b80801084f6e07eb6
institution Kabale University
issn 2076-3417
language English
publishDate 2025-07-01
publisher MDPI AG
record_format Article
series Applied Sciences
spelling doaj-art-af74977fbcd3455b80801084f6e07eb62025-08-20T03:58:27ZengMDPI AGApplied Sciences2076-34172025-07-011514786510.3390/app15147865Optimizing Contrastive Learning with Semi-Online Triplet MiningPrzemysław Buczkowski0Marek Kozłowski1Piotr Brzeziński2National Information Processing Institute, 00-608 Warsaw, PolandNational Information Processing Institute, 00-608 Warsaw, PolandVive Textile Recycling, 25-663 Kielce, PolandContrastive learning is a machine learning technique in which models learn by contrasting similar and dissimilar data points. Its goal is to learn a representation of data in such a way that similar instances are close together in the representation space, while dissimilar instances are far apart. Our industrial use case focuses on a special case of contrastive learning called triplet learning. Building triplets with adequate difficulty is crucial to effective training convergence in such a setup. By combining online and offline mining techniques, we propose a method of mining hard triplets that is both performant and memory-inexpensive. Our experiments demonstrate that the method leads to improved identity pairing (which is the specific case of clustering) both on a real-life industry shoe dataset and on a generated benchmark one.https://www.mdpi.com/2076-3417/15/14/7865machine learningcontrastive learningrepresentation learningtriplet learningtriplet mining
spellingShingle Przemysław Buczkowski
Marek Kozłowski
Piotr Brzeziński
Optimizing Contrastive Learning with Semi-Online Triplet Mining
Applied Sciences
machine learning
contrastive learning
representation learning
triplet learning
triplet mining
title Optimizing Contrastive Learning with Semi-Online Triplet Mining
title_full Optimizing Contrastive Learning with Semi-Online Triplet Mining
title_fullStr Optimizing Contrastive Learning with Semi-Online Triplet Mining
title_full_unstemmed Optimizing Contrastive Learning with Semi-Online Triplet Mining
title_short Optimizing Contrastive Learning with Semi-Online Triplet Mining
title_sort optimizing contrastive learning with semi online triplet mining
topic machine learning
contrastive learning
representation learning
triplet learning
triplet mining
url https://www.mdpi.com/2076-3417/15/14/7865
work_keys_str_mv AT przemysławbuczkowski optimizingcontrastivelearningwithsemionlinetripletmining
AT marekkozłowski optimizingcontrastivelearningwithsemionlinetripletmining
AT piotrbrzezinski optimizingcontrastivelearningwithsemionlinetripletmining