Self-Supervised Representation Learning: Introduction, advances, and challenges
Open accessThe DOI matches its Crossref record; 111 of the 125 references it cites were found in a registry or an index, and 1 are not checked yet.
Abstract
Self-supervised representation learning (SSRL) methods aim to provide powerful, deep feature learning without the requirement of large annotated data sets, thus alleviating the annotation bottleneck—one of the main barriers to the practical deployment of deep learning today. These techniques have advanced rapidly in recent years, with their efficacy approaching and sometimes surpassing fully supervised pretraining alternatives across a variety of data modalities, including image, video, sound, text, and graphs. This article introduces this vibrant area, including key concepts, the four main families of approaches and associated state-of-the-art techniques, and how self-supervised methods are applied to diverse modalities of data. We further discuss practical considerations including workflows, representation transferability, and computational cost. Finally, we survey major open challenges in the field, that provide fertile ground for future work.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source doesn’t support the claim as written.
In the citing paper
“The various self-supervised pretexts are divided into five broad families: transformation prediction, masked prediction, instance discrimination, clustering, and contrastive instance discrimination.”
From Self-Supervised Learning for Time Series Analysis: Taxonomy, Progress, and Prospects (2024) · cites this paper as [28]
Checked by a reasoning model — may be wrong.
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“Contrastive learning [56-58] is in fact a training procedure used in recent artificial intelligence models as a way to efficiently learn semantic representations.”
From How Adversarial REM Dreams May Facilitate Creativity, and Why We Become Aware of Them (2024) · cites this paper as [58]
The cited source says what the claim attributes to it.
In the citing paper
“For example, Ericsson et al. [28] provides a very comprehensive review of pretext tasks for multiple modalities, including image, video, text, audio, time series, and graph.”
From Self-Supervised Learning for Time Series Analysis: Taxonomy, Progress, and Prospects (2024) · cites this paper as [28]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 4×
“State of the art architectures usually start with CNN representation encoding hθ(·), with ResNet [1] being widely used, before appending task-specific”
InfluentialVaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404 · cited 3×
“long been used to model the context while recent works have moved to Transformer-based architectures with self-attention [2], which allow longer range connections to be made across words in a sentence, but require more data for training.”
Exploiting Generative AI to Scale up Intelligent Tutoring Systems
Jakubův, J., Chvalovský, K., Goertzel, Z., Kaliszyk, C., Olšák, M., Piotrowski, B., Schulz, S., Suda, M., & Urban, J. (2023) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/lipics.itp.2023.19 · Cited by 76,658
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173 · cited 9×
“Self-supervised strategies have been leveraged successfully to improve sample efficiency of learning across a variety of modalities, from image [5]–[7], video [8], [9], speech [10], [11], text [12], [13], and graphs [14], [15].”
InfluentialImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333 · cited 2×
“As in much of computer vision, ImageNet [35] is the most typical source data set for self-supervised pretraining [5]–[7], [40], consisting of 1.28 million training images across 1,000 object categories, with the most commonly used resolution at 224 × 224.”
InfluentialBERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1423 · Cited by 54,325 · cited 9×
“Self-supervised strategies have been leveraged successfully to improve sample efficiency of learning across a variety of modalities, from image [5]–[7], video [8], [9], speech [10], [11], text [12], [13], and graphs [14], [15].”
InfluentialAn Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al. (2020) · arXiv · doi:10.48550/arxiv.2010.11929 · Cited by 47,472 · cited 1×
“One notable version is a vision transformer [58], which is increasingly being used by recent self-supervised methods on image data [43].”
The Elements of Statistical Learning
Hastie, T., Tibshirani, R., & Friedman, J. (2009) · Springer Series in Statistics · doi:10.1007/978-0-387-84858-7 · Cited by 25,970 · cited 2×
“These range from classic shallow approaches such as Gaussian mixtures [21] to deep methods such as variational autoencoders (VAEs) and generative adversarial networks (GANs) [18].”
Semi-supervised classification with graph convolutional networks
Kipf, T. N., & Welling, M. (2016) · arXiv · doi:10.48550/arxiv.1609.02907 · Cited by 24,853 · cited 1×
“Graph NNs (GNNs) [89] have advanced all these tasks significantly, especially where massive-labeled data are available.”
Distributed Representations of Words and Phrases and their Compositionality
Mikolov, T., Sutskever, I., Chen, K., Corrado, G., & Dean, J. (2013) · arXiv · doi:10.48550/arxiv.1310.4546 · Cited by 24,533 · cited 6×
“With data annotation being a major bottleneck, NLP was the first discipline to make major and successful use of self-supervision [22].”
Influential
Research works printed but not linked: 4
- 0, Youtube-8m A large and diverse labeled video dataset for video understanding research
- gokaslan, 2019, OpenWebText Corpus
- radford, 2019, Language Models are Unsupervised Multitask Learners
- goodfellow, 2016, Deep Learning
Cited by
Items 101–200 of 641
A focus on molecular representation learning for the prediction of chemical properties
Harnik, Y., & Milo, A. (2024) · Chemical Science · doi:10.1039/d4sc90043j · Cited by 28
“4,6 The design of a pretraining encoder depends on the format of its inputs; for example, molecules or reactions can be represented as graphs, strings, and images.”
BackgroundWong, D. R., Hill, A. S., & Moccia, R. (2025) · Bioinformatics · doi:10.1093/bioinformatics/btaf317 · Cited by 27
“For single-cell biology, this can be a foundational understanding of complex gene networks learnable via self-supervision 22 .”
Unsupervised Learning of Disentangled Representation via Auto-Encoding: A Survey
Eddahmani, I., Pham, C.-H., Napoléon, T., Badoc, I., Fouefack, J.-R., & Elbouz, M. (2023) · Sensors · doi:10.3390/s23042362 · Cited by 27
Ruan, S., Wang, R., Shen, X. Y., Liu, H., Xiao, B., Shi, J., Zhang, K., Huang, Z., Liu, Y., Chen, E., et al. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2506.19769 · Cited by 27
“Moreover, techniques like self-supervised representation learning [135] and contrastive learning [136] offer the potential for capturing and utilizing cross-modal relationships, providing richer and more detailed super-vision to refine alignment precision.”
Deldari, S., Xue, H., Saeed, A., He, J., Smith, D. V., & Salim, F. D. (2022) · arXiv · doi:10.48550/arxiv.2206.02353 · Cited by 26
“However, it is still infeasible to apply the cross-entropy objective function across large datasets [49].”
InfluentialProgress and new challenges in image-based profiling
Serrano, E., Peters, J., Wagner, J., Graham, R. E., Chen, Z.-H., Feng, B. Y., Miranda, G., Kalinin, A. A., Vulliard, L., Tomkinson, J., et al. (2026) · Molecular Systems Biology · doi:10.1038/s44320-026-00197-7 · Cited by 7
“Self-supervised learning is a training strategy that relies on unannotated image data by de fi ning objective functions that guide the model to learn meaningful representations from the data itself (Ericsson et al, 2021).”
A Survey of Pathology Foundation Model: Progress and Future DirectionsPreprint
Xiong, C., Chen, H., & Sung, J. J. �. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2504.04045 · Cited by 26
“SSL leverages unlabeled data by automatically generating supervisory signals through pretext tasks [ Ericsson et al. , 2022 ] .”
Zhang, Z., Ren, Z., Tao, C., Zhang, Y., Peng, C., & Li, H. (2023) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2023.3336285 · Cited by 25
He, C., Cheng, J., Chen, G., Guan, Q., Fei, X., & Tang, Y. Z. (2024) · Information Fusion · doi:10.1016/j.inffus.2024.102254 · Cited by 25
Moody, J. B., Poitrasson-Rivière, A., Renaud, J. M., Hagio, T., Alahdab, F., Al-Mallah, M. H., Vanderver, M. D., Goonewardena, S. N., Ficaro, E. P., & Murthy, V. L. (2025) · NEJM AI · doi:10.1056/aioa2500164 · Cited by 24
Pandiyan, V., Wróbel, R., Richter, R. A., Leparoux, M., Leinenbach, C., & Shevchik, S. A. (2023) · Materials & Design · doi:10.1016/j.matdes.2023.112458 · Cited by 24
Dense lead contrast for self-supervised representation learning of multilead electrocardiograms
Liu, W., Li, Z., Zhang, H., Chang, S., Wang, H., He, J. L., & Huang, Q. (2023) · Information Sciences · doi:10.1016/j.ins.2023.03.099 · Cited by 24
Dual-Kernel Adapter: Expanding Spatial Horizons for Data-Constrained Medical Image Analysis
Zhu, Z., Zhu, H., Lu, S., Li, X., Meng, Y., Jin, G., Yin, L., Hu, L., Wang, D., Liu, L., et al. (2026) · International Conference on Learning Representations · doi:10.48550/arxiv.2602.18888 · Cited by 24
“…2019; Kim et al., 2022; Kora et al., 2022), semi-supervised learning (Huynh et al., 2022; Chebli et al., 2018; Wang et al., 2020a; Han et al., 2024a; Zhou et al., 2019), and self-supervised representation learning (Ericsson et al., 2022; Jiao et al., 2020; Ye et al., 2024; Krishnan et al., 2022).”
Guan, C., Chen, F., Yuan, L., Zhang, Z., & Yu, Y. (2024) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2024.3420791 · Cited by 23
“Motivated by recent advances in state representation learning [54, 33], which reveals that auxiliary representation objectives could facilitate policy learning [14], we aim at efficiently aggregating information as compact representations for policy by designing a novel communication framework Multi-Agent communicatio…”
Shi, G., Lu, H., Hui, H., & Tian, J. (2025) · Medical Image Analysis · doi:10.1016/j.media.2024.103442 · Cited by 21
How Well Do Self-Supervised Models Transfer to Medical Imaging?
Anton, J., Castelli, L., Chan, M. F., Outters, M., Tang, W. H., Cheung, V., Shukla, P., Walambe, R. A., & Kotecha, K. V. (2022) · Journal of Imaging · doi:10.3390/jimaging8120320 · Cited by 21
“See [27] for a more comprehensive discussion and survey of SSL techniques.”
BackgroundLearning Representations for Multilead Electrocardiograms From Morphology-Rhythm Contrast
Liu, W., Zhang, H., Chang, S., Wang, H., He, J. L., & Huang, Q. (2024) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2024.3369152 · Cited by 21
“Recently, self-supervised representation learning (SSRL) has made rapid progress in unlabeled data utilization [16].”
BackgroundInfluentialDash, B. M., Bouamama, B. O., Pékpé, K. M., & Boukerdja, M. (2023) · Neurocomputing · doi:10.1016/j.neucom.2024.127871 · Cited by 21
Scene-Centric Unsupervised Panoptic Segmentation
Hahn, O., Reich, C., Araslanov, N., Cremers, D., Rupprecht, C., & Roth, S. (2025) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52734.2025.02280 · Cited by 21
“Self-supervised representation learning focuses on learning generic feature extractors from unlabeled data, aiming for expressive features that facilitate a broad range of down-stream tasks [25].”
FNBench: Benchmarking Robust Federated Learning against Noisy LabelsPreprint
Jiang, X., Li, J., Wu, N., Wu, Z., Li, X., Sun, S., Xu, G., Wang, Y., Li, Q., & Liu, M. (2024) · doi:10.36227/techrxiv.172503083.36644691/v1 · Cited by 20
Zaheer, Q., Wang, J., Shah, S. M. A. H., Ehsan, H., Shah, S. F. H., Ai, C., Kuang, J., Wang, W., & Qiu, S. (2025) · Journal of Civil Structural Health Monitoring · doi:10.1007/s13349-025-00960-8 · Cited by 20
Dar, S. S., Karandikar, M. K., Rehman, M. Z. U., Bansal, S., & Kumar, N. (2024) · Knowledge-Based Systems · doi:10.1016/j.knosys.2024.112526 · Cited by 20
ViLReF: An Expert Knowledge Enabled Vision-Language Retinal Foundation ModelPreprint
Yang, S., Du, J., Guo, J., Zhang, W., Liu, H., Li, H., & Wang, N. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2408.10894 · Cited by 20
“Sample pairs are selected during pre-training dynamically [17].”
Flap: Fast Language-Audio Pre-Training
Yeh, C.-F., Huang, P.-Y., Sharma, V., Li, S.-W. D., & Gosh, G. (2023) · IEEE Workshop on Automatic Speech Recognition and Understanding (ASRU) · doi:10.1109/asru57964.2023.10389692 · Cited by 19
“Recently, self-supervised representation learning (SSL) [2, 3] has emerged as a prominent research area in hope of reducing human annotations.”
BackgroundExploring Fragment Adding Strategies to Enhance Molecule Pretraining in AI-Driven Drug Discovery
Meng, Z., Chen, C., Zhang, X., Zhao, W. H., & Cui, X. (2024) · Big Data Mining and Analytics · doi:10.26599/bdma.2024.9020003 · Cited by 19
A Survey of Deep Learning: From Activations to Transformers
Schneider, J., & Vlachos, M. (2024) · International Conference on Agents and Artificial Intelligence (ICAART) · doi:10.5220/0012404300003636 · Cited by 19
Visual DNA: Representing and Comparing Images Using Distributions of Neuron Activations
Ramtoula, B., Gadd, M., Newman, P. M., & Martini, D. D. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.01069 · Cited by 18
“The DNA creation exploits the recent progress in selfsupervised representation learning [13, 18] and extracts image descriptors directly from patterns of neuron activations in neural networks (NNs).”
BackgroundZhang, Z., Lu, H., Ma, S., Peng, J., Lin, C., Li, N., & Dong, B. (2024) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2024.106788 · Cited by 18
Li, T., Guo, Y., Zhao, Z., Chen, M., Lin, Q., Hu, X., Yao, Z., & Hu, B. (2024) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2024.3380357 · Cited by 18
“Unsupervised/self-supervised learning, an important branch of machine learning, can explore the relationships in data by maximizing intra-sample similarity [18], [19].”
BackgroundCAD - Contextual Multi-modal Alignment for Dynamic AVQA
Nadeem, A., Hilton, A., Dawes, R., Thomas, G., & Mustafa, A. (2024) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv57701.2024.00709 · Cited by 18
“On the other hand, self-supervised learning (SSL) is used to train transformer models using unlabeled data, unlike supervised learning that relies on labeled data [22].”
MethodYoon, J., & Choi, M.-K. (2023) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw59228.2023.00333 · Cited by 18
“We expect that more distinct feature representations can be developed for keyframe extraction with the help of strong self-supervised methods [9].”
BackgroundReview of multimodal machine learning approaches in healthcarePreprint
Krones, F., Marikkar, U., Parsons, G., Szmul, A., & Mahdi, A. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2402.02460 · Cited by 18
“To mitigate the disadvantages of supervised pre-training, self-supervised learning aims to train neural networks to learn data representations by leveraging the inherent properties of the data [69].”
BackgroundJointly spatial-temporal representation learning for individual trajectories
Fei, H., Lv, J., & Yue, Y. (2024) · Computers Environment and Urban Systems · doi:10.1016/j.compenvurbsys.2024.102144 · Cited by 18
“Representation learning is regarded as a crucial technique for foundation models because it offers a unified vector representation of information to effectively capture and comprehend a subject matter (Bengio et al., 2013; Bommasani et al., 2022; Ericsson et al., 2022).”
BackgroundContrastive Audio-Language Learning for Music
Manco, I., Benetos, E., Quinton, E., & Fazekas, G. (2022) · arXiv · doi:10.48550/arxiv.2208.12208 · Cited by 17
What Comes After Transformers? A Selective Survey Connecting Ideas in Deep LearningGPT
Schneider, J. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-87330-0_4 · Cited by 17
GSSCL: A framework for Graph Self-Supervised Curriculum Learning based on clustering label smoothing
Fu, Y.-G., Chen, X., Xu, S., Li, J., Yao, X. F., Huang, Z., & Wang, Y. (2024) · Neural Networks · doi:10.1016/j.neunet.2024.106787 · Cited by 17
Open set classification of sound event
You, J., Wu, W., & Lee, J. (2024) · Scientific Reports · doi:10.1038/s41598-023-50639-7 · Cited by 17
“In the self-supervised training stage, the raw audio sequences first transform into different audio sequences, but the intra meaning of audio is not modified.”
Selective HuBERT: Self-Supervised Pre-Training for Target Speaker in Clean and Mixture Speech
Lin, J., Ge, M., Wang, W., Li, H., & Feng, M. (2024) · IEEE Signal Processing Letters · doi:10.1109/lsp.2024.3383794 · Cited by 17
“S ELF-SUPERVISED pre-training has gained much popularity among the researchers in recent years [1], [2], [3]. pre-training By pre-training the models on large amounts of unannotated speech data, not only is the cost of annotation saved, but the models also generate universal speech representations…”
BackgroundPhysics-Guided Dual Self-Supervised Learning for Structure-Based Material Property Prediction
Fu, N., Wei, L., & Hu, J. (2024) · The Journal of Physical Chemistry Letters · doi:10.1021/acs.jpclett.4c00100 · Cited by 17
“It is a barrier to the development of supervised learning DNNs, especially for the applications that data and annotations are rare, costly, or time-consuming to collect [11].”
BackgroundFeed-Forward SceneDINO for Unsupervised Semantic Scene Completion
Jevtić, A., Reich, C., Wimbauer, F., Hahn, O., Rupprecht, C., Roth, S., & Cremers, D. (2025) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51701.2025.00638 · Cited by 17
“Self-supervised representation learning (SSL) aims to extract general features from data without annotations, facilitating various downstream tasks such as segmentation [24].”
Backdoor Contrastive Learning via Bi-level Trigger Optimization
Sun, W., Zhang, X., Lü, H., Chen, Y.-C., Wang, T., Chen, J., & Lin, L. (2024) · International Conference on Learning Representations · doi:10.48550/arxiv.2404.07863 · Cited by 17
“The superior performance of CL makes it a popular choice in modern machine learning designs (Ericsson et al., 2022), where a feature extractor is pre-trained on large-scale unlabeled data, based on which different predictors can be customized to serve a variety of downstream tasks.”
BackgroundBundele, V., Sarıtaş, K., Kargi, B., Çal, O. A., Tezören, K., Ghaderi, Z., & Lensch, H. P. A. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2412.19124 · Cited by 17
“Self-supervised learning has emerged as a powerful solution to these limitations, enabling models to learn rich representations from unlabeled data prior to task-specific fine-tuning with minimal labeled samples [2, 6, 22, 25, 39, 41, 73].”
Bian, S., Liu, M., Ray, L. S. S., Zhou, B., Guo, B., Yu, Z., Plötz, T., Lukowicz, P., Yuan, S., & Rey, V. F. (2026) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3810230 · Cited by 17
He, X., Jin, J., Chen, D., & Feng, Y. (2024) · Journal of Manufacturing Processes · doi:10.1016/j.jmapro.2024.02.006 · Cited by 16
Hu, J., Huang, Z., Yin, X., Ruan, W., Cheng, G., Dong, Y., & Huang, X. (2025) · neural information processing systems · doi:10.52202/085713-3419 · Cited by 16
Atitallah, S. B., Driss, M., Boulila, W., & Koubâa, A. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-70819-0_7 · Cited by 16
Fang, C., Sandino, C. M., Mahasseni, B., Minxha, J., Pouransari, H., Azemi, E., Moin, A., & Zippi, E. L. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2410.16424 · Cited by 16
“Pretraining models with self-supervised objectives is a popular and effective strategy in machine learning [Ericsson et al., 2022, Gui et al., 2023].”
MethodA Probabilistic Model Behind Self-Supervised Learning
Bizeul, A., Schölkopf, B., & Allen, C. (2024) · Transactions on machine learning research · doi:10.3929/ethz-b-000712860 · Cited by 16
Comparing self-supervised learning techniques for wearable human activity recognition
Ek, S., Presotto, R., Civitarese, G., Portet, F., Lalanda, P., & Bettini, C. (2025) · CCF Transactions on Pervasive Computing and Interaction · doi:10.1007/s42486-024-00182-9 · Cited by 15
“Lastly, the pretext task of predictive approaches involves the prediction of some properties of data samples.”
Self-supervised Dense Representation Learning for Live-Cell Microscopy with Time Arrow Prediction
Gallusser, B., Stieber, M., & WEIGERT, M. G. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-43993-3_52 · Cited by 15
PixelDINO: Semi-Supervised Semantic Segmentation for Detecting Permafrost Disturbances in the Arctic
Heidler, K., Nitze, I., Grosse, G., & Zhu, X. X. (2024) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2024.3448294 · Cited by 15
“Combining these two paradigms, semi-supervised learning trains models on both labeled and unlabeled data at the same time [17], [18].”
BackgroundCollaborative Self-Supervised Evolution for Few-Shot Remote Sensing Scene Classification
Liu, Y., Li, J., Gong, M., Liu, H., Sheng, K., Zhang, Y., Tang, Z., & Zhou, Y. (2024) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2024.3434418 · Cited by 15
“Since the ability of self-supervised learning that utilizes a large amount of unlabeled data can alleviate the difficulty of label acquisition, it is widely used in various fields [56], [57], [58].”
BackgroundLiu, S., Yu, T., Zhou, J. W., Xing, G., & Chen, H. (2025) · Information Fusion · doi:10.1016/j.inffus.2025.103225 · Cited by 15
ssVERDICT: Self‐supervised VERDICT‐MRI for enhanced prostate tumor characterization
Sen, S., Singh, S., Pye, H., Moore, C. M., Whitaker, H. C., Punwani, S., Atkinson, D., Panagiotaki, E., & Slator, P. J. (2024) · Magnetic Resonance in Medicine · doi:10.1002/mrm.30186 · Cited by 15
“These methods require the neural network to predict one part of the input data given another part, 14 removing the requirement for explicitly labeled training data and thus the risk of introducing bias to the estimates.”
First Complexity Results for Evolutionary Knowledge Transfer
Scott, E. O., & Jong, K. A. D. (2023) · Foundations of Genetic Algorithms · doi:10.1145/3594805.3607137 · Cited by 15
“The idea that reusing and co-opting past solutions is a powerful way to solve new and difficult problems has a long history in artificial intelligence [11, 18, 39, 44, 57, 71, 74, 79], evolutionary biology [24, 25, 36, 58], and the study of technological innovation [3, 4].”
BackgroundInfMasking: Unleashing Synergistic Information by Contrastive Multimodal Interactions
Wen, L., Dai, Q., Liu, J., Zheng, J., Dai, Y., Wang, D., Kang, Z., Wang, J., Xu, Z., & Duan, J. (2025) · neural information processing systems · doi:10.52202/085713-0525 · Cited by 15
Computational protein design with data‐driven approaches: Recent developments and perspectives
Liu, H., & Chen, Q. (2022) · Wiley Interdisciplinary Reviews Computational Molecular Science · doi:10.1002/wcms.1646 · Cited by 14
CoRe: Contrastive and Restorative Self-Supervised Learning for Surface Defect Inspection
Wu, H., Li, B., Tian, L., Sun, Z., Dong, C., & Liao, W. (2023) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2023.3291776 · Cited by 14
“However, no large-scale defect image dataset is available for SSL in defect inspection tasks, thus not guaranteeing the generalization and quality of learned representation [17], [18].”
BackgroundAn Effective Masked Transformer Model for Automatic Modulation Recognition
Kong, W., Jiao, X., Xu, Y., & Yang, Q. (2025) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2025.3550729 · Cited by 14
“The approach of pre-training the model with unlabeled data and then fine-tuning the model with a small amount of labeled data on a downstream task has emerged as a feasible solution [11].”
Dai, Q., Zhao, C., & Huang, B. (2024) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2024.3477968 · Cited by 14
“Self-supervised learning leverages data itself to create pseu-dolabels, forming pretext tasks to predict any parts of the observed inputs, which has achieved significant success in representation learning in the fields of computer vision and natural language processing [31].”
Kang, S., Kim, S.-W., & Seong, J. (2024) · NeuroImage · doi:10.1016/j.neuroimage.2024.120737 · Cited by 14
Self-supervised Representation Learning for Speech Processing
Lee, H.-Y., Mohamed, A., Watanabe, S., Sainath, T. N., Livescu, K., Li, S.-W. D., Yang, S.-W., & Kirchhoff, K. (2022) · doi:10.18653/v1/2022.naacl-tutorials.2 · Cited by 14
“We also encourage the audience to read the papers of some well-known SSL techniques before the tutorial, which are listed below: (Ericsson et al., 2021; Rogers et al., 2020; Liu et al., 2021c; Qiu et al., 2020).”
BackgroundFEET: A Framework for Evaluating Embedding TechniquesPreprint
Lee, S. A., Lee, J. K., & Chiang, J. N. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2411.01322 · Cited by 14
“These models are trained on vast datasets using self-supervised methods [Ericsson et al., 2022], with little to no explicit supervision, enabling them to capture a wide range of knowledge.”
MethodA Survey of Pathology Foundation Model: Progress and Future Directions
Xiong, C., Chen, H., & Sung, J. J. �. (2025) · International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2025/1193 · Cited by 14
Spectral-Spatial Self-Supervised Learning for Few-Shot Hyperspectral Image ClassificationPreprint
Chen, W., Zhang, Y., Xiao, Z., Chu, J., & Wang, X. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2505.12482 · Cited by 14
“SSL can learn powerful feature representations without labeled data [29].”
Self-supervised machine learning approach for autism detection in young children using MEG signals
Barik, K., Dey, S., Watanabe, K., Hirosawa, T., Yoshimura, Y., Kikuchi, M., Bhattacharya, J., & Saha, G. (2024) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2024.106671 · Cited by 13
“This process is similar to learning one’s mother tongue at a young age before formal supervised education [35].”
InfluentialSelf-supervised learning for efficient seismic facies classification
Chikhaoui, K., & Alfarraj, M. (2024) · Geophysics · doi:10.1190/geo2023-0508.1 · Cited by 13
Yang, B., & Li, X. (2024) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2024.3458811 · Cited by 13
“B. Audio Self-Supervised Representation Learning Self-supervised representation learning [27] has been successfully applied to audio/speech processing in recent years [28], which has shown effectiveness in a wide range of downstream applications like automatic speech recognition and sound event…”
MethodTorrents, J., Costa, T., & de Polavieja, G. D. (2026) · eLife · doi:10.7554/elife.107602 · Cited by 7
Stuardi, C., Gheller, C., Vazza, F., & Botteon, A. (2024) · Monthly Notices of the Royal Astronomical Society · doi:10.1093/mnras/stae2014 · Cited by 13
End-to-End and Self-Supervised Learning for ComParE 2022 Stuttering Sub-Challenge
Sheikh, S. A., Sahidullah, M., Ouni, S., & Hirsch, F. (2022) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3503161.3551588 · Cited by 13
“In SSL classification systems, the model is first pre-trained on some pre-auxiliary task to capture rich embeddings from the innate structure of the data [4, 8, 16, 23].”
MethodBeyond Shannon: Semantic Information Theory and Methodology
Zhang, P., Niu, K., Liang, Z., Wang, C., Wu, J., Liu, Y., Xu, W., Ma, N., Xu, X., & Zhang, R. (2026) · IEEE Transactions on Network Science and Engineering · doi:10.1109/tnse.2026.3676901 · Cited by 13
“In particular, deep autoencoders and related representation-learning architectures [80], [81], [82], [83] learn compact latent representations that support semantic extraction and compression.”
Egbuna, I. K., Otoibhili, P. A., Abdulkareem, R. O., Ogbozor, F. I., Oyegue, P. E., Azih, N. K., & Akomolafe, O. B. (2025) · International Journal of Biological and Pharmaceutical Sciences Archive · doi:10.53771/ijbpsa.2025.10.1.0052 · Cited by 13
TripletCLIP: Improving Compositional Reasoning of CLIP via Synthetic Vision-Language NegativesPreprint
Patel, M., Kusumba, A., Cheng, S., Kim, C., Gokhale, T., Baral, C., & Yang, Y. (2024) · arXiv (Cornell University) · doi:10.13016/m2ksck-meo0 · Cited by 13
A Comparison Of Deep Learning MOS Predictors For Speech Synthesis Quality
Ragano, A., Benetos, E., Chinen, M., Martinez, H. B., Reddy, C. K. A., Skoglund, J., & Hines, A. (2023) · Irish Signals and Systems Conference · doi:10.1109/issc59246.2023.10162088 · Cited by 13
“To mitigate the scarcity of annotated corpora, new speech synthesis MOS prediction techniques based on self-supervised learning (SSL) [7] emerged.”
MethodGao, M., Wei, Y., Xie, Y., & Zhang, Y. (2024) · Mathematics · doi:10.3390/math12091290 · Cited by 13
“Self-supervised learning (SSL) is a technique used in representation learning [24], thereby allowing models to discover feature detection or classification representations in raw data automatically.”
MethodRobust and explainable framework to address data scarcity in diagnostic imaging
Zhao, Z., Alzubaidi, L., Zhang, J., Duan, Y., Naseem, U., & Gu, Y. (2025) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2025.111052 · Cited by 12
Self-supervised contrastive learning achieves 12-lead ECG classification
Ma, K., Zhang, T., Zhang, H., & Huang, W. X. (2025) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2025.108420 · Cited by 12
A lightweight scheme of deep appearance extraction for robust online multi-object tracking
Li, Y., Liu, Y., Zhou, C., Xu, D., & Tao, W. (2023) · The Visual Computer · doi:10.1007/s00371-023-02901-2 · Cited by 12
“Combining representation [21] and metric learning [22], multi-ID loss [23] and circle loss [24]areusedtodesignjointlossfunctionsforit.”
Difference-Masking: Choosing What to Mask in Continued Pretraining
Wilf, A., Akter, S., Mathur, L., Liang, P. P., Mathew, S., Shou, M., Nyberg, E., & Morency, L. (2023) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2023.findings-emnlp.881 · Cited by 12
Kim, T., Choi, J., Kwang-il, K., & Ko, M. (2025) · JMIR Medical Informatics · doi:10.2196/57298 · Cited by 12
“To address the risk of overfitting due to the limited data, we used a self-supervised learning (SSL) strategy [31,32].”
Multimodal Learning for Embryo Viability Prediction in Clinical IVF
Kim, J., Shi, Z., Jeong, D., Knittel, J., Yang, H. Y., Song, Y., Li, W., Li, Y., Ben‐Yosef, D., Needleman, D. J., et al. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-72086-4_51 · Cited by 13
“One solution is to pre-train modality-specific encoders separately with pretext tasks using self-supervised learning [9] and then fine-tune the encoders with multimodal transformers by supervised learning for the downstream task.”
MethodXiong, Y., He, T., Zhu, W., Liao, Y., Xu, Q., Niu, Y., & Chen, Z. (2024) · Sustainable Energy Technologies and Assessments · doi:10.1016/j.seta.2024.103989 · Cited by 12
Morfessor-enriched features and multilingual training for canonical morphological segmentation
Rouhe, A., Grönroos, S.-A., Virpioja, S., Creutz, M., & Kurimo, M. (2022) · SIGMORPHON Workshop on Computational Research in Phonetics, Phonology, and Morphology (SIGMORPHON) · doi:10.18653/v1/2022.sigmorphon-1.16 · Cited by 12
Knowledge Accumulation in Continually Learned Representations and the Issue of Feature Forgetting
Hess, T., Verwimp, E., van de Ven, G. M., & Tuytelaars, T. (2024) · Transactions on machine learning research · doi:10.48550/arxiv.2304.00933 · Cited by 12
“The recent success (Ericsson et al., 2022) of these learning mechanisms sparked interest for their application in continual learning as well (Fini et al., 2022; Cha et al., 2021; Marsocci & Scardapane, 2023; Wang et al., 2023), and they have been found to be less prone to forgetting Wang et al.…”
BackgroundOurang, S. A., Sohrabniya, F., Sadr, S., Lee, J., Ramzy, N., Law, A., Lam, E. W. N., Khan, A. A., & Nosrat, A. (2025) · International Endodontic Journal · doi:10.1111/iej.70065 · Cited by 11
Self-supervised learning from noisy and incomplete data
Tachella, J. A., & Davies, M. E. (2026) · Foundations and Trends® in Signal Processing · doi:10.1108/ftsig-10-2025-0133 · Cited by 11
Can Self-Supervised Representation Learning Methods Withstand Distribution Shifts and Corruptions?
Chhipa, P. C., Holmgren, J. R., De, K., Saini, R., & Liwicki, M. (2023) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw60793.2023.00481 · Cited by 11
“In recent years self-supervised representation learning (SSL) methods [18] have garnered interest in computer vision applications.”
BackgroundVILLS: Video-Image Learning to Learn Semantics for Person Re-Identification
Huang, S., Kathirvel, R. P., Guo, Y., Chellappa, R., & Peng, C. (2025) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv61041.2025.00582 · Cited by 11
“Recent progress in self-supervised learning have further improved image-based ReID methods by allowing models to learn rich, transferable representations [13, 19, 57, 67].”
Foundation Models for Environmental Science: A Survey of Emerging FrontiersPreprint
Yu, R., Chen, S., Xie, Y., Yao, H., Willard, J., & Jia, X. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2504.04280 · Cited by 11
Lead-fusion Barlow twins: A fused self-supervised learning method for multi-lead electrocardiograms
Liu, W., Pan, S., Li, Z., Chang, S., Huang, Q., & Jiang, N. (2024) · Information Fusion · doi:10.1016/j.inffus.2024.102698 · Cited by 11
Subhan, R. M., Lee, Y.-D., & Koo, I. (2026) · Applied Sciences · doi:10.3390/app16031448 · Cited by 11
SA-WavLM: Speaker-Aware Self-Supervised Pre-training for Mixture Speech
Lin, J., Ge, M., Ao, J., Deng, L., & Li, H. (2024) · Interspeech (USB) · doi:10.21437/interspeech.2024-787 · Cited by 11
“Self-supervised learning (SSL) based pre-training have greatly advanced over the past few years [1, 2, 3].”
BackgroundToward Fully Self-Supervised Multi-Pitch EstimationPreprint
Cwitkowitz, F., & Duan, Z. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2402.15569 · Cited by 11
“This more closely resembles the self-supervised learning (SSL) paradigm in machine learning, which has been garnering increasing attention due to its profound success in learning rich representations for complex data spanning various modalities (Ericsson et al., 2022).”
BackgroundAbbas, T., Linjawi, M., Iqbal, I. A., Alghushairy, O., Alsini, R. A. I., & Daud, A. Y. (2025) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2025.108391 · Cited by 10
SeisBERT: A pretrained seismic image representation model for seismic data interpretation
Pham, N. H., Di, H., Zhao, T., & Abubakar, A. (2025) · The Leading Edge · doi:10.1190/tle44020096.1 · Cited by 10
Bai, L., Zhang, X., Qin, W., Long, J., Wang, H., Dong, X., & Du, S. (2025) · IEEE Geoscience and Remote Sensing Magazine · doi:10.1109/mgrs.2025.3588505 · Cited by 10
Analyzing Generalization in Policy Networks: A Case Study with the Double-Integrator System
Zhang, R., Han, H., Lv, M., Yang, Q., & Cheng, J. P. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i15.29623 · Cited by 10
Ha, S. V., Jaensch, S., Kańduła, M. M., Herman, D., Czodrowski, P., & Ceulemans, H. (2025) · Scientific Reports · doi:10.1038/s41598-025-05914-0 · Cited by 10
Aligning Speech-Text Representations via Contrastive Modality Translation
Chien, J., & Liu, P.-Y. (2025) · IEEE Transactions on Audio Speech and Language Processing · doi:10.1109/taslpro.2025.3603910 · Cited by 10
“These representations capture a wide range of acoustic and linguistic characteristics including prosody, speaker identity, and semantic content, often through the processes such as vector quantization or latent discretization [4].”
Metadata
- Authors
- Ericsson, L., Gouk, H., Loy, C. C., & Hospedales, T. M.
- Affiliations
- University of Edinburgh (GB)
- Nanyang Technological University (SG)
- Venue
- IEEE Signal Processing Magazine
- Published
- 2022-05
- Volume
- 39
- Issue
- 3
- Pages
- 42-62
- ISSN
- 1053-5888 (ISSN-L)
- 1558-0792
- OpenAlex
- W3207924272
- Cited by
- 433
- References
- 125
- Citations by year
- 2026: 89 · 2025: 105 · 2024: 132 · 2023: 86 · 2022: 20
- FWCI
- 28.22
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Signal Processing Magazine (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- Edinburgh Research Explorer (University of Edinburgh) (submittedVersion, cc-by-nd)
- DR-NTU (Nanyang Technological University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- Engineering and Physical Sciences Research Council (EP/S000631/1, EP/L016427/1, EP/R026173/1)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Physical Sciences › Computer Science › Artificial Intelligence › Advanced Graph Neural Networks
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning and Algorithms
- Keywords
- self-supervised representation learning, pre-training, deep feature learning, image representation, video representation, sound representation, text representation, graph representation, computational cost, open problems
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.