A Survey on Deep Learning for Multimodal Data Fusion
The DOI matches its Crossref record; 100 of the 100 references it cites were found in a registry or an index.
Abstract
With the wide deployments of heterogeneous networks, huge amounts of data with characteristics of high volume, high variety, high velocity, and high veracity are generated. These data, referred to multimodal big data, contain abundant intermodality and cross-modality information and pose vast challenges on traditional data fusion methods. In this review, we present some pioneering deep learning models to fuse these multimodal big data. With the increasing exploration of the multimodal big data, there are still some challenges to be addressed. Thus, this review presents a survey on deep learning for multimodal data fusion to provide readers, regardless of their original community, with the fundamentals of multimodal deep learning fusion method and to motivate new multimodal data fusion techniques of deep learning. Specifically, representative architectures that are widely used are summarized as fundamental to the understanding of multimodal deep learning. Then the current pioneering multimodal data fusion deep learning models are summarized. Finally, some challenges and future topics of multimodal data fusion deep learning models are described.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
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?
In the citing paper
“For more generally applicable models, several deep learning frameworks have been extended and enable prediction using multimodal data (Gao et al. 2020).”
From Similarity-based multimodal regression (2024) · cites this paper as Gao et al. 2020
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 1×
“ResNet (He et al., 2016) A shortcut connection is used to cross several layers to back propagate the network loss to previous layers.”
MethodImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 6×
“The image subnetwork is a representative deep convolutional neural network, such as Alexnet and Inception, which effectively encodes the image input into a concise representation.”
BackgroundInfluentialHochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318 · cited 3×
“LSTM (Hochreiter & Schmidhuber, 1997) The memory block is introduced to model the long-time dependency well.”
BackgroundInfluentialLeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126
Gradient-based learning applied to document recognition
Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998) · Proceedings of the IEEE · doi:10.1109/5.726791 · Cited by 58,805
MobileNetV2: Inverted Residuals and Linear Bottlenecks
Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L.-C. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00474 · Cited by 28,755
Reducing the dimensionality of data with neural networks
Hinton, G. E., & Salakhutdinov, R. R. (2006) · Science · doi:10.1126/science.1127647 · Cited by 20,928
GAN(Generative Adversarial Nets)
(2017) · Journal of Japan Society for Fuzzy Theory and Intelligent Informatics · doi:10.3156/jsoft.29.5_177_2 · Cited by 19,270
A fast learning algorithm for deep belief nets
Hinton, G. E., Osindero, S., & Teh, Y.-W. (2006) · Neural Computation · doi:10.1162/neco.2006.18.7.1527 · Cited by 16,336 · cited 1×
“Then the gradients of the weight, the visible bias, and the hidden bias are computed in the following forms: Unfortunately, in those gradient-computing equations, the probability x P ( x ) P ( h i = 1 | x ) is difficult to compute (Hinton, Osindero, & Teh, 2006).”
MethodVisualizing and understanding convolutional networks
Zeiler, M. D., & Fergus, R. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-10590-1_53 · Cited by 15,839 · cited 1×
“Similar to the fully connected architecture, the CNN is also trained to fit the training data, using the same algorithm (LeCun et al., 1989; LeCun, Bot-tou, Bengio, & Haffner, 1998; Zeiler & Fergus, 2014).”
Method
Research works printed but not linked: 1
- Goodfellow, I. J., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., … Bengio, Y. (2014). Generative adversarial nets. In Z.Ghahramani, M.Welling, C.Cortes, N. D.Lawrence, & K. Q.Weinberger (Eds.), Advances in neural information processing systems, 27 (pp. 2672–2680). Red Hook, NY: Curran.
Cited by
Items 801–837 of 837
Scalable and fair deep neural networks for imbalanced data using a multilevel graph-based approach
Bennett, R., & Razzaghi, T. (2026) · Expert Systems with Applications · doi:10.1016/j.eswa.2026.134330
Nguyen, B.-D., Hoang, V.-D., & Nguyen, H. D. (2026) · Network Modeling Analysis in Health Informatics and Bioinformatics · doi:10.1007/s13721-026-00871-8
“…several surveys have reviewed multimodal learning in general (Baltrušaitis et al. 2018 ) and biomedical multimodal AI more broadly (Acosta et al. 2022 ; Gao et al. 2020 ; Kline et al. 2022 ), the specific role of gating and MoE in biomedi - cal multimodal fusion remains insufficiently synthesized.”
He, M., Wang, H., Wu, D., Wang, Y., & Fu, X. (2026) · Optics & Laser Technology · doi:10.1016/j.optlastec.2026.116372
Wang, J. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7465226
Wan, X., Yang, D., Hu, H., Liao, J., Wang, Y., Deng, P., Tao, Q., & Wu, Z. (2026) · Microchemical Journal · doi:10.1016/j.microc.2026.119721
Zhang, K., Wei, Z., Chen, M., Qiu, Z., & Lang, L. (2026) · Structural Health Monitoring · doi:10.1177/14759217261483593
Fuzzy Multi-View Learning: Principles, Methodologies, and Future Directions
Li, A.-D., Liu, H.-B., Hu, F.-P., & Zhang, W. (2026) · Symmetry · doi:10.3390/sym18101587
A multimodal deep learning framework for physics-guided anomaly detection and time-series prediction
Pessoa, R. W. S., Amblard, J., Russo, A., & Nogueira, I. R. (2026) · Chemical Engineering Science · doi:10.1016/j.ces.2026.125216
Cui, J., Mou, Y., Xiao, H., Qin, Y., Chen, D., & Yu, H. (2026) · Translational Psychiatry · doi:10.1038/s41398-026-04486-5
meng, X., & Lin, J. (2026) · Pattern Recognition · doi:10.1016/j.patcog.2026.114998
Ensemble-Based Machine Learning for Drug-Induced Autoimmunity Prediction Using Molecular Descriptors
Fouda, M. I., Sabor, A., & Mohammed, A. (2026) · Intelligent Methods, Systems, and Applications (IMSA) · doi:10.1109/imsa70415.2026.11700376
“In many practical scenarios, different modalities[25] provide complementary and correlated perspectives of the same underlying concept.”
Balanced Speech Emotion Recognition for Edge and Cloud: A Practical Accuracy-Efficiency Tradeoff
Ahmed, Y., Elsherif, A., Darwish, F., & Mohammed, A. (2026) · Intelligent Methods, Systems, and Applications (IMSA) · doi:10.1109/imsa70415.2026.11700355
“Integrating multiple modalities [22] enables the model to leverage complementary and mutually informative representations of the same underlying concept, leading to more robust and accurate performance.”
Muntashir, A. W., Aufaristama, M., & Dharmawan, I. A. (2026) · IEEE Access · doi:10.1109/access.2026.3738150
“Multi-modal learning is utilized when data must be integrated from diverse sources [50], and attention mechanisms are used to identify prominent areas in complex scenes [48].”
Hadi, A. A., & Supreethi, K. P. (2026) · Discover Informatics · doi:10.1007/s44564-026-00022-1
Применение нейросетевых методов с алгоритмической предобработкой для интеграции гетерогенных данных
Соколов, А. А., & Саркисова, И. О. (2026) · Вестник Воронежского государственного технического университета · doi:10.36622/1729-6501.2026.22.3.007
Feng, K., Tang, C., Ma, W., & Yankai, S. U. N. (2026) · Neurocomputing · doi:10.1016/j.neucom.2026.135296
Rangel, P., Moosavi, S. R., Jeter, A., Khatibi, E., Wang, Z., Firouzi, F., Chakrabarty, K., & Rahmani, A. M. (2026) · IEEE International Conference on Omni-layer Intelligent Systems (COINS) · doi:10.1109/coins69995.2026.11715579
“Beyond ECG-only learning, multimodal physiological modeling has gained increasing attention because different biosignals capture complementary aspects of cardiovascular dynamics [11, 12].”
Multimodal semantic communications
Qin, Z., Ying, J., Tao, X., & Tong, W. (2026) · Nature Reviews Electrical Engineering · doi:10.1038/s44287-026-00336-0
Chen, H. (2026) · International Conference on Biomedical Engineering and Bioinformatics for Health (ICBEBH) · doi:10.1109/icbebh71590.2026.11722448
“Early fusion combines raw features before training — simple and enabling direct cross-modal interaction — but is sensitive to scale, noise, and alignment differences across modalities [9][10].”
Nature Mental Health Earth–Brain–Health Commission
Schumann, G., Luterbacher, J., Marquand, A. F., Groß, C., Hunt, X., Stahl, B. C., Brandt, L., Чэн, Б., Damodharan, D., Ji, J. S., et al. (2026) · Nature Mental Health · doi:10.1038/s44220-026-00737-y
Intelligent Visibility Estimation Under Sea Fog Conditions Based on Deep Learning
-, Z. Q., Hu, W., Ren, P., Tian, H., Ruan, C., & Song, Y. (2026) · Journal of Marine Science and Engineering · doi:10.3390/jmse14201888
Martinis, M. C., Popello, V., Lazzaro, I., Zucco, C., & Cannataro, M. (2026) · AI · doi:10.3390/ai7100424
Yao, S., yao, Xu, L., Le, Y., Liao, D., Wu, J., & Wan, S. (2026) · Water Resources Research · doi:10.1029/2025wr040618
“Landsat data, with higher spatial resolution relative to coarser remote sensing data sets, excels at capturing fine‐scale wetland features, such as subtle landscape variations and clear water‐land boundaries (Li et al., 2020; Wulder et al., 2022).”
张, 智. (2026) · Hans Journal of Data Mining · doi:10.12677/hjdm.2026.164013
Deep Learning Methods for Video Classification and Tagging
Tuyet, V. T. H., & Bình, N. T. (2026) · Advances in computational intelligence and robotics book series · doi:10.4018/979-8-2600-1393-9.ch004
Hwang, Y., Ryu, G., Lee, M., So, Y., Yu, J., & Mun, B. (2026) · Plants · doi:10.3390/plants15203171
Wang, X., Yang, F., Liu, L., Huang, Y., Dong, W., & Yao, D. (2026) · Sensors and Actuators A Physical · doi:10.1016/j.sna.2026.118676
Human-Centric Multimodal Fusion for Real-Time Driving Hazard Detection
Ma, T., Feng, A., Xiong, X., Gao, Z., Long, Z., Petrosian, O., & Xu, X. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37679-4_8
Guan, Y., Li, S., Fan, J., Cui, Y., Zhang, D., Yang, L., & Kang, Y. (2026) · Journal of Composites Science · doi:10.3390/jcs10100559
Shu, B., Liu, Y., Shu, G., & Zhang, H. (2026) · Energy and Buildings · doi:10.1016/j.enbuild.2026.118392
Zheng, C., Wang, J., Yan, Y., Zhang, L., Dong, W., Sun, X., Sun, B., & Wang, S. (2026) · Trends in Food Science & Technology · doi:10.1016/j.tifs.2026.106194
Reji, A. A. A., Anandaraj, M. D., Jose, G. J. A., Muneshwara, M. S., & Anand, R. (2026) · 2026 Conference on Latest Innovations in Computing and Knowledge (CLICK) · doi:10.1109/click68749.2026.11741754
“The fused vector is passed through dense layers with BatchNorm and ReLU to produce the final classification output: This latefusion, attentionbased strategy dynamically weights modality contributions, capturing inter-modal dependencies across visual, auditory, and temporal dimensions [17].”
Szewczyk, D., & Lech, W. (2022) · Data Engineering and Applications · doi:10.64972/dea.2022.v1i3.4336d:65-79
Wind Turbine Digital Twin Virtual Sensing Construction Based on a Koopman Autoencoder Model
Al-Buflasa, J., Al-Harmoodi, E., & Al-Riyami, K. (2023) · Data Engineering and Applications · doi:10.64972/dea.2023.v2i2.3531d:1-14
Pokorný, O., & Růžička, V. (2022) · Data Engineering and Applications · doi:10.64972/dea.2022.v1i2.4238d:108-120
ІНТЕЛЕКТУАЛЬНІ СИСТЕМИ ТА ЦИФРОВІ ТЕХНОЛОГІЇ
Kashtan, V., Dr.Sci., V. H., Dr.Sci., V. H., Ph.D., A. V., Dr.Sci., L. K., Dr.Sci., S. P., Nelasa, H., Mikhalev, K., Ph.D., А. N., Ph.D., О. І., et al. (2026) · doi:10.36059/978-966-397-663-1
Wu, J., Nie, J., Yang, X., Rogers, K. M., Xu, X., Li, C., Feng, P., Luo, Z., Mei, H., Wang, P., et al. (2026) · Food Chemistry · doi:10.1016/j.foodchem.2026.151987
Metadata
- Authors
- Gao, J., Li, P., Chen, Z., & Zhang, J.
- Affiliations
- Dalian University of Technology (CN)
- Corresponding author
- Gao, Jing
- Li, Peng
- Chen, Zhikui
- Zhang, Jianing
- Venue
- Neural Computation
- Publisher
- MIT Press
- Published
- 2020-05-01
- Volume
- 32
- Issue
- 5
- Pages
- 829-864
- ISSN
- 0899-7667 (ISSN-L)
- 1530-888X
- OpenAlex
- W3011727199
- Cited by
- 821
- References
- 99
- Citations by year
- 2026: 152 · 2025: 213 · 2024: 161 · 2023: 129 · 2022: 90 · 2021: 57 · 2020: 19
- FWCI
- 5.99
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- Neural Computation (publishedVersion)
- PubMed (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Multimodal Machine Learning Applications
- Physical Sciences › Engineering › Media Technology › Advanced Image Fusion Techniques
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Keywords
- multimodal fusion, multimodal deep learning, heterogeneous networks, deep learning architectures, fusion model
- MeSH terms
- Deep Learning, Algorithms, Artificial Intelligence, Surveys and Questionnaires, Neural Networks, Computer
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.