An Introduction to Deep Learning for the Physical Layer
Open accessThe DOI matches its Crossref record; 90 of the 92 references it cites were found in a registry or an index, and 2 are not checked yet; 8 of the 32 sources its claims cite were checked for support.
Abstract
We present and discuss several novel applications of deep learning for the physical layer. By interpreting a communications system as an autoencoder, we develop a fundamental new way to think about communications system design as an end-to-end reconstruction task that seeks to jointly optimize transmitter and receiver components in a single process. We show how this idea can be extended to networks of multiple transmitters and receivers and present the concept of radio transformer networks as a means to incorporate expert domain knowledge in the machine learning model. Lastly, we demonstrate the application of convolutional neural networks on raw IQ samples for modulation classification which achieves competitive accuracy with respect to traditional schemes relying on expert features. This paper is concluded with a discussion of open challenges and areas for future investigation.
Verified 2026-11-19 · 2 references not checked yet
Overview
CoChat Summary
TL;DR
The paper introduces deep learning for the physical layer by treating a communications system as an autoencoder that jointly learns transmitter and receiver implementations end-to-end. It also extends this idea to multi-user interference channels, proposes radio transformer networks (RTNs) to inject expert domain knowledge, and shows CNNs on raw IQ samples can outperform expert-feature modulation classifiers.
Key findings
- A learned autoencoder for a (7,4) system at rate R = 4/7 achieves the same block error rate (BLER) as a Hamming (7,4) code with maximum likelihood decoding, while an (8,8) autoencoder outperforms uncoded BPSK over the full E_b/N_0 range, indicating learned joint coding and modulation.
- For the two-user AWGN interference channel, the autoencoder matches time-sharing for (1,1) and (2,2), but achieves gains of about 0.7 dB for (4,4) and 1 dB for (4,8) at a BLER of 10^-3.
- An autoencoder (8,4) with an RTN over a multipath fading channel with L = 3 Rayleigh taps outperforms differential BPSK with maximum likelihood sequence estimation and a Hamming (7,4) code, whereas the plain autoencoder struggles to match it; RTNs also show faster training convergence.
- For modulation classification on 1.2M sequences of 128 complex-valued IQ samples across ten modulation schemes and SNRs from -20 dB to 18 dB, the CNN outperforms a boosted expert-feature classifier by around 4 dB in the low-to-medium SNR range, with almost identical high-SNR performance; a single tree is about 6 dB worse at medium SNR and 3.5% worse at high SNR.
Methods
- The transmitter, channel, and receiver are represented as one deep neural network trained as an autoencoder: one-hot message input, dense layers, normalization layer for transmit constraints, additive noise channel layer, and softmax receiver output, trained with categorical cross-entropy using SGD/Adam at learning rate 0.001 and fixed E_b/N_0 = 7 dB.
- The autoencoder concept is extended to multiple transmitter-receiver pairs in a two-user interference channel, with dynamic loss weights α_t per mini-batch to balance the two cross-entropy losses; performance is compared against 2^{2k/n}-QAM with time-sharing.
- RTNs combine a learned parameter estimator, a differentiable parametric transform, and a discriminative network; the paper demonstrates phase-offset compensation and multipath inverse filtering. For modulation classification, a VGG-like CNN with 324,330 trainable parameters is trained on the RML2016.10b dataset.
Limitations
- The paper states that scalability to large message sets M is a major challenge: for k = 100 bits, training complexity is prohibitive because the autoencoder must see at least every message once.
- The paper states that optimal data representations, loss functions, training SNR, and hyperparameters are not well understood; training at one SNR does not necessarily generalize to others, and no common communications benchmarks or datasets comparable to MNIST or ImageNet exist.
- The paper states that end-to-end learning over real channels requires system identification because backpropagation assumes a known channel transfer function; complex-valued neural networks are not supported by current DL libraries. Evident from the text: the reported results are simulation-based, use small (n,k) settings, assume known channel models during training, and the paper explicitly leaves detailed performance comparisons for various channel types and parameters to future work.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures











Verification
Claims the paper makes
6 Valid · 4 Not valid · 18 Not checked · 65 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“For instance in [22], it was shown that training with a single codeword is sufficient since the structure of the code is embedded in the NN through the Tanner graph.”
What the sources say
- Supportedcites [22] [22] E. Nachmani, Y. Be'ery, and D. Burshtein, 'Learning to decode linear codes using deep learning,' in Proc. IEEE Annu. Allerton Conf. Commun., Control, and Computing (Allerton) , 2016, pp. 341-346. - source
The Introduction says the decoder uses a Tanner graph representing the code and that one codeword suffices for training. Section III explains the precise condition: over a binary memoryless symmetric channel, the decoder’s error rate is independent of the transmitted codeword, so training can use noisy versions of a single codeword. (LLM, conf=0.91)
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“We study the use of NNs on complex-valued IQ samples for the problem of modulation classification and show that convolutional neural networks (CNNs), which are the cornerstone of most DL systems for computer vision, can outperform traditional classification techniques based on expert features.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“Training at low SNR for instance does not allow for the discovery of structure important in higher SNR scenarios.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: unclear — It is unclear why the paper says it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Another advantage of RTNs is faster training convergence which can be seen from Fig. 10 that compares the validation loss of the autoencoder with and without RTN as a function of the training epochs.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Similarly, Fig. 4b shows a (4 , 2) system which leads to a rotated 16-PSK constellation.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
39 more claims
Create a free account to see all 93 claims and to verify the 4 references still open.
- Probably not a substantial claim. 29
- It may state two separate claims as one. 24
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ A. Autoencoders for end-to-end communications systems · p. 5
“Similarly, Fig. 4b shows a (4 , 2) system which leads to a rotated 16-PSK constellation.”
§ A. Autoencoders for end-to-end communications systems · p. 6
“Figure 4: Constellations produced by autoencoders using parameters ( n, k ) : (a) (2 , 2) (b) (2 , 4) , (c) (2 , 4) with average power constraint, (d) (7 , 4) 2-dimensional t-SNE embedding of received symbols.”
Citations
Cites
Very Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 1×
“dense/fully-connected layers and terminated with a dense softmax layer for our classifier (similar to a VGG architecture [55]).”
BackgroundAdam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946 · cited 4×
“Training was done at a fixed value of Eb/N0 = 7 dB (see Section IV-B) using Adam [46] with learning rate 0.”
BackgroundMethodInfluentialA Mathematical Theory of Communication
Shannon, C. E. (1948) · Bell System Technical Journal · doi:10.1002/j.1538-7305.1948.tb01338.x · Cited by 88,486
Scikit-learn: Machine Learning in Python
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Müller, A., Nothman, J., Louppe, G., et al. (2012) · arXiv · doi:10.48550/arxiv.1201.0490 · Cited by 71,016 · cited 1×
“11, we compare the classification accuracy of the CNN against that of extreme gradient boosting7 with 1000 estimators, as well as a single scikit-learn decision tree [56], operating on a mix of 16 analog and cumulant expert features as proposed in [53] and [57].”
Method(2014) · Cited by 33,419
(2008) · Journal of Machine Learning Research · Cited by 33,051
Human-level control through deep reinforcement learning
Mnih, V., Kavukcuoglu, K., Silver, D., Rusu, A. A., Veness, J., Bellemare, M. G., Graves, A., Riedmiller, M., Fidjeland, A. K., Ostrovski, G., et al. (2015) · Nature · doi:10.1038/nature14236 · Cited by 30,357
(2007) · Electronics: A First Course · doi:10.4324/9780080477527-64 · Cited by 24,923
Pan, S. J., & Yang, Q. (2010) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2009.191 · Cited by 23,387 · cited 1×
“learning [64] is one appealing candidate for adapting an endto-end communications system trained on a statistical model to”
BackgroundMultilayer feedforward networks are universal approximators
Hornik, K., Stinchcombe, M., & White, H. (1989) · Neural Networks · doi:10.1016/0893-6080(89)90020-8 · Cited by 21,728 · cited 1×
“Third, it has been shown that NNs are universal function approximators [13] and recent work has shown a remarkable”
Background
Research works printed but not linked: 2
- [35] I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning . MIT Press, 2016. -
- [52] J. Estaran et al. , 'Artificial neural networks for linear and non-linear impairment mitigation in high-baudrate IM/DD systems,' in Proc. 42nd European Conf. Optical Commun. (ECOC) . VDE, 2016, pp. 1-3. -
Cited by
Items 1,301–1,400 of 2,995
Machine Learning for Optimal Resource Allocation
Pesavento, M., & Bahlke, F. (2019) · Wiley eBooks · doi:10.1002/9781119562306.ch5 · Cited by 9
Deep Learning Interference Cancellation in Wireless NetworksPreprint
Zhou, Y., Samiee, A., Zhou, T., & Jalali, B. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2009.05533 · Cited by 14
“Although hand-crafted functional blocks and coding schemes are proven effective to guarantee reliable data transfer, currently deep learning-based approaches have drawn increasing attention in the communication system modeling [1, 2].”
BackgroundA Channel Coding Benchmark for Meta-LearningPreprint
Li, R., Bohdal, O., Mishra, R., Kim, H., Li, D., Lane, N. D., & Hospedales, T. M. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2107.07579 · Cited by 11
“Thus there has been significant interest in applying deep learning to improve channel decoding (and coding itself) [25].”
BackgroundSpectrum Challenges in the Internet of Things: State of the Art and Next Steps
Restuccia, F., Melodia, T., & Ashdown, J. (2022) · Wiley eBooks · doi:10.1002/9781119892199.ch19 · Cited by 8
Robust Generation of Channel Distributions with Diffusion Models
Kim, M., Fritschek, R., & Schaefer, R. F. (2024) · International Conference on Communications · doi:10.1109/icc51166.2024.10622251 · Cited by 8
“An end-to-end (E2E) framework was proposed resembling the autoencoder (AE) model with a wireless channel block between the encoder and decoder and integrating modulation and demodulation [1], [2].”
MethodSecuring MIMO Wiretap Channel With Learning-Based Friendly Jamming Under Imperfect CSI
Tuan, B. M., Nguyen, D. N., Trung, N. L., Nguyen, V., Huynh, N. V., Hoang, D. T., Krunz, M., & Dutkiewicz, E. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3536702 · Cited by 10
“It is worth noting that the communication and security tasks are transformed into the classification with the AE, Softmax activation function, and cross-entropy-based security loss function [32].”
Qian, Y., Bai, Y., Mei, Z., Zhang, S., Ni, Y., Shi, L. X., & Shu, F. (2025) · IEEE Transactions on Communications · doi:10.1109/tcomm.2025.3578841 · Cited by 11
“Restrictions apply. significantly threatens the covertness of CCSs [10].”
Rapid prediction of complex nonlinear dynamics in Kerr resonators using the recurrent neural network
Huang, T., Chen, L., Lü, M. K., Pan, J., Xu, C., Wang, P., & Shum, P. P. (2025) · Frontiers of Optoelectronics · doi:10.1007/s12200-025-00164-4 · Cited by 7
“Deep learning technology possesses unique advantages and is poised to address challenges posed by traditional mathematical models, thereby leading to innovative optical communication models [17, 18].”
Applying Deep Neural Networks for Duty Cycle Estimation
Al-Tahmeesschi, A., Umebayashi, K., Iwata, H., López‐Benítez, M., & Lehtomäki, J. J. (2020) · IEEE Wireless Communications and Networking Conference · doi:10.1109/wcnc45663.2020.9120720 · Cited by 8
“This is mainly due to the advancement in massively parallel GPU architecture and high-level languages [13].”
BackgroundPower information network intrusion detection based on deep learning and cloud computing
Zhao, X., & Zhou, J. (2021) · Journal of Intelligent & Fuzzy Systems · doi:10.3233/jifs-219104 · Cited by 8
End-to-End Deep Learning IRS-assisted Communications Systems
Alawad, M. A., Hamdan, M. Q., & Hamdi, K. A. (2021) · IEEE Vehicular Technology Conference · doi:10.1109/vtc2021-fall52928.2021.9625398 · Cited by 7
Training Deep Filters for Physical-Layer Frame Synchronization
Kaladè, Š., Crockett, L. H., & Stewart, R. W. (2022) · IEEE Open Journal of the Communications Society · doi:10.1109/ojcoms.2022.3185973 · Cited by 8
Deep learning-based Human Body Communication baseband transceiver for WBAN IEEE 802.15.6
Ali, A., Ahmed, S. M., Sayed, M. S., & Shalaby, A. (2022) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2022.105169 · Cited by 7
Dynamic speaker localization based on a novel lightweight R–CNN model
Çatalbaş, M. C., & Dobrišek, S. (2023) · Neural Computing and Applications · doi:10.1007/s00521-023-08251-3 · Cited by 9
Blind Detection of Broadband Signal Based on Weighted Bi-Directional Feature Pyramid Network
Guo, S., Yao, J., Wu, P., Yang, J., Wu, W., & Lin, Z. (2023) · Sensors · doi:10.3390/s23031525 · Cited by 7
Intelligent resource allocation decision using deep learning and optimization techniques for HetNets
Victoria, D. R. S., Princy Magdaline, P., & Glory Ratna Mary, D. (2023) · Wireless Networks · doi:10.1007/s11276-023-03360-2 · Cited by 7
“means to enhance performance in complicated communications settings that are challenging to model using conventional mathematical methods [6].”
BackgroundA deep learning-based adaptive receiver for full-duplex systems
Shammaa, M., Mashaly, M. E., & El-mahdy, A. (2023) · AEU - International Journal of Electronics and Communications · doi:10.1016/j.aeue.2023.154822 · Cited by 9
Sultana, R., Rana, V., & Chou, R. A. (2023) · IEEE International Symposium on Information Theory (ISIT) · doi:10.1109/isit54713.2023.10206773 · Cited by 7
“(e0, d0) is implemented with an autoencoder as in [33].”
MethodAnwar, D. N., Ahmad, R., Salameh, H. B., Elgala, H., & Ayyash, M. (2023) · Neural Computing and Applications · doi:10.1007/s00521-023-09017-7 · Cited by 7
“To address the problem of sub-optimization in the conventional block-based [6, 7] communication structure, T. O’Shea introduced channel auto-encoder (AE) as an end-to-end learning framework to jointly optimize transmitter and receiver components of communication system [8].”
MethodA Real-World Dataset Generator for Specific Emitter Identification
Muller, B. P., Wong, L. J., Clark, W. H., & Michaels, A. J. (2023) · IEEE Access · doi:10.1109/access.2023.3322105 · Cited by 8
“for signal detection, classification, and estimation [2], [3], enabling enhanced security and radio spectrum sharing”
BackgroundImage super resolution based channel estimation for future wireless communication
Ahmad, M., Shakeel, T., & Shin, S. Y. (2023) · Computer Networks · doi:10.1016/j.comnet.2023.110057 · Cited by 8
AI/ML Optimized High-Order Modulations
Madadi, P., Cho, J., Zhang, J. C., & Burghal, D. (2023) · International Conference on Communications · doi:10.1109/icc45041.2023.10279091 · Cited by 7
“Specifically, the approach employs an autoencoder based end-to-end learning [12] and performs joint optimization of the transmitter and receiver neural networks (NN) using the binary cross entropy (BCE) as the loss function.”
MethodPerformance Evaluation of AI/ML Model to Enhance Beam Management in 5G-Advanced System
Xu, J., Nakamura, I., Feng, R., Liu, L., & Chen, L. (2023) · International Conference on Mobile Computing and Ubiquitous Network (ICMU) · doi:10.23919/icmu58504.2023.10412253 · Cited by 12
“Considering the physical layer of wireless communications, AI is first introduced in [1], where the conventional module based communication process is reshaped as jointly optimization of AI based transmitter and receiver, which is regarded as an end-to-end communication.”
BackgroundKapsis, T. T., Lyras, N. K., & Panagopoulos, A. D. (2024) · Electronics · doi:10.3390/electronics13030647 · Cited by 7
Federated learning based modulation classification for multipath channels
Bhardwaj, S., Kim, D.-H., & Kim, D. (2024) · Parallel Computing · doi:10.1016/j.parco.2024.103083 · Cited by 8
Low Complexity High Speed Deep Neural Network Augmented Wireless Channel Estimation
Haq, S. A. U., Singh, V., Tanaji, B. T., & Darak, S. J. (2024) · International Conference on VLSI Design · doi:10.1109/vlsid60093.2024.00045 · Cited by 7
“With the evolution of deep learning (DL), various works have explored its usefulness for wireless PHY [1]–[9].”
BackgroundMethodMa, X., & Ding, S. (2024) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2024.3394695 · Cited by 7
“Restrictions apply. that adapt to changing conditions and estimation of channel performance or QoS in different layers [16], [17], [18], [19].”
BackgroundDigital Modulation Identification in Spatial Modulation MIMO Systems
Beknadj, D., Tayakout, H., Tounsi, M., Bouchibane, F. Z., & Boutellaa, E. (2024) · International Conference on Image and Signal Processing and their Applications (ISPA) · doi:10.1109/ispa59904.2024.10536863 · Cited by 8
“[ 43] thought that employing the CNN trained on IQ samples could facilitate the automatic features extraction and efficiently classify the modulated signals.”
MethodEnd-to-End Learning for SLP-based ISAC Systems
Zheng, Y., Liu, R., Li, M., & Liu, Q. (2024) · IEEE Wireless Communications and Networking Conference · doi:10.1109/wcnc57260.2024.10571266 · Cited by 7
“For example, the local optimization of each module does not imply the end-to-end global optimization of the whole system [5], which points out a promising breakthrough for addressing performance bottlenecks in end-to-end research.”
BackgroundGraph model-aided optimal iterative decoding technique for LDPC in optical fiber communication
Tian, Q., Pan, Y., Xin, X., Zhu, L., Li, Z., Wang, C., Dong, Z., Gao, R., Tian, F., Wang, F., et al. (2024) · Optics Express · doi:10.1364/oe.534637 · Cited by 8
Self-Supervised Radio Pre-training: Toward Foundational Models for Spectrogram Learning
Aboulfotouh, A., Eshaghbeigi, A., Karslidis, D., & Abou-Zeid, H. (2024) · Globecom · doi:10.1109/globecom52923.2024.10901287 · Cited by 7
“Deep learning has showcased promising results when applied in wireless communication [5].”
Deep Learning Autoencoders for Reducing PAPR in Coherent Optical Systems
Alnaseri, O., Al-Saedi, I. R. K., Himeur, Y., & Li, H. (2025) · International Conference on Communications · doi:10.1109/icc52391.2025.11161223 · Cited by 8
BS-UE Association and Power Allocation in Heterogeneous Massive MIMO Systems
Yan, H., & Lu, I. (2020) · IEEE Access · doi:10.1109/access.2020.3029518 · Cited by 7
“Note that, as far as the authors know, NN approach for BS-UE association in HM-MIMO systems is currently not available even though the applications of NN have been broadly reported in various aspects of wireless communications such as channel estimation [21], detection [22], positioning [23], exploration of noise corr…”
BackgroundBlind identification of space–time block codes based on deep learning
Zhang, L., Zhang, Y., Yan, W., & Ma, L. (2021) · Chinese Journal of Aeronautics · doi:10.1016/j.cja.2020.10.037 · Cited by 7
Deep Learning Based Channel Estimation with Flexible Delay and Doppler Networks for 5G NR
Saikrishna, P., Chavva, A. K. R., Beniwal, M., & Goyal, A. (2021) · IEEE Wireless Communications and Networking Conference · doi:10.1109/wcnc49053.2021.9417313 · Cited by 9
“The potential applications of the DL to replace various conventional approaches in wireless communications are described in [2]–[5].”
BackgroundLSTM and ResNets Deep Learning Aided End-to-End Intelligent Communication Systems
Zhang, H., Zhang, L., Jiang, Y., & Wu, Z. (2021) · Information Communication Technologies Conference (ICTC) · doi:10.1109/ictc51749.2021.9441657 · Cited by 8
“Recently, the first basic DL-aided end-to-end transmission system is presented in [4].”
BackgroundMethodResultInfluentialJin, G., Hu, Y.-X., Jiao, Y., Wen, J., & Song, Q.-Z. (2021) · Complex · doi:10.1155/2021/5534385 · Cited by 5
Toma, O. H., & López‐Benítez, M. (2021) · IEEE Access · doi:10.1109/access.2021.3109861 · Cited by 7
“, 4-dimensional vector, the s-th element of which is equal to one and zero otherwise [23]).”
BackgroundHybrid neural coded modulation: Design and training methods
Lim, S. H., Han, J., Noh, W., Song, Y., & Jeon, S.-W. (2022) · ICT Express · doi:10.1016/j.icte.2022.01.018 · Cited by 7
Bari, M. F., Agrawal, P., Chatterjee, B., & Sen, S. (2022) · Frontiers in Electronics · doi:10.3389/felec.2022.856284 · Cited by 7
“Some other prominent works for RF fingerprinting using deep learning are mentioned in (O’Shea and Hoydis 2017; Wang et al., 2017; Wang et al., 2018; Zhang et al., 2019).”
BackgroundData-Driven Estimation of Capacity Upper Bounds
Häger, C., & Agrell, E. (2022) · IEEE Communications Letters · doi:10.1109/lcomm.2022.3207385 · Cited by 7
“These approaches have their origin in the seminal work [12], where the authors propose to reinterpret the communication problem as a reconstruction task using parameterized transmitters and receivers, similar to autoencoders (AEs) in machine learning.”
BackgroundFrame Size Optimization Using a Machine Learning Approach in WLAN Downlink MU-MIMO Channel
Kassa, L., Deng, J., Davis, M., & Cai, J. (2022) · International Conference on Data Science and Machine Learning (DSML) · doi:10.5121/csit.2022.121521 · Cited by 7
“Moreover, for the growing diversification of services, users, and the constantly changing channel and traffic dynamics in a networking system, a ML solution is relevant and should be adopted in more effective ways to speed up the decision-making process [4–7].”
MethodDNN-Based Physical-Layer Network Coding for Visible Light Communications
Wang, X., Zhang, R., Xie, X., & Lu, L. (2022) · Photonics · doi:10.3390/photonics10010023 · Cited by 7
Deep Learning Based Interference Exploitation in 1-Bit Massive MIMO Precoding
Hossienzadeh, M., Aghaeinia, H., & Kazemi, M. (2023) · IEEE Access · doi:10.1109/access.2023.3244928 · Cited by 10
Polar Coded Integrated Data and Energy Networking: A Deep Neural Network Assisted End-to-End Design
Xiang, L., Cui, J., Hu, J., Yang, K., & Hanzo, L. (2023) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2023.3262624 · Cited by 7
“Contributions this work [3]–[5] [8] [6] [9]–[11]”
BackgroundGeneralized hybrid LiFi-WiFi UniPHY learning framework towards intelligent UAV-based indoor networks
Ahmad, R., Anwar, D. N., Salameh, H. B., Elgala, H., Ayyash, M., Almajali, S., & El-Khazali, R. (2024) · International Journal of Intelligent Networks · doi:10.1016/j.ijin.2024.05.008 · Cited by 7
Kozlenko, M., & Vialkova, V. (2020) · IEEE International Conference on Advanced Trends in Radioelectronics, Telecommunications and Computer Engineering (TCSET) · doi:10.1109/tcset49122.2020.235501 · Cited by 7
“In the work [14] several novel applications of DL for the physical layer are presented and discussed.”
Performance Evaluation of Physical Attacks against E2E Autoencoder over Rayleigh Fading Channel
Albaseer, A. M., Çiftler, B. S., & Abdallah, M. M. (2020) · IEEE International Conference on Informatics, IoT, and Enabling Technologies (ICIoT) · doi:10.1109/iciot48696.2020.9089601 · Cited by 9
“For experimental setup, we use the MLP and CNN autoencoders to apply white-box and black-box attacks as done in [3], conducting a training process.”
BackgroundMethodInfluentialA Review on Deep Learning-Based Channel Estimation Scheme
Ranjan, A., Singh, A. K., & Sahana, B. C. (2020) · Advances in intelligent systems and computing · doi:10.1007/978-981-15-4032-5_90 · Cited by 7
A Survey about Deep Learning for Constellation Design in Communications
Morales, M. J. L., Chen–Hu, K., & Armada, A. G. (2020) · International Symposium on Communication Systems, Networks and Digital Signal Processing (CSNDSP) · doi:10.1109/csndsp49049.2020.9249528 · Cited by 7
“Among the applications of deep learning in communications, which is an emerging field, there is a technique called end-to-end learning which aims at learning transmitter and receiver architectures to properly communicate under any imperfection or intractability of the channel [9].”
BackgroundMethodInfluentialThe Performance Evaluation of Big Data-Driven Modulation Classification in Complex Environment
Cai, Z., Wang, J., & Ma, M. (2021) · IEEE Access · doi:10.1109/access.2021.3054756 · Cited by 7
“It is found that the modulation classification algorithm is mainly divided into two parts: automatic classification algorithm based on Deep Learning and manual calculation classification algorithm [3]–[6].”
MethodCodebook Design for Noncoherent Communications Using Autoencoders
Sergienko, A. B., & Apalina, P. V. (2021) · IEEE Conference of Russian Young Researchers in Electrical and Electronic Engineering · doi:10.1109/elconrus51938.2021.9396178 · Cited by 7
“In [2], the use of autoencoder networks was introduced for endto-end joint optimization of a communication system.”
MethodBari, M. F., Chatterjee, B., & Sen, S. (2021) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas51556.2021.9401403 · Cited by 7
Artificial Neuronal Networks for Empowering Radio Transceivers: Opportunities and Challenges
Mohammadi, H., & Marojevic, V. (2021) · IEEE Vehicular Technology Conference · doi:10.1109/vtc2021-fall52928.2021.9625494 · Cited by 10
Multimodal Deep Learning Model for Specific Emitter Identification
Zhong, Y., Zhang, L., & Pu, W. (2021) · IEEE International Conference on Signal and Image Processing (ICSIP) · doi:10.1109/icsip52628.2021.9688616 · Cited by 7
“Recently, many new SEI classifiers [12]–[15] have been studied to reveal and separate the emitter UMOP characteristics.”
BackgroundEfficient FPGA Implementation of an ANN-Based Demapper Using Cross-Layer Analysis
Ney, J., Hammoud, B., Dörner, S., Herrmann, M., Clausius, J., ten Brink, S., & Wehn, N. (2022) · Electronics · doi:10.3390/electronics11071138 · Cited by 9
Deep Learning Based Near-Orthogonal Superposition Code for Short Message Transmission
Bian, C., Yang, M., Hsu, C., & Kim, H.-S. (2022) · International Conference on Communications · doi:10.1109/icc45855.2022.9838685 · Cited by 7
“A DNN-based learned code was originally introduced in [10] where the encoder learns a joint coding and modulation scheme generating a length-7 codeword from a length-16 onehot input to achieve the performance similar to (7,4) hamming code.”
MethodBinarized Coherent Optical Receiver Based on Opto-Electronic Neural Network
Yu, Z., Zhao, X., Yang, S., Chen, H., & Chen, M. (2019) · IEEE Journal of Selected Topics in Quantum Electronics · doi:10.1109/jstqe.2019.2931251 · Cited by 7
“ANN can be applied for signal recovery in the receiver which has been demonstrated in [46]–[49].”
Minimum Euclidean distance evaluation using deep neural networks
Lee, J.-H. (2019) · AEU - International Journal of Electronics and Communications · doi:10.1016/j.aeue.2019.152964 · Cited by 7
Deep-Learning-Based Frame Format Detection for IEEE 802.11 Wireless Local Area Networks
Kim, M., Zhang, Z., Kim, D., & Choi, S. (2020) · Electronics · doi:10.3390/electronics9071170 · Cited by 9
“Recently, several meaningful studies on applying deep learning to communications focused on physical layer signal processing, including signal detection, modulation classification, and channel decoding [6,7].”
BackgroundAI-enhanced constellation design for NOMA system: A model driven method
Wang, S., Yu, H., Yuan, Y., Liu, G., & Fei, Z. (2020) · China Communications · doi:10.23919/jcc.2020.11.009 · Cited by 7
“Recently, Artificial Intelligence (AI) has gained significant interest in wireless network PHY-layer design to enhance the performance [11]-[13].”
BackgroundDeep Learning Based Parallel Detector for MIMO Systems
Huang, D., Jiang, X., Chen, S., Wu, Y. P., & Bai, E. (2020) · International Conference on Communication, Image and Signal Processing (CCISP) · doi:10.1109/ccisp51026.2020.9273487 · Cited by 8
“Especially, DL based technologies also have many successful applications in wireless communication [5], [6].”
BackgroundJoint Channel Estimation and Data Decoding using SVM-based ReceiversPreprint
Akın, S., Penner, M., & Peissig, J. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2012.02523 · Cited by 8
“In this regard, the authors in [7] explored several new fundamental opportunities offered by machine learning, in particular deep learning, in the physical layer.”
BackgroundSionna Research Kit: A GPU-Accelerated Research Platform for AI-RAN
Cammerer, S., Marcus, G., Zirr, T., Aoudia, F. A., Maggi, L., Hoydis, J., & Keller, A. (2025) · IEEE International Conference on Machine Learning for Communication and Networking (ICMLCN) · doi:10.1109/icmlcn64995.2025.11140427 · Cited by 6
“Artificial intelligence (AI) for wireless communications has received significant attention from both academia [2] and industry [3].”
Self-citationNOMA Channel State Estimation: Deep Learning Approaches
Khaled, S., Abdelhameed, D., Essai, M. H., Zaini, H. G., Elnazer, S., & Heshmat, S. (2025) · IEEE Access · doi:10.1109/access.2025.3608136 · Cited by 6
Constellation Shaping for OFDM-ISAC Systems: From Theoretical Bounds to Practical Implementation
Geiger, B., Liu, F., Lu, S., Rode, A., Gaviria, D. G., Muth, C., & Schmalen, L. (2026) · IEEE Transactions on Communications · doi:10.1109/tcomm.2026.3669491 · Cited by 6
“…the sensing constraint (C2) into the bitwise AE framework [15], enabling the joint end-to-end optimization of both geometric and probabilistic shaping for S&C. Autoencoders have become a standard and state-of-the-art tool for constellation shaping in communication systems, see, e.g., [29], [30].”
End‐to‐End Multi‐Domain and Multi‐Step Jamming Prediction in Wireless Communications
Su, Z., Qi, N., Jia, L., Chen, J., Liu, Y., & Sun, W. (2021) · Electronics Letters · doi:10.1049/ell2.12128 · Cited by 6
“In recent years, deep neural networks have proved with excellent performance [13].”
BackgroundBrennsteiner, S., Arslan, T., Thompson, J., & McCormick, A. C. (2022) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2022.3225505 · Cited by 8
“a) optimization or extension of existing algorithms or by the introduction of new machine learning-based algorithms [3], [4]; b) joint optimization of various algorithms by removing classical block boundaries [5].”
BackgroundMethodOn Assessing Vulnerabilities of the 5G Networks to Adversarial Examples
Zolotukhin, M., Miraghaei, P., Zhang, D., & Hämäläinen, T. (2022) · IEEE Access · doi:10.1109/access.2022.3225921 · Cited by 6
“The first fundamental study in this area [25] proposes to interpret an end-to-end communication system as an autoencoder, where both the transmitter and receiver are implemented as deep neural networks.”
BackgroundGros, A., Moeyaert, V., & Mégret, P. (2022) · IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) · doi:10.1109/pimrc54779.2022.9977690 · Cited by 6
“The EMD methodology is suitable for non linear and non stationary signals that are typical features of real telecommunication signals [7] [8].”
MethodA New Approach for an End-to-end Communication System Using Variational Auto-encoder (VAE)
Alawad, M. A., Hamdan, M. Q., Hamdi, K. A., Foh, C. H., & Quddus, A. U. (2022) · Globecom · doi:10.1109/globecom48099.2022.10001178 · Cited by 6
“3 : 1) VAE input: The hot vector in [1] can be replaced with a new concept we call it image hot vector (IHV) in the same way that [14] used to represent the constellation of a symbol.”
BackgroundMethodInfluentialImproving ADMM-based massive MIMO detectors via deep learning
Tiba, I. N., & Zhang, Q. (2023) · Digital Signal Processing · doi:10.1016/j.dsp.2023.104027 · Cited by 8
BERT-Aided Sequence Detection With Natural Redundancy
Wang, Z., Yu, H., Zhu, Z., Du, J., Shen, C., & Shen, Z. (2023) · IEEE Communications Letters · doi:10.1109/lcomm.2023.3297129 · Cited by 7
“performance of sequence detection [8], [9], [10].”
A Deep Learning Approach for Universal NPRACH Detection With Inter-Cell Interference
Li, R., Xue, J., Sun, J. Z., & Chatzinotas, S. (2023) · IEEE Transactions on Communications · doi:10.1109/tcomm.2023.3337272 · Cited by 6
“As we know, deep learning (DL) has been widely used in wireless communication to enhance the system performance [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53].”
BackgroundSiriwardana, G. K., Jayawardhana, H. D., Bandara, W. U., Atapattu, S., & Herath, V. R. (2023) · Moratuwa Engineering Research Conference (MERCon) · doi:10.1109/mercon60487.2023.10355464 · Cited by 6
“DL is particularly effective in extracting complex, high-dimensional features from wireless signals that cannot be accurately modeled using traditional methods [12].”
BackgroundEffects of diagram plane on neural network based modulation recognition
Leblebici, M., Çalhan, A., & Ci̇ci̇oğlu, M. (2024) · Applied Soft Computing · doi:10.1016/j.asoc.2024.111412 · Cited by 6
Wang, J., Wang, C., Zhang, H., Zhang, W., & Ng, D. W. K. (2023) · Globecom · doi:10.1109/globecom54140.2023.10437948 · Cited by 6
“The confusion matrices reveal that WBFM signals are easily mis-classified as AM-DSB signals due to the similarity in their analog nature and the small observation window that capture only the carrier tones during the silent period [26].”
BackgroundZhu, Y., Ye, J., Yan, L., Zhou, T., Yu, X., Zou, X., & Pan, W. (2024) · Journal of Lightwave Technology · doi:10.1109/jlt.2024.3403129 · Cited by 6
A lightweight detection algorithm for tooth cracks in optical images
Xie, Z., Hu, X., Guo, L., Lin, W., Liu, J., Zhang, C., Ge, G., Tang, Y., & Wang, W. (2024) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2024.109153 · Cited by 6
Concept Drift Detection for Deep Learning Aided Receivers in Dynamic Channels
Uzlaner, N., Raviv, T., Shlezinger, N., & Todros, K. (2024) · IEEE International Workshop on Signal Processing Advances in Wireless Communications (SPAWC) · doi:10.1109/spawc60668.2024.10694247 · Cited by 7
“The direct approach to tackle this, coined joint learning [6], trains the DNN over a broad range of channel conditions, seeking a non-coherent receiver at the cost of performance degradation compared with coherent operation.”
BackgroundU-Shaped Error Correction Code Transformers
Nguyen, D.-T., & Kim, S. (2024) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2024.3482349 · Cited by 6
“On the other hand, model-free approaches [8], [9], [10], [11], [12], [13], [14], [48], which were pioneered by the authors of [8], were not subject to the complexity of the Tanner-graph but instead hindered by the curse of dimensionality, especially with significantly large code length.”
Vijay, E. V., & Aparna, K. (2024) · Physica Scripta · doi:10.1088/1402-4896/ad8cb2 · Cited by 7
Sindal, S. S., & Trivedi, Y. N. (2024) · Physical Communication · doi:10.1016/j.phycom.2024.102534 · Cited by 6
Deep Learning Based Joint Multi-User MISO Power Allocation and Beamforming Design
Vahapoglu, C., O’Shea, T. J., Roy, T., & Ulukuş, Ş. (2024) · IEEE Military Communications Conference (MILCOM) · doi:10.1109/milcom61039.2024.10773663 · Cited by 6
“This includes tasks such as interference management, transceiver chain design, error-correcting algorithms design to provide reliable data transfer [1].”
MethodSelf-citationBilal, N., & Velmurugan, T. (2024) · International Journal of Communication Systems · doi:10.1002/dac.6060 · Cited by 6
A Hybrid Quantum-Classical Autoencoder Framework for End-to-End Communication Systems
Zhang, B. L., Zheng, G., & Huynh, N. V. (2024) · IEEE Wireless Communications Letters · doi:10.1109/lwc.2024.3524330 · Cited by 12
“The E2E paradigm is pioneered in [2], where the authors have proposed a foundational channel autoencoder architecture for the entire communication system, demonstrating near-optimal block error rate (BLER) performance compared to existing baseline modulation and coding schemes.”
InfluentialEfficient Deep Learning-Based Detection Scheme for MIMO Communication Systems
Ibarra-Hernández, R. F., Castillo-Soria, F. R., Gutierrez, C. A., Puerto-Flores, J. A. D., Acosta‐Elías, J., Rodríguez-Abdalá, V. I., & Palacios-Luengas, L. (2025) · Sensors · doi:10.3390/s25030669 · Cited by 6
Chen, S., Tan, C. W., Zhai, X., & Poor, H. V. (2025) · IEEE Transactions on Green Communications and Networking · doi:10.1109/tgcn.2025.3541338 · Cited by 7
“However, as discussed in [24], [25], machine learning techniques are unlikely to completely replace resource allocation algorithms in communication systems, which are often designed from an optimization theory perspective.”
InfluentialEnhanced End-to-End and Consistent Time-Frequency Analysis for Tracking
Ding, M., Ding, Y., Dongye, G., & Lv, P. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3543196 · Cited by 6
“In signal processing, DL-based approaches also have shown considerable promise, especially in fields like speech recognition [36], [37], [38], signal estimation, and modulation recognition [39].”
Multi-User Visible Light Communications With Probabilistic Constellation Shaping and Precoding
Nguyen, T. K., Pham, T. V., Le, H. D., Nguyen, C. T., & Pham, A. T. (2025) · IEEE Transactions on Communications · doi:10.1109/tcomm.2025.3543208 · Cited by 6
“The idea of using end-to-end learning for physical layer design has been introduced in [44], where the end-to-end performance of communication systems can be optimized by utilizing a deep neural network (NN) known as AE.”
Niu, X., Bai, B. B., Guo, N., Zhang, W., & Han, W. (2025) · Entropy · doi:10.3390/e27040373 · Cited by 20
“Recent advances in AI-empowered communication have reimagined communication systems as a joint optimization of transmitter and receiver through end-to-end training for reconstruction tasks [12,13].”
A Novel PAPR Reduction Scheme Based on Deep Autoencoder Network for FBMC Systems
Cheng, X., Chen, D., Li, S., He, S., & Dejin, K. (2025) · IEEE Access · doi:10.1109/access.2025.3562066 · Cited by 6
End-to-End Deep Learning in Phase Noisy Coherent Optical Link
Alnaseri, O., & Himeur, Y. (2024) · IEEE Future Networks World Forum (FNWF) · doi:10.1109/fnwf63303.2024.11028865 · Cited by 9
Adaptive Meta-learning-based Adversarial Training for Robust Automatic Modulation Classification
Bamdad, A., Owfi, A., & Afghah, F. (2025) · IEEE International Conference on Communications workshops · doi:10.1109/iccworkshops67674.2025.11162364 · Cited by 6
“Limitations in the availability of training samples and training resources often result in a few-shot learning scenario, a factor largely neglected in the current AMC literature [1, 4, 5, 7, 9].”
Machine Learning Based Featureless Signalling
Shakeel, I. (2018) · IEEE Military Communications Conference (MILCOM) · doi:10.1109/milcom.2018.8599694 · Cited by 7
“An interesting ML-based concept for the physical layer has been proposed in [1].”
BackgroundMethodInfluentialGrand Challenges in Signal Processing for Communications
She, C., Cheng, P., Li, A., & Li, Y. (2021) · Frontiers in Signal Processing · doi:10.3389/frsip.2021.664331 · Cited by 7
A Low-Complexity Transceiver Design for Terahertz Communication based on Deep Learning
Che, B., Li, X., Chen, Z., & He, Q. (2022) · IEEE International Conference on Communications workshops · doi:10.1109/iccworkshops53468.2022.9814636 · Cited by 6
Comprehensive Literature Survey for mm-Wave Massive MIMO Using Machine Learning for 6G
Devnikar, R., & Hendre, V. (2022) · Lecture notes in electrical engineering · doi:10.1007/978-981-16-7985-8_80 · Cited by 6
Sparse Code Multiple Access Scheme Based on Variational Learning
Yuan, Q., Wang, Z., Li, D., Guo, Q., & Xiang, W. (2022) · IEEE Transactions on Communications · doi:10.1109/tcomm.2022.3214236 · Cited by 6
“In [12], the authors propose an end-to-end communication system based on the principles of the autoencoder.”
BackgroundMethodBig data network security defense mode of deep learning algorithm
Yu, Y. (2022) · Open Computer Science · doi:10.1515/comp-2022-0257 · Cited by 6
Metadata
- Authors
- O'Shea, T., & Hoydis, J.
- Affiliations
- Nokia (France) (FR)
- Venue
- IEEE Transactions on Cognitive Communications and Networking
- Published
- 2017-12
- Volume
- 3
- Issue
- 4
- Pages
- 563-575
- ISSN
- 2332-7731 (ISSN-L)
- OpenAlex
- W2734408173
- Semantic Scholar
- 78989616eeeac55b202e3e4205225e7135054185
- Cited by
- 2,980
- References
- 69
- Citations by year
- 2026: 307 · 2025: 327 · 2024: 305 · 2023: 368 · 2022: 432 · 2021: 409 · 2020: 406 · 2019: 322 · 2018: 95 · 2017: 9
- FWCI
- 122.29
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE Transactions on Cognitive Communications and Networking (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Wireless Signal Modulation Classification
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Advanced Wireless Communication Techniques
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Advanced Wireless Communication Technologies
- Keywords
- end-to-end reconstruction, domain knowledge integration, convolutional neural networks, automatic modulation classification, IQ samples
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.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-11-07, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-11-07 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
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.