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 1,001–1,100 of 2,995
Radio Signal Automatic Modulation Classification based on Deep Learning and Expert Features
Yao, T., Chai, Y., Wang, S., Miao, X., & Bu, X. (2020) · IEEE Information Technology, Networking, Electronic and Automation Control Conference (ITNEC) · doi:10.1109/itnec48623.2020.9085077 · Cited by 13
“If the selected features and classifier are not suitable, the recognition rate may be very poor, that is to say, the generalization based on artificial feature selection is weak [8].”
BackgroundJoint Constellation Design and Multiuser Detection for Grant-Free NOMA
Ma, Z., Wu, W. R., Jian, M., Gao, F., & Shen, X. (2021) · IEEE Transactions on Wireless Communications · doi:10.1109/twc.2021.3108666 · Cited by 15
“The idea was first pioneered in [30], and has subsequently led to many extensions [31]– [33].”
BackgroundCombined Deep Learning and SOR Detection Technique for High Reliability in Massive MIMO Systems
Jang, J., Park, C.-Y., Shin, B.-S., & Song, H. (2021) · IEEE Access · doi:10.1109/access.2021.3125002 · Cited by 15
“Many studies that use deep learning (DL) algorithms in the field of communication systems including the physical layer have been conducted due to the successful cases of DL application in various fields, and this has led to the emergence of DL-based detection techniques [5]–[8].”
BackgroundAutoencoder-based Communications with Reconfigurable Intelligent Surfaces
Erpek, T., Sagduyu, Y. E., Alkhateeb, A., & Yener, A. (2021) · IEEE International Symposium on Dynamic Spectrum Access Networks · doi:10.1109/dyspan53946.2021.9677378 · Cited by 20
Deep-learning methods for integrated sensing and communication in vehicular networks
Zhang, Z., Chang, Q., Xing, J., & Chen, L. (2023) · Vehicular Communications · doi:10.1016/j.vehcom.2023.100574 · Cited by 18
Machine learning methods for SIR prediction in cellular networks
Rozenblit, O., Haddad, Y., Mirsky, Y., & Azoulay, R. (2018) · Physical Communication · doi:10.1016/j.phycom.2018.08.005 · Cited by 14
Deep Reinforcement Learning-Based Power Control in Full-Duplex Cognitive Radio Networks
Meng, X., İnaltekin, H., & Krongold, B. S. (2018) · Globecom · doi:10.1109/glocom.2018.8647699 · Cited by 16
Learning-Based Iterative Interference Cancellation for Cognitive Internet of Things
Liu, Y., Kuai, X., Yuan, X., Liang, Y., & Zhou, L. (2019) · IEEE Internet of Things Journal · doi:10.1109/jiot.2019.2915598 · Cited by 13
Visible Light Communication Using Deep Learning Techniques
Pachpande, P. G., Khadr, M. H., Hussein, A. F., & Elgala, H. (2018) · IEEE Sarnoff Symposium · doi:10.1109/sarnof.2018.8720493 · Cited by 13
“AE based VLC system is designed as the parameters listed in Table I and II.”
BackgroundMethodInfluentialA Deep Neural Network Method for Automatic Modulation Recognition in OFDM with Index Modulation
Liu, F., Zhou, Y., & Liu, Y. (2019) · IEEE Vehicular Technology Conference · doi:10.1109/vtcspring.2019.8746286 · Cited by 13
“This structure is proved to be a kind of effective model for modulation types recognition [18].”
BackgroundAlfarozi, S. A. I., Pasupa, K., Hashizume, H., Woraratpanya, K., & Sugimoto, M. (2019) · IEEE Access · doi:10.1109/access.2019.2952465 · Cited by 15
“[10] T. O’Shea and J. Hoydis, ‘‘An introduction to deep learning for the physical layer,’’ IEEE Trans.”
BackgroundAccelerating wireless channel autoencoders for short coherence-time communications
Morocho-Cayamcela, M. E., & Lim, W. (2020) · Journal of Communications and Networks · doi:10.1109/jcn.2020.000011 · Cited by 13
“S. Dörner et al. [17], extended the existing ideas to continuous end-to-end data transmission over-the-air which eases the restriction to short block lengths found on [16].”
BackgroundMethodInfluentialThe Performance Analysis of Signal Recognition Using Attention Based CNN Method
Yin, Z., Chen, B., Zhen, W., Wang, C., & Zhang, T. (2020) · IEEE Access · doi:10.1109/access.2020.3038208 · Cited by 13
“There is a trend in AMR applied to image processing and I/Q sequence rather than crafting expert features [7].”
BackgroundEnd-to-End Learning in Optical Fiber Communications: Experimental Demonstration and Future Trends
Karanov, B., Oliari, V., Chagnon, M., Liga, G., Alvarado, A., Aref, V., Lavery, D., Bayvel, P., & Schmalen, L. (2020) · European Conference on Optical Communication · doi:10.1109/ecoc48923.2020.9333265 · Cited by 14
Jing, S., & Xiao, C. (2021) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2021.3096466 · Cited by 13
On Deep Learning Classification of Digitally Modulated Signals Using Raw I/Q Data
Snoap, J. A., Popescu, D. C., & Spooner, C. M. (2022) · IEEE Consumer Communications and Networking Conference · doi:10.1109/ccnc49033.2022.9700688 · Cited by 17
“Recently, deep learning (DL) approaches to classifying digitally modulated signals have also been proposed, such as those in [4], [5], which use the raw I/Q signal components for training and signal recognition/classification, or the alternative approaches in [6]–[10], in which the amplitude/phase…”
BackgroundPower Allocation in Multi-User Cellular Networks: Deep Reinforcement Learning ApproachesPreprint
Meng, F. Y., Chen, P., Wu, L., & Cheng, J. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1901.07159 · Cited by 13
“These methods are usually modelfree and data-driven [11], [12], and the solutions are obtained through data learning instead of model-oriented analysis and design.”
MethodAn End-to-End Block Autoencoder For Physical Layer Based On Neural NetworksPreprint
Mu, T., Chen, X., Chen, L., Yin, H., & Wang, W. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1906.06563 · Cited by 16
“In [6], CNNs accomplish classification tasks for different modulation schemes.”
BackgroundChannel charting based beamformingPreprint
Magoarou, L. L., Yassine, T., Paquelet, S., & Crussière, M. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2212.12340 · Cited by 13
“One can cite for example detection [1], [2], channel estimation [3], [4], [5], [6] and even end-to-end system learning [7], [8] (see [9], [10] for exhaustive surveys on this topic).”
BackgroundJee, J., & Park, H. (2023) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2023.3327276 · Cited by 15
Joint estimation model for FSO channel parameters and performance evaluation based on CNNs
Chen, D., Wang, R., Wang, C., Gao, Y., & Chen, H. (2024) · Applied Optics · doi:10.1364/ao.514064 · Cited by 13
Li, Z., Huang, O., Yan, A., Li, G., Dong, B., Shen, W., Xing, S., Shi, J., Li, Z., Shen, C., et al. (2024) · Journal of Lightwave Technology · doi:10.1109/jlt.2024.3488121 · Cited by 13
Sums: Sniffing Unknown Multiband Signals Under Low Sampling Rates
Peng, J., Chen, Z., Lin, Z., Yuan, H., Fang, Z., Bao, L., Song, Z., Li, Y., Ren, J., & Gao, Y. (2024) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2024.3509861 · Cited by 27
Learning-Aided Physical Layer Attacks Against Multicarrier Communications in IoT
Nooraiepour, A., Bajwa, W. U., & Mandayam, N. B. (2020) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2020.2990657 · Cited by 13
“In particular, the authors in [15] proposed an end-to-end learning of communication systems based on deep neural networks (DNNs).”
BackgroundMethodModified gate activation functions of Bi-LSTM-based SC-FDMA channel equalization
Mohamed, M. A., Hassan, H. A., Essai, M. H., Esmaiel, H., Mubarak, A. S., & Omer, O. A. (2023) · Journal of Electrical Engineering · doi:10.2478/jee-2023-0032 · Cited by 13
Grant-Free Non-Orthogonal Multiple Access: A Key Enabler for 6G-IoTPreprint
Abbas, R., Huang, T., Shahab, M. B., Shirvanimoghaddam, M., Li, Y., & Vucetic, B. S. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2003.10257 · Cited by 14
“We note here that the optimal choice of loss function remains an open problem [6].”
BackgroundMethodInfluentialEnd-to-End Learning of Geometrical Shaping Maximizing Generalized Mutual Information
Gümüş, K., Alvarado, A., Chen, B., Häger, C., & Agrell, E. (2020) · TU/e Research Portal · doi:10.1364/ofc-2020-w3d.4 · Cited by 12
THz ISCI: Terahertz Integrated Sensing, Communication and Intelligence
Wu, Y., Han, C. Z., & Chen, Z. (2021) · International Conference on Infrared, Millimeter, and Terahertz Waves (IRMMW-THz) · doi:10.1109/irmmw-thz50926.2021.9567142 · Cited by 12
Two-stage deep learning framework for sRGB image white balance
Farghaly, M., Mansour, R. F., & Sewisy, A. A. E.-M. (2022) · Signal Image and Video Processing · doi:10.1007/s11760-022-02230-2 · Cited by 12
“Due to the remarkable results achieved by deep learning meth-ods in several research fields, such as computer networking and communications [21–23], we propose a deep learning-based method to solve our problem.”
BackgroundDeep learning-driven MIMO: Data encoding and processing mechanism
Song, Z., & Ma, J. (2022) · Physical Communication · doi:10.1016/j.phycom.2022.101976 · Cited by 17
ResNet-WGAN-Based End-to-End Learning for IoV Communication With Unknown Channels
Zhao, J., Mu, H., Zhang, Q., & Zhang, H. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3274209 · Cited by 12
Survey of Research on Application of Deep Learning in Modulation Recognition
Sun, Y., & Wu, W. (2023) · Wireless Personal Communications · doi:10.1007/s11277-023-10826-1 · Cited by 14
“Researchers have also proved that the recognition method based on DL is easier to be applied to time series data, and the accuracy is several times higher than that of the best performance of the traditional feature-based classifier [69, 123].”
Machine Learning Based Beam Selection for Maximizing Wireless Network Capacity
Geranmayeh, P., & Grass, E. (2024) · IEEE Access · doi:10.1109/access.2024.3381542 · Cited by 15
Pramitarini, Y., Perdana, R. H. Y., Shim, K., & An, B. (2024) · IEEE Access · doi:10.1109/access.2024.3392255 · Cited by 13
Deep Learning-Based Detection for Marker Codes Over Insertion and Deletion Channels
Ma, G., Jiao, X., Mu, J., Han, H., & Yang, Y. (2024) · IEEE Transactions on Communications · doi:10.1109/tcomm.2024.3394039 · Cited by 14
“Motivated by the application of deep learning for physical layer communications [23], [24], [25], [26], [27], [28], we propose deep learning-based detection methods for marker code over insertion and deletion channels to solve the performance degradation problem due to CSI uncertainty or inaccurate…”
MethodRode, A., Farsi, M., Lauinger, V., Karlsson, M., Agrell, E., Schmalen, L., & Häger, C. (2024) · Optical Fiber Technology · doi:10.1016/j.yofte.2024.104047 · Cited by 17
“Indeed, treating a communication system as a denoising AE can facilitate transmitter optimizations, and, more generally, joint transmitter and receiver design [110].”
Deep Learning-Assisted Multi-Dimensional Modulation and Resource Mapping for Advanced OFDM Systems
Kim, J.-H., Lee, B., Lee, H., Kim, Y., & Lee, J. (2018) · IEEE Globecom Workshops (GC Wkshps) · doi:10.1109/glocomw.2018.8644281 · Cited by 14
Böhm, S., & König, H. (2024) · Computer Communications · doi:10.1016/j.comcom.2024.02.006 · Cited by 14
Deep Learning-based Pilot Adaptation and Channel Estimation in OFDM Systems
Shahmohammadi, M., Sebghati, M., & Zareian, H. (2024) · Wireless Personal Communications · doi:10.1007/s11277-024-10937-3 · Cited by 12
Li, J., Yang, Y., & Yan, Z. (2024) · IEEE Wireless Communications · doi:10.1109/mwc.013.2300448 · Cited by 12
“However, [1] and [2] only discussed the traditional PHYSKG schemes, while in [3] and [4], DL applications related to key generation were not discussed.”
BackgroundTransformer-empowered receiver design of OFDM communication systems
Yue, B., Qiu, S., Yang, C. X., Peng, L., & Zhang, Y. (2024) · Computer Communications · doi:10.1016/j.comcom.2024.107960 · Cited by 15
A Review of Reinforcement Learning for Semantic Communications
Xiao, Y., Xiumei, F., Yau, K. A., Xie, Z., Rui, M., & Gang, Y. (2025) · Journal of Network and Systems Management · doi:10.1007/s10922-025-09927-y · Cited by 20
“Based on the DeepSC [16], we proposes a SC system model as shown in Fig.”
Enabling Simulation-Based Optimization through Machine Learning: A Case Study on Antenna Design
Testolina, P., Lecci, M., Rebato, M., Testolin, A., Gambini, J., Flamini, R., Mazzucco, C., & Zorzi, M. (2019) · Globecom · doi:10.1109/globecom38437.2019.9013240 · Cited by 12
“This can be applied at many levels, starting from the pure, physical channel model, to the entire transmitterchannel-receiver chain [7].”
BackgroundLearning-Augmented Streaming Codes are Approximately Optimal for Variable-Size Messages
Rudow, M., & Rashmi, K. V. (2022) · IEEE International Symposium on Information Theory (ISIT) · doi:10.1109/isit50566.2022.9834539 · Cited by 14
Jointly Learned Symbol Detection and Signal Reflection in RIS-Aided Multi-user MIMO Systems
Wang, L., Shlezinger, N., Alexandropoulos, G. C., Zhang, H., Wang, B., & Eldar, Y. C. (2021) · Asilomar Conference on Signals, Systems and Computers (ACSSC) · doi:10.1109/ieeeconf53345.2021.9723108 · Cited by 13
“Recent years have witnessed growing interest in the application of DNNs for Receiver (RX) design; see detailed surveys in [4]–[7].”
BackgroundA Robust Graph Convolutional Neural Network-Based Classifier for Automatic Modulation Recognition
Ghasemzadeh, P., Hempel, M., & Sharif, H. (2022) · International Wireless Communications and Mobile Computing Conference (IWCMC) · doi:10.1109/iwcmc55113.2022.9825340 · Cited by 14
Dong, Y., Wang, H., & Yao, Y. (2022) · International Conference on Communications · doi:10.1109/icc45855.2022.9838452 · Cited by 16
“1 illustrates a typical end-to-end autoencoder communication system [1], which is implemented in this paper.”
BackgroundMethodInfluentialA Multi-Rate Approach for Nonlinear Pre-Distortion Using End-to-End Deep Learning in IM-DD Systems
Minelli, L., Forghieri, F., Nespola, A., Straullu, S., & Gaudino, R. (2022) · Journal of Lightwave Technology · doi:10.1109/jlt.2022.3216591 · Cited by 12
“This ANN can be seen equivalently as a FFNN whose inputs are entries of a tap delay line.”
BackgroundMethodInfluentialA Residual Learning-Aided Convolutional Autoencoder for SCMA
Jiang, F., Chang, D.-W., Ma, S., Hu, Y., & Xu, Y. (2023) · IEEE Communications Letters · doi:10.1109/lcomm.2023.3260881 · Cited by 12
“In recent years, deep learning has been used to solve various problems in communication systems because of its powerful image recognition and classification capabilities [8], and it has also been widely studied in SCMA systems.”
BackgroundMeta-Learning for Wireless Interference Identification
Owfi, A., Afghah, F., & Ashdown, J. (2023) · IEEE Wireless Communications and Networking Conference · doi:10.1109/wcnc55385.2023.10119039 · Cited by 12
“In [8], a CNN-based model was used for classifying ten different single carrier modulation schemes based on their radio frequency time-series data.”
BackgroundMethodInfluentialXu, T., & Darwazeh, I. (2023) · Communications Engineering · doi:10.1038/s44172-023-00108-w · Cited by 14
“For machine learning applications aiming for both transmitter and receiver sides, end-to-end autoencoder 41,42 is the representative technique.”
MethodMitigating nonlinear distortions of high-powered LEDs for VLC using deep neural networks
Abhaynarasimha, K. S., Mani, V. V., & Sellathurai, M. (2023) · Optics Communications · doi:10.1016/j.optcom.2023.129997 · Cited by 14
Variational autoencoder-enhanced deep neural network-based detection for MIMO systems
Omondi, G., & Olwal, T. O. (2023) · e-Prime - Advances in Electrical Engineering Electronics and Energy · doi:10.1016/j.prime.2023.100335 · Cited by 17
Varasteh, M., Hoydis, J., & Clerckx, B. (2019) · IEEE International Workshop on Signal Processing Advances in Wireless Communications (SPAWC) · doi:10.1109/spawc.2019.8815474 · Cited by 14
“Some examples include decoder design for existing channel codes [12], blind channel equalization [13], learning physical layer signal representation for SISO [11] and, OFDM systems [14].”
BackgroundMethodSelf-citationErnest, T. Z. H., Madhukumar, A. S., Sirigina, R. P., & Krishna, A. K. (2020) · Vehicular Communications · doi:10.1016/j.vehcom.2020.100235 · Cited by 15
Deep Learning Aided Signal Detection in OFDM Systems with Time-Varying Channels
Yao, R., Wang, S., Zuo, X., Xu, J., & Qi, N. (2019) · IEEE Pacific Rim Conference on Communications, Computers and Signal Processing (PACRIM) · doi:10.1109/pacrim47961.2019.8985060 · Cited by 16
“For example, several novel ways of applying deep learning to the physical layer have been discussed in [5]–[7].”
BackgroundA Deep Learning Based Channel Estimation for High Mobility Vehicular Communications
Mehrabi, M., Mohammadkarimi, M., Ardakani, M., & Jing, Y. (2020) · 2012 International Conference on Computing, Networking and Communications (ICNC) · doi:10.1109/icnc47757.2020.9049735 · Cited by 12
A Novel Intelligent SIC Detector for NOMA Systems Based on Deep Learning
Fu, J., Xiao, Y., Liu, H., Yang, P., & Zhang, B. (2021) · IEEE Vehicular Technology Conference · doi:10.1109/vtc2021-spring51267.2021.9449008 · Cited by 12
“In addition, more applications of deep learning in physical layer communication can be referred in [17] and [18].”
BackgroundBeck, E., Bockelmann, C., & Dekorsy, A. (2021) · IEEE Transactions on Communications · doi:10.1109/tcomm.2021.3114682 · Cited by 15
“Notably, the authors of the early work [16] demonstrate a complete communication system design by interpreting transmitter, channel and receiver as an autoencoder which is trained end-to-end similar to one DNN.”
BackgroundMethodSoufleri, E., & Roy, K. (2021) · IEEE Access · doi:10.1109/access.2021.3116418 · Cited by 14
“Deep Neural Networks are state-of-the-art solutions for solving a plethora of complex tasks, varying from image classification, object detection, and voice recognition tasks to communication and networking applications [1], [2].”
BackgroundTowards recurrent neural network with multi-path feature fusion for signal modulation recognition
Lei, Z., Jiang, M., Yang, G., Guan, T., Huang, P. F., Gu, Y., Xu, Z., & Ye, Q. (2022) · Wireless Networks · doi:10.1007/s11276-021-02877-8 · Cited by 12
“Recently, with the help of the rapid development of DL technology, a series of DL-based AMR approaches are proposed [7–10].”
MethodEnd-to-End Learning-Based Framework for Amplify-and-Forward Relay Networks
Gupta, A. K., & Sellathurai, M. (2021) · IEEE Access · doi:10.1109/access.2021.3085901 · Cited by 12
“However, none of theworks [1], [5]–[17], have analyzed and contrasted the bit-wise AE-based designed constellations over that of the symbol-wise AE by analyzing the coded-modulation designs either for the P2P or relay networks.”
BackgroundMethodInfluentialDeep Learning Based Spatial User Mapping on Extra Large MIMO ArraysPreprint
Amiri, A., Manchón, C. N., & de Carvalho, E. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2002.00474 · Cited by 12
“Different problems in the physical layer such as rate maximization, modulation schemes, power control are investigated using machine learning techniques [8].”
BackgroundMethodFueling the Next Quantum Leap in Cellular Networks: Embracing AI in 5G Evolution towards 6GPreprint
Lin, X., Chen, M., Rydén, H., Jeong, J., Lee, H., Sundberg, M., Timo, R. N., Razaghi, H. S., & Poor, H. V. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2111.10663 · Cited by 15
“Utilizing AI can help to approximate the true underlying signal characteristics [12].”
BackgroundCNN-based Channel Estimation using NOMA for mmWave Massive MIMO System
Anu, T. S., & Raveendran, T. (2023) · IEEE Statistical Signal Processing Workshop (SSP) · doi:10.1109/ssp53291.2023.10207968 · Cited by 12
“This features leads to utilize deep learning in communication system [6].”
BackgroundPEACH: Proactive and Environment-Aware Channel State Information Prediction with Depth Images
Ayvaşık, S., Mehmeti, F., Babaians, E., & Kellerer, W. (2023) · Proceedings of the ACM on Measurement and Analysis of Computing Systems · doi:10.1145/3579450 · Cited by 14
Deep Learning Based Joint Hybrid Precoding and Combining Design for mmWave MIMO Systems
Liu, F., Li, X., Yang, X., Shi, H., Shi, B., & Du, R. (2024) · IEEE Systems Journal · doi:10.1109/jsyst.2024.3357712 · Cited by 12
A review on Recent Trends in Deep Learning-based Channel Estimation Techniques
Susandhika, M., & Edward, A. S. (2024) · International Conference on Data Engineering and Communication Systems (ICDECS) · doi:10.1109/icdecs59733.2023.10503562 · Cited by 11
End-to-end learning of adaptive coded modulation schemes for resilient wireless communications
Davey, C. P., Shakeel, I., Deo, R. C., Sharma, E., Salcedo‐Sanz, S., & Soar, J. (2024) · Applied Soft Computing · doi:10.1016/j.asoc.2024.111672 · Cited by 11
Kenanoğlu, R., Mert, İ., Baydar, C., Köse, Ö., & Yağlı, H. (2024) · Energy · doi:10.1016/j.energy.2024.131901 · Cited by 12
Automatic classification of multi-carrier modulation signal using STFT spectrogram and deep CNN
Tailor, A., Dua, M., & Verma, P. (2024) · Physica Scripta · doi:10.1088/1402-4896/ad538a · Cited by 13
Deep learning in wireless communications for physical layer
Zhao, J., Liu, C., Liao, J., & Wang, D. (2024) · Physical Communication · doi:10.1016/j.phycom.2024.102503 · Cited by 16
COMSPLIT: A Communication-Aware Split Learning Design for Heterogeneous IoT Platforms
Ninković, V., Vukobratović, D., Mišković, D., & Zennaro, M. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3487246 · Cited by 20
“Following a similar approach to [50], we keep the channel conditions constant during training (i.e., erasure probability for the erasure channel or noise variances for the AWGN channel), which can be considered as a neural network hyperparameter.”
Exploring Active Inference for Efficient Resource Allocation in AAV-Enabled Cognitive NOMA Uplink
Obite, F., Krayani, A., Alam, A. S., Marcenaro, L., Nallanathan, A., & Regazzoni, C. S. (2024) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2024.3510577 · Cited by 11
“In [36], [37], the authors explored model-free and data-driven deep learning (DL) methods to minimize computational complexity using available input and output training datasets [38].”
Deep learning approaches and data augmentation for melanoma detection
Alzamel, M., Iliopoulos, C. S., & Lim, Z. (2024) · Neural Computing and Applications · doi:10.1007/s00521-024-10590-8 · Cited by 11
“Each layer is a made of several 2D planes that are composed of independent neurons [20, 21].”
Model-Driven Deep-Learning for End-to-End Optimization in Fiber-Terahertz Communication Systems
Li, Z., Wang, C., Jia, J., Huang, O., Dong, B., Li, G., Xing, S., Zhou, Y., Shi, J., Li, Z., et al. (2024) · Journal of Lightwave Technology · doi:10.1109/jlt.2024.3519360 · Cited by 14
Wang, D., Du, L., Li, Z., Cheng, N., Si, J., & Shen, W. (2025) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2025.3614368 · Cited by 11
“Following the introduction of deep learning into the communication field by O’shea et al. [17], [18], researchers have perceived the potential of deep learning (DL) in the domain of electromagnetic [19], [20].”
Khan, M. A., Liu, S., Bilal, M., & Hassan, A. (2025) · Ocean Engineering · doi:10.1016/j.oceaneng.2025.123720 · Cited by 11
Integrated Radio Sensing Capabilities for 6G Networks: AI/ML Perspective
Shatov, V., Schieler, S., Muth, C., Mateos-Ramos, J. M., de Almeida, I. B. F., Euchner, F., Semper, S., ten Brink, S., Fettweis, G. P., Häger, C., et al. (2026) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2026.3668458 · Cited by 14
“The classifier architecture in [6] itself is a straightforward CNN composed of narrowing convolutional layers followed by FC layers and a dense Softmax output layer.”
InfluentialBand Assignment in Dual Band Systems: A Learning-Based Approach
Burghal, D., Wang, R., & Molisch, A. F. (2018) · IEEE Military Communications Conference (MILCOM) · doi:10.1109/milcom.2018.8599747 · Cited by 14
A Deep Learning Wireless Transceiver with Fully Learned Modulation and Synchronization
Schmitz, J., von Lengerke, C., Airee, N., Behboodi, A., & Mathar, R. (2019) · IEEE International Conference on Communications workshops · doi:10.1109/iccw.2019.8757051 · Cited by 12
“A deep neural network (DNN) autoencoder was proposed in [2] as a first attempt to endto-end learning of communication systems.”
BackgroundMethodResearch Status of Active Control of Hydraulic Fluid Pulsation
Xia, S., Hua, Z., & Huayong, Y. (2019) · Journal of Mechanical Engineering · doi:10.3901/jme.2019.24.216 · Cited by 14
Qi, P., Lin, X., Shi, H., Bao, J., Li, X., & Zhuang, Y. (2020) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2020.3015241 · Cited by 11
“formance improvements [12], [13], since neural networks have been proven to be universal function approximators [14].”
RecNet: Deep Learning-Based OFDM Receiver with Semi-Blind Channel Estimation
Liu, C., & Arslan, T. (2020) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas45731.2020.9180878 · Cited by 12
Generalization Bounds and Algorithms for Learning to Communicate Over Additive Noise Channels
Weinberger, N. (2021) · IEEE Transactions on Information Theory · doi:10.1109/tit.2021.3129080 · Cited by 13
“The approach to this problem is coarsely categorized into either one of two types [88].”
BackgroundJoint Compensation of CFO and IQ Imbalance in OFDM Receiver: A Deep Learning Based Approach
Liu, S., Wang, T., & Wang, S. (2021) · IEEE/CIC International Conference on Communications in China (ICCC) · doi:10.1109/iccc52777.2021.9580359 · Cited by 15
“In [16], an autoencoder-based transceiver is proposed to learn the end-to-end signal processing to achieve global optimization.”
MethodRNN-Based Twin Channel Predictors for CSI Acquisition in UAV-Assisted 5G+ Networks
Shehzad, M. K., Rose, L., & Assaad, M. (2021) · Globecom · doi:10.1109/globecom46510.2021.9685990 · Cited by 12
“Motivated by the above issues and promising benefits of machine learning (ML) [11], [12], this paper focuses on employing ML in UAV-enabled wireless networks.”
BackgroundA Knowledge-Based Deep Learning Detection Scheme for Downlink SCMA Systems
Zheng, Y., Hou, X., Wang, H., Jiang, M., & Zhang, S. (2022) · IEEE Wireless Communications Letters · doi:10.1109/lwc.2022.3231602 · Cited by 11
“Deep learning (DL) techniques have produced significant success in wireless communications, such as channel estimation, channel decoding, and modulation classification [9].”
BackgroundSemantic Communications: A New Paradigm for Networked Intelligence
Zhang, Y., Wang, F., Xu, W., & Liu, C. (2022) · Machine learning for signal processing ... · doi:10.1109/mlsp55214.2022.9943480 · Cited by 11
“Besides existing source coding and channel coding algorithms, the DL-enabled joint-source-channel (JSC) coding schemes have also been demonstrated to be efficient and are widely used in existing works of semantic communications [13,14].”
MethodDesigning Learning-Based Adversarial Attacks to (MIMO-)OFDM Systems With Adaptive Modulation
Zheng, K., & Ma, X. (2023) · IEEE Transactions on Wireless Communications · doi:10.1109/twc.2023.3240970 · Cited by 14
“Deep learning has been developed in recent years [10], [11].”
Adaptive Uplink Data Compression in Spectrum Crowdsensing Systems
Zeng, Y., Calvo-Palomino, R., Giustiniano, D., Bovet, G., & Banerjee, S. (2023) · IEEE/ACM Transactions on Networking · doi:10.1109/tnet.2023.3239378 · Cited by 11
“However, recent advances in deep learning-based communication design show that in a learning-based modulation scheme, the constellation points can be asymmetric around the origin point [28], so the importance of I and Q readings can differ from each other.”
BackgroundAutomatic Modulation Classification Based on CNN, LSTM and Attention Mechanism
Wang, Z., Wang, P., & Lan, P. (2022) · IEEE International Conference on Computer and Communications (ICCC) · doi:10.1109/iccc56324.2022.10065667 · Cited by 16
“With the rapid development of deep learning in the field of image recognition and speech recognition, its application in modulation recognition has also been studied [9].”
BackgroundAuto-Encoder Based Orthogonal Time Frequency Space Modulation and Detection With Meta-Learning
Park, J., Hong, J., Kim, H.-J., & Jeong, B. J. (2023) · IEEE Access · doi:10.1109/access.2023.3271993 · Cited by 13
“Recently, deep learning based auto-encoders (AEs) have been successfully applied to solving complex problems in communication systems [25], [26], [27], [28], [29] and references therein.”
BackgroundDefending AI-Based Automatic Modulation Recognition Models Against Adversarial Attacks
Tang, H., Çatak, F. Ö., Kuzlu, M., Çatak, E., & Zhao, Y. (2023) · IEEE Access · doi:10.1109/access.2023.3296805 · Cited by 12
“early DL based AMR studies [17], the authors proposed a Convolutional Neural Network (CNN) to extract features from a modulated signal for modulation classification.”
BackgroundToward the Intelligent OFDM Receiving Method With Hybrid Knowledge and Data Driven in IoT
Wang, B., Dai, H. C., Zhou, H., & Yuan, Z. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3358418 · Cited by 11
“In terms of the end-to-end physical layer of communication, [39] proposes a new way of thinking by considering communication as an end-to-end reconfiguration optimization task using autoencoders to conjointly learn transmitter and receiver implementations and signal coding without any prior…”
BackgroundSoltani, S., Ghafourian, E., Salehi, R., Martín, D. R., & Vahidi, M. (2024) · Intelligent Automation & Soft Computing · doi:10.32604/iasc.2024.042693 · Cited by 14
A Neural Network for Estimating CQI in 5G Communication Systems
Vankayala, S. K., & Shenoy, K. G. (2020) · IEEE Wireless Communications and Networking Conference Workshops (WCNCW) · doi:10.1109/wcncw48565.2020.9124744 · Cited by 11
“These approaches have been successfully used in fields such as image processing [8], [9], speech processing [10], [11] and are currently finding applications in communication system design [12], [13].”
MethodFritschek, R., Schaefer, R. F., & Wunder, G. (2020) · IEEE International Workshop on Signal Processing Advances in Wireless Communications (SPAWC) · doi:10.1109/spawc48557.2020.9154239 · Cited by 12
Deep Learning-Based Precoder Design in MIMO Systems With Finite-Alphabet Inputs
Zhu, X., Zhang, X., Zeng, W., & Xie, J. (2020) · IEEE Communications Letters · doi:10.1109/lcomm.2020.3011978 · Cited by 13
“Thanks to the powerful prediction and data analytic ability, deep learning techniques are shown to have better performance than conventional ones in many aspects [9], including power control [10] and channel estimation [11].”
BackgroundVision-Aided Radio: User Identity Match in Radio and Video Domains Using Machine Learning
Pinho, V., de Campos, M. L. R., Garcia, L. U., & Popescu, D. G. (2020) · IEEE Access · doi:10.1109/access.2020.3038926 · Cited by 13
“In [2]–[6] some interesting use-cases of machine learning in the field of wireless communication and networking are surveyed: MAC layer protocols designed with reinforcement learning, deep neural networks for MIMO detection, UE positioning with neural networks, and others.”
Background
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.