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 1,701–1,800 of 2,995
Huang, P., Li, Y., Gong, H., & Koara, H. (2025) · Photonics · doi:10.3390/photonics12100965 · Cited by 6
Ultralightweight AMC via Robust Geometric Median Filter Pruning for Edge IoT Devices
Shi, C., Zhang, X., Tang, T., Wang, Y., Gui, G., & Jo, M. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3617238 · Cited by 3
“O’Shea and Hoydis [31] proposed a DL-based AMC approach that directly utilizes the received in-phase and quadrature (IQ) signals as input data.”
Diffusion model-based underwater acoustic communication receiver
Liu, Y., Zhao, Y., Gerstoft, P., & Ma, L. (2025) · The Journal of the Acoustical Society of America · doi:10.1121/10.0039870 · Cited by 3
AI‐Enhanced Signal Detection and Channel Estimation for Beyond 5G and 6G Wireless Networks
Daha, M. Y., Babu, B., Qureshi, R., & Hadi, M. U. (2025) · Transactions on Emerging Telecommunications Technologies · doi:10.1002/ett.70306 · Cited by 3
“Research work presented in [23] discusses the employment of DL-based end-to-end systems to achieve ultra-low latency that minimizes the complexity of designing conventional trans-mitters and receivers.”
Adaptive End-to-End Transceiver Design for NextG Pilot-Free and CP-Free Wireless Systems
Cheng, J., Chen, W., & Ai, B. (2025) · IEEE Journal on Selected Areas in Communications · doi:10.1109/jsac.2025.3648383 · Cited by 7
“A novel E2E learning paradigm has been proposed in [10], enabling joint optimization of transmitter and receiver tailored to specific channel environments.”
Yang, W., Sun, X., Kang, K., Dai, J., & Zhao, P. (2025) · Journal of King Saud University - Computer and Information Sciences · doi:10.1007/s44443-025-00382-y · Cited by 3
“The foundational work by O’Shea et al. (2017) initiated convolutional neural networks (CNNs) for signal feature extraction in modulation classification.”
Denoising 5G limited channel state information
Earle, B., Al-Habashna, A., Wainer, G., Li, X., & Xue, G. (2026) · Wireless Networks · doi:10.1007/s11276-025-04076-1 · Cited by 3
Kumar, A., Гаур, Н., & Nanthaamornphong, A. (2026) · Results in Optics · doi:10.1016/j.rio.2026.100992 · Cited by 3
Khan, F. M. A., Hallaq, M., Abou-Zeid, H., Erak, O., Waqar, O., Hassan, S. A., Alhussein, O., & Hossain, E. (2026) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2026.3682638 · Cited by 8
“Classical baselines include least-squares and MMSE channel estimators followed by linear equalizers (e.g., ZF/MMSE) or nonlinear detectors (e.g., sphere decoding) [6].”
SNR-aware semantic image transmission with deep learning-based channel estimation in fading channels
Salim, M. M., Abdalzaher, M. S., Muqaibel, A. H., Elsayed, H. A. E. A., Rabei, K. M., & Lee, I. (2026) · Physical Communication · doi:10.1016/j.phycom.2026.103116 · Cited by 6
Xu, H. J., Liang, Y., Xie, R., & Kong, Y. (2026) · Sensors · doi:10.3390/s26113609 · Cited by 4
“The pioneering work of O’Shea and Hoydis [1] demonstrated that communication systems can be modeled as differentiable end-to-end learning frameworks, thereby enabling joint optimization across traditionally separated modules.”
InfluentialA Parallel Turbo Decoder Based on Recurrent Neural Network
Zhang, L., Fu, W., Shi, F., Zhou, C., & Liu, Y. (2022) · Wireless Personal Communications · doi:10.1007/s11277-022-09779-8 · Cited by 4
“[21] used deep learning based methods for channel coding, which designed the physical layer through deep learning.”
MethodUtilizing Cognitive Radio and Reinforcement Learning for Multiple Preamble Detection in 5G-PRACH
Zehra, S. S., Qureshi, R. I., Aamir, M., & Magarini, M. (2022) · International Symposium on Wireless Personal Multimedia Communications (WPMC) · doi:10.1109/wpmc55625.2022.10014807 · Cited by 3
“Layer Applications Learning Method Algorithms PHY & MAC Channel Estimation and Prediction [3] Supervised Learning CNN Symbol Detection [4] Supervised Learning RNN, DNN Channel Coding [5] Supervised Learning RNN, DNN End-to-end Learning [6] Supervised Learning Auto-encoder Dynamic Spectrum Sensing [7] Reinforcement Lea…”
MethodState-of-the-art in PHY layer deep learning for future wireless communication systems and networks
Koufos, K., Haloui, K. E., Zhou, C., Frascolla, V., & Dianati, M. (2023) · CRC Press eBooks · doi:10.1201/9781003190691-7 · Cited by 8
Ma, C., Jiang, X., Li, P., & Liu, J. (2023) · Symmetry · doi:10.3390/sym15030676 · Cited by 3
MINE-based Geometric Constellation Shaping in AWGN Channel
Wang, Q., Ji, X., Qian, L., Liu, Z., Du, X., & Kam, P. (2023) · IEEE/CIC International Conference on Communications in China (ICCC) · doi:10.1109/icccworkshops57813.2023.10233820 · Cited by 3
“To deal with these problems, the deep learning (DL) approach is introduced into physical layer communications [17], [18].”
BackgroundSparse Adversarial Attacks against DL-Based Automatic Modulation Classification
Jiang, Z., Zeng, W., Zhou, X., Feng, P., Chen, P., Yin, S., Han, C., & Li, L. (2023) · Electronics · doi:10.3390/electronics12183752 · Cited by 3
Gao, G., Hu, X., Li, B., Wang, W., & Ghannouchi, F. M. (2023) · Electronics · doi:10.3390/electronics12224668 · Cited by 3
Application of machine learning in wireless communication
Hua, S., Wang, Q., & Xu, X. (2023) · Theoretical and Natural Science · doi:10.54254/2753-8818/12/20230452 · Cited by 3
Deep Learning-Based SC-FDMA Channel Equalization
Mohamed, M. A. (2024) · International Journal of Electrical and Electronic Engineering & Telecommunications · doi:10.18178/ijeetc.13.1.67-79 · Cited by 3
Assessing Time Offset and Classifier Impacts on Preamble-based Cross-Collection SEI
Tyler, J. H., Reising, D. R., Fadul, M. K. M., & Sartipi, M. (2024) · IEEE Consumer Communications and Networking Conference · doi:10.1109/ccnc51664.2024.10454900 · Cited by 5
“…is attributed to the concerns surrounding IoT security mentioned above and the success of Deep Learning (DL) in detecting modulation schemes [17], designing systems [17]–[21], managing spectrum [22], [23], flourishing in the presence of increasing information, and learning emitter-specific…”
BackgroundNeural Constellation Shaping and Back-Off Training for Memoryless Power Amplifiers
Kopyto, D., & Bauch, G. (2024) · IEEE Wireless Communications and Networking Conference · doi:10.1109/wcnc57260.2024.10571132 · Cited by 3
“End-to-end learning using autoencoders [5] offers the possibility to directly estimate the channel characteristics of unknown channels and optimizing the constellation with respect to the learned channel.”
BackgroundTaha, M. A., Fadul, M. K. M., Tyler, J. H., Reising, D. R., & Loveless, T. D. (2024) · EURASIP Journal on Information Security · doi:10.1186/s13635-024-00175-2 · Cited by 3
“During the past 6 to 7 years, deep learning (DL) has received a lot of attention within the SEI community due to its demonstrated successes in the areas of modulation detection [14], system design [14– 18], and spectrum management [19, 20] areas.”
BackgroundMethodWang, Z., Zhou, F., Han, X., & Zhang, L. (2024) · International Conference on Signal Processing and Intelligent Computing (SPIC) · doi:10.1109/spic62469.2024.10691420 · Cited by 3
“And DSCGAN-CNN[5] compared CGAN[8]-CNN[5] and CNN[5] and DSCGAN-CLDNN[11] compared CGAN[8]-CLDNN[11] and CLDNN[11] to select the optimal results of the network.”
MethodInfluentialHuang, L., & Chen, W. (2024) · IEEE Transactions on Communications · doi:10.1109/tcomm.2024.3487794 · Cited by 3
“The landscape of adaptive modulation and coding (AMC) technologies has seen transformative changes with the advent of deep neural networks (DNN) [17].”
Automatic Modulation Recognition Using Parallel Feature Extraction Architecture
Tang, H., Zhao, Y., Kuzlu, M., Luo, C., Çatak, F. Ö., & Wang, W. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-71467-2_18 · Cited by 3
CFGAN: A Conditional Filter Generative Adversarial Network for Signal Modulation Mode Recognition
Zhou, F., Wang, J., Zhang, L., Wang, Y., Chen, X., & Zhang, P. (2024) · Electronics · doi:10.3390/electronics14010012 · Cited by 3
“In automatic modulation recognition, O’Shea et al. first applied shallow CNN to feature extraction of wireless signals, and achieved more than 70% average accuracy [8].”
InfluentialSyed, Q. S., Hussain, S., & Bashir, I. (2025) · Advances in wireless technologies and telecommunication book series · doi:10.4018/979-8-3693-6725-4.ch009 · Cited by 3
Li, Q., Chen, B., Lei, Y., & Liang, Z. (2025) · Optics Express · doi:10.1364/oe.545577 · Cited by 3
A Novel Approach for Robust Automatic Modulation Recognition Based on Reversible Column Networks
Jing, D., Xu, T., Han, L., Yin, H., Li, L., Zhang, Y., Li, M., Pan, M., & Guo, L. (2025) · Electronics · doi:10.3390/electronics14030618 · Cited by 3
Advanced CNN-RNN Model Based Automatic Modulation Classification on Resource-Constrained End Devices
Tripathi, P., Bhola, B., Kumar, R., & Turlapaty, A. C. (2024) · IEEE Conference on Information and Communication Technology (CICT) · doi:10.1109/cict64037.2024.10899613 · Cited by 3
Adaptive Pulse Shaping and Equalization for OFDM in Time-Frequency Doubly Selective Channels
Zhang, X., Zheng, Z., Wang, S., Guo, J. H., & Fei, Z. (2025) · IEEE Transactions on Communications · doi:10.1109/tcomm.2025.3557974 · Cited by 3
“Deep learning (DL) has emerged as a pivotal role in the physical layer (PHY) of wireless communications [14], [15], enhancing the functionality of PHY modules and revolution-0090-6778 izing the architecture of communication systems through the adoption of autoencoders (AEs).”
Self-Evolving Wireless Communications: A Novel Intelligence Trend for 6G and Beyond
Qian, L., Yang, P., Wu, G., Tang, W., & Chen, Z. (2024) · International Conference on Future Communications and Networks (FCN) · doi:10.1109/fcn64323.2024.10984554 · Cited by 5
“…types of machine learning problems and algorithms in cognitive radio are summarized in [17]; deep learning is used in the physical layer in [18], treating communication as an end-to-end reconfiguration optimization task; and domain knowledge has been integrated into deep learning models in [19].”
A Novel Autoencoder-Based Design for Channel Estimation in Maritime OFDM Systems
Yang, Y., Chao, W., Ma, L., Meng, F., & Hu, Z. (2025) · Electronics · doi:10.3390/electronics14173454 · Cited by 3
Deep Randomized Distributed Function Computation (DeepRDFC): Neural Distributed Channel Simulation
Bergström, D., & Günlü, O. (2025) · IEEE International Symposium on Information Theory (ISIT) · doi:10.1109/isit63088.2025.11195680 · Cited by 9
“As in [27], we use a large batch size ( 2 14 ) and a low learning rate.”
InfluentialReliable and Reproducible Radio Experiments in FIT/CorteXlab SDR Testbed: Initial Findings
Cardoso, L. S., Oubejja, O., Villemaud, G., Risset, T., & Gorce, J.-M. (2018) · Lecture notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering · doi:10.1007/978-3-319-76207-4_19 · Cited by 3
A DNN-based Multi-Objective Precoding for Gaussian MIMO Networks
Zhang, X., & Vaezi, M. (2020) · Globecom · doi:10.1109/globecom42002.2020.9322490 · Cited by 3
“…[1] but not limited to: spectrum intelligence which manages the radio resource deceptively [2], [3], [7]; transmission intelligence which focus on reaching wireless coverage and channel throughput [4], [5]; and, network intelligence which enhances communication quality in a system-level [8].”
BackgroundMethodA Kalman-based Autoencoder Framework for End-to-End Communication Systems
Hu, B., Wang, J., Chen, X., Zhang, G., & Li, R. (2021) · IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) · doi:10.1109/pimrc50174.2021.9569618 · Cited by 3
Autoencoder Communications with Optimized Interference Suppression for NextG RAN
Davaslıoğlu, K., Erpek, T., & Sagduyu, Y. E. (2022) · IEEE Future Networks World Forum (FNWF) · doi:10.1109/fnwf55208.2022.00037 · Cited by 4
“The AE-based communications jointly trains the transmitter and receiver functionalities as two DNNs, one for the encoder and the other one for the decoder [3].”
BackgroundJoint Channel Estimation and Signal Detection using Latent Space Representations in VAE
Wong, C. Y. X., Jaward, M. H., Baskaran, V. M., Chee, C. H., & Sim, M. L. (2022) · International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS) · doi:10.1109/ispacs57703.2022.10082835 · Cited by 3
“Other advantages of Deep Learning-based algorithms over existing conventional physical layer approaches have been extensively discussed in [6].”
BackgroundDeep learning based sequence detection with natural redundancy for memory sources
Wang, Z., Yu, H., Du, J., Shen, Z., Zhu, Z., & Shen, C. (2023) · IET Communications · doi:10.1049/cmu2.12618 · Cited by 3
Wireless Communication Using Autoencoder
Harshitha, Y. S., Bhuvan, T. S., Sowkhya, G. R., N, R. K. K., & Karjigi, V. (2023) · International Conference on Smart Systems for Applications in Electrical Sciences (ICSSES) · doi:10.1109/icsses58299.2023.10199263 · Cited by 4
“Auto encoder enhances the performance, reliability, and security of wireless communication systems [14].”
BackgroundDeepIC+: Learning Codes for Interference Channels
Chahine, K., Jiang, Y., Cho, J., & Kim, H. (2023) · IEEE Transactions on Wireless Communications · doi:10.1109/twc.2023.3302617 · Cited by 7
“The proposed method is similar as the one described in [2] and is shown in Algorithm 1.”
BackgroundMethodInfluentialA Comprehensive Analysis on Unconstraint Video Analysis Using Deep Learning Approaches
Bhushanam, P. N., & Selva Kumar, S. (2023) · International Conference on Edge Computing and Applications (ICECAA) · doi:10.1109/icecaa58104.2023.10212227 · Cited by 3
“predominates in videos widely available in archives and expanding online [1].”
BackgroundEnd-to-End Learning for Free Space Optical Communication with BiLSTM-based Channel Model
Zhang, T., Zhang, L., Chen, W., Fu, M., Zhou, P., Li, S., & Wang, D. (2023) · IEEE International Symposium on Broadband Multimedia Systems and Broadcasting (BMSB) · doi:10.1109/bmsb58369.2023.10211249 · Cited by 4
“In recent years, a foundational way of interpreting the entire system as an autoencoder has been designed and introduced in radio frequency (RF) communication systems[3] and optical fiber communication systems [4].”
BackgroundAdversarial Learning for Implicit Semantic-Aware Communications
Lu, Z., Xiao, Y., Sun, Z., Li, Y., Shi, G., Chen, X., Bennis, M., & Poor, H. V. (2023) · International Conference on Communications · doi:10.1109/icc45041.2023.10279317 · Cited by 4
“The source user, for instance, can use the joint source-channel encoder proposed in [17] to encode the explicit semantic information.”
BackgroundRNN BCJR: a fully trainable version of the additive BCJR algorithm
Montorsi, G., & Ripani, B. (2023) · International Conference on Communications · doi:10.1109/icc45041.2023.10279315 · Cited by 4
“A survey on the adoption of DL on the physical layer is provided in [2].”
BackgroundiDeepRx Enabled 100 Gb/s DFT-s-OFDM Data Transmission Over 220 GHz Testbed
Qi, W., Ye, C., Korpi, D., Gong, C., Jin, Y., & Yang, T. (2023) · International Conference on Communications · doi:10.1109/icc45041.2023.10278696 · Cited by 3
“Machine learning (ML) aided radio link performance enhancements have been widely studied, including ML-based channel estimation [2], [3], channel state information (CSI) compression and feedback [4], and even the end-to-end learning by optimizing the transmitter and the receiver simultaneously [5].”
BackgroundImage Semantic Communication over Fading Channel: A Learned Broadcast Approach
Ma, K., Shao, S., & Tao, M. (2023) · IEEE International Workshop on Signal Processing Advances in Wireless Communications (SPAWC) · doi:10.1109/spawc53906.2023.10304507 · Cited by 3
“We introduce an NN as one auto-encoder (AE) [16] to optimize constellation design for the current fading channel.”
MethodDeep Learning-Based Demodulation in Impulse Noise Channels
Andersson, A., Hägglund, K., & Axell, E. (2023) · IEEE Military Communications Conference (MILCOM) · doi:10.1109/milcom58377.2023.10356304 · Cited by 3
“In particular, neural networks show promise as they allow for tailoring a specific network or system to the intended environment [10].”
BackgroundLoss Design for Single-carrier Joint Communication and Neural Network-based Sensing
Muth, C., Geiger, B., Gaviria, D. G., & Schmalen, L. (2024) · International ITG Workshop on Smart Antennas · doi:10.1109/wsa61681.2024.10511513 · Cited by 3
“ML approaches are have been studied separately for communication systems [4], [5], and in the context of radar [6], [7].”
BackgroundDan, 陈. C., Rui, 王. W., Feier, 艾. A., & Linhai, 汤. T. (2024) · Acta Optica Sinica · doi:10.3788/aos231188 · Cited by 4
Ghous, M., Nguyen, T. L., Do, T. N., & Kaddoum, G. (2024) · IEEE Access · doi:10.1109/access.2024.3427118 · Cited by 7
Scalable Syndrome-based Neural Decoders for Bit-Interleaved Coded Modulations
Rovella, G. D. B., Benammar, M., Benaddi, T., & Méric, H. (2024) · IEEE International Conference on Machine Learning for Communication and Networking (ICMLCN) · doi:10.1109/icmlcn59089.2024.10625146 · Cited by 4
“Early works in channel decoding [1], [2] promptly faced the curse of dimensionality , where the space of valid codewords was too large for a common Deep Neural Network (DNN) to explore and learn .”
BackgroundEnd-to-End Communication With Task-Driven CSI Acquisition in Multiple-Antenna Systems
Zhang, J., Guo, J., Wen, C.-K., & Jin, S. (2024) · IEEE Wireless Communications Letters · doi:10.1109/lwc.2024.3452486 · Cited by 4
“In contrast, DL simplifies the design of end-to-end (E2E) communication systems, aiming for global optimization [2], [3], [4], [5], [6], [7], [8].”
BackgroundModulation Recognition Method for Wireless Signals Based on Joint Neural Networks
Wang, X., Wang, J., & Lu, X. (2024) · IEEE Access · doi:10.1109/access.2024.3453418 · Cited by 4
Edge Convolution Graph Neural Network Assisted Power Allocation for Wireless IoT Networks
Kim, J., Cho, Y., & Kim, J. (2024) · IEEE Access · doi:10.1109/access.2024.3457805 · Cited by 3
Exploiting Side Information for Improved Online Learning Algorithms in Wireless Networks
Hanawal, M. K., & Darak, S. J. (2024) · IEEE Transactions on Wireless Communications · doi:10.1109/twc.2024.3485893 · Cited by 3
“Hence, they are preferred over dataset-dependant, compute and memory-intensive machine learning and deep learning-based approaches [5].”
Karanov, B. (2024) · Optical Fiber Technology · doi:10.1016/j.yofte.2024.104069 · Cited by 4
Rapid Online Bayesian Learning for Deep Receivers
Gusakov, Y., Simeone, O., Routtenberg, T., & Shlezinger, N. (2025) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49660.2025.10888154 · Cited by 4
“While pre-training DNN on multiple channel distribution through joint learning [4], [5] can handle some variations, it often leads to degraded performance, learning a non-coherent receiver.”
Doğan-Tusha, S., Bouanani, F. E., & Qaraqe, M. (2025) · IEEE Transactions on Machine Learning in Communications and Networking · doi:10.1109/tmlcn.2025.3567370 · Cited by 5
Kobovich, A., Yaakobi, E., & Weinberger, N. (2025) · IEEE International Symposium on Information Theory (ISIT) · doi:10.1109/isit63088.2025.11195301 · Cited by 5
“A prominent method in this domain is end-to-end learning , introduced in [19].”
Yapay Sinir Ağları İle Kablosuz Yerel Alan Ağlarında Veri Trafiği Optimizasyonu
Koçak, C., & Karakurt, H. B. (2019) · Journal of Polytechnic · doi:10.2339/politeknik.443219 · Cited by 3
Machine Learning in Wireless Communication: A Survey
Sharma, N. V., & Yadav, N. S. (2019) · Advances in wireless technologies and telecommunication book series · doi:10.4018/978-1-5225-7458-3.ch007 · Cited by 3
Implementation of conventional communication system in deep learning
Satyanarayana, P., Devi., C., V, Sowjanya, P. S., Babu, S., Kumar, N. S., & S, M. N. V. (2017) · International Journal of Engineering & Technology · doi:10.14419/ijet.v7i1.1.10835 · Cited by 3
Unsupervised Machine Intelligence for Automation of Multi-Dimensional Modulation
Ko, Y., & Choi, J. (2019) · IEEE Communications Letters · doi:10.1109/lcomm.2019.2932417 · Cited by 3
“tions [10]–[13].”
Deep Learning Based Optical Camera Communications
Walter, I., Khadr, M. H., & Elgala, H. (2019) · OSA Advanced Photonics Congress · doi:10.1364/sppcom.2019.spth3e.2 · Cited by 3
Application of deep learning for power control in the interference channel: a RNN-based approach
Deng, Z., Sang, Q., Pan, Y., & Xin, Y. (2019) · ACM Conference on Research in Adaptive and Convergent Systems (RACS) · doi:10.1145/3338840.3355668 · Cited by 4
“The brief introduction of the applications of deep learning in wireless physical layer, such as modulation recognition, channel decoding and sequence detection, was proposed in [5].”
BackgroundA multi-level network for radio signal modulation classification
Peng, L., Qu, W., Zhao, Y. L., & Wu, Y. (2019) · International Conference on Artificial Intelligence, Information Processing and Cloud Computing (AIIPCC) · doi:10.1145/3371425.3371482 · Cited by 3
Deep Learning Aided Friendly Coexistence of WiFi and LTE in Unlicensed Bands
Gu, H., Wang, Y., Hong, S., & Gui, G. (2019) · International Conference on Wireless Communications and Signal Processing (WCSP) · doi:10.1109/wcsp.2019.8928141 · Cited by 3
“Generally, the CNN model [27] is a neural network containing multiple hidden layers, each of which has several twodimensional planes composed of several independent neurons.”
BackgroundDeep Ensemble Learning: A Communications Receiver Over Wireless Fading Channels
Al-Baidhani, A., & Fan, H. H. (2019) · IEEE Global Conference on Signal and Information Processing · doi:10.1109/globalsip45357.2019.8969302 · Cited by 3
“Lately, some DL algorithms have been used to model wireless communication systems [2-8], including an end-toend AutoEncoder (AE) [2-3], a receiver under fast-fading [4], MIMO detection [5-6], channel decoding [7-9].”
MethodA Learning Approach to Cooperative Communication System Design
Lu, Y., Cheng, P., Chen, Z., Mow, W. H., & Li, Y. (2020) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp40776.2020.9054093 · Cited by 3
“The network parameters used are similar to that of [2].”
BackgroundMethodInfluentialWang, J., Wang, Y., Li, W., Gui, G., Gacanin, H., & Adachi, F. (2020) · IEEE Vehicular Technology Conference · doi:10.1109/vtc2020-fall49728.2020.9348790 · Cited by 4
Deep Learning Based Detection and Channel Tracking for MIMO Systems
Yang, J., & Li, H. (2020) · International Conference on Computer and Communication Systems (ICCCS) · doi:10.1109/icccs49078.2020.9118499 · Cited by 4
“In the field of signal detection [7], channel estimation [8], and modulation recognition [9], data-driven methods using deep learning achieve better performance.”
MethodUncooperative RF Direction Finding with I/Q Data
Shanthakumar, V. A., Banerjee, C., Mukherjee, T., & Pasiliao, E. L. (2020) · International Conference on Information System and Data Mining (ICISDM) · doi:10.1145/3404663.3404668 · Cited by 3
“Deep networks have also been used for tasks such as modulation detection and for various other tasks in the physical layer [20, 21].”
MethodAn Efficient ADMM-Aided Deep Learning-Based Signal Detector for Uplink Massive MIMO
Huang, H., Cioffi, J. M., & Hashemi, S. A. (2020) · IEEE International Conference on Communications workshops · doi:10.1109/iccworkshops49005.2020.9145413 · Cited by 3
“This paper addresses the challenge of retaining low complexity but nearing optimal detector performance through deep-learning methods, motivated by other big data uses to solve complicated wireless transmission problems in [5] and [6].”
BackgroundTrainable Projected Gradient Detector for Sparsely Spread Code Division Multiple Access
Takabe, S., Yamauchi, Y., & Wadayama, T. (2020) · IEEE International Conference on Communications workshops · doi:10.1109/iccworkshops49005.2020.9145193 · Cited by 3
“Recently, deep unfolding proposed by Gregor and LeCun [11] has attracted great interests as another DL-based approach [12] in addition to an end-to-end approach [13].”
MethodPhysical Adversarial Attacks Against Deep Learning Based Channel Decoding Systems
Babu, S. A., & Ameer, P. M. (2020) · IEEE Region 10 Symposium (TENSYMP) · doi:10.1109/tensymp50017.2020.9230666 · Cited by 3
“For example, [2] describes end-to-end learning of the autoencoder-based communication systems.”
BackgroundA Novel Neural Network Denoiser for BCH Codes
Zhu, H., Cao, Z., Zhao, Y., & Li, D. (2020) · IEEE/CIC International Conference on Communications in China (ICCC) · doi:10.1109/iccc49849.2020.9238971 · Cited by 3
“The physical layer of communication systems is also a typical application scenario of DL [7].”
BackgroundYang, L., Yin, X., & Rodríguez-Piñeiro, J. (2020) · International Conference on Wireless Communications and Signal Processing (WCSP) · doi:10.1109/wcsp49889.2020.9299800 · Cited by 3
“Some DL frameworks that have been proposed in wireless communication systems, including detection of communication signals under various conditions [7], [8], modulation recognition [9], channel encoding and decoding [10], [11], etc.”
BackgroundLabel-free Optimization for Passive Beamforming in IRS-assisted MISO Systems
Chu, L., He, H., Pei, L., Yu, W., & Qiu, R. C. (2020) · IEEE International Conference on Communication Technology (ICCT) · doi:10.1109/icct50939.2020.9295770 · Cited by 3
“With superior nonlinear mapping capability, the deep learning learning based solutions [15], [16] have recently drawn increasing attention for wireless communication problems, such as channel-independent detections [11], [12], [17], noise-resisting decoding [13], etc.”
BackgroundDeep Learning-Based Bootstrap Detection Scheme for Digital Broadcasting System
Kang, T.-H., Lee, W., Baek, M.-S., Bae, B., & Song, H. (2021) · IEEE Access · doi:10.1109/access.2021.3051906 · Cited by 3
“Recently, deep learning technology has received attention in communication engineering area [14]–[21].”
BackgroundDigital signal modulation recognition method based on high-order cumulants and wavelet transform
Wang, A., & Liu, P. (2020) · International Conference on Algorithms, Computing and Artificial Intelligence (ACAI) · doi:10.1145/3446132.3446423 · Cited by 3
“At present, deep learning and neural networks have been studied by many scholars, and have been gradually applied to the field of signal modulation recognition, but the research about this still needs to be carried out [9-12].”
BackgroundSVM‐based online learning for interference‐aware multi‐cell mmWave vehicular communications
Li, D., Zhu, J., Zhao, H., Wang, X., & Jiang, R. (2021) · IET Communications · doi:10.1049/cmu2.12138 · Cited by 4
“For example, the autoencoders are used in [10] to model an end-to-end communication system comprising the encoding, channel and decoding blocks for MIMO interference channels.”
An Introduction to Neural Networks in SCMA
Madhura, K., Rukmini, M. S. S., & Raut, R. D. (2021) · Wireless Personal Communications · doi:10.1007/s11277-021-08222-8 · Cited by 3
TC-MIMONet: A Learning-based Transceiver for MIMO Systems with Temporal Correlations
Chen, C., Wang, Z., Mao, Y., Wu, H., Bai, B. B., & Zhang, G. (2021) · IEEE Vehicular Technology Conference · doi:10.1109/vtc2021-spring51267.2021.9448981 · Cited by 3
“The end-to-end learning framework for communication systems was first proposed for additive white Gaussian noise (AWGN) channels [6].”
BackgroundMethodInfluentialCooperative Spectrum Sensing using DQN in CRN
Moneesh, M., Tejaswi, T., Yeshwanth, T., Harshitha, M., & Chakravarthy, G. (2021) · ICST Transactions on Mobile Communications and Applications · doi:10.4108/eai.14-7-2021.170290 · Cited by 3
“Importance of Deep Learning for physical layer is discussed by the authors in [14].”
BackgroundSuper-resolution reconstruction algorithm for aerial image data management based on deep learning
Bing, X., & Niu, F. (2021) · Distributed and Parallel Databases · doi:10.1007/s10619-021-07356-9 · Cited by 3
Doppler Spread Estimation Based on Machine Learning for an OFDM System
Yoon, E., Kwon, S., Yun, U., & Kim, S. Y. (2021) · Wireless Communications and Mobile Computing · doi:10.1155/2021/5586029 · Cited by 3
Generative Adversarial Network-based Channel Modeling for Free-Space Optical Communication
Chen, W., Wang, D., Wang, D., Song, Y., Li, J., & Zhang, M. (2021) · Opto-Electronics and Communications Conference (OECC) · doi:10.1364/oecc.2021.s4b.2 · Cited by 3
Deep Neural Networks for Multicarrier Waveforms Classification in UAV Networks
Zerhouni, K., Amhoud, E. M., & Noubir, G. (2021) · International Symposium on Wireless Communication Systems · doi:10.1109/iswcs49558.2021.9562225 · Cited by 3
Neural Network-Assisted Robust Symbol Detection Under Intersymbol Interference
Yang, J., Du, Q., & Jiang, Y. F. (2021) · IEEE/CIC International Conference on Communications in China (ICCC) · doi:10.1109/iccc52777.2021.9580317 · Cited by 4
Parameter Estimation and Classification via Supervised Learning in the Wireless Physical Layer
McClintick, K. W., Wernsing, G. M., Ferreira, P. V. R., & Wyglinski, A. M. (2021) · IEEE Access · doi:10.1109/access.2021.3128813 · Cited by 3
“O’Shea and Hoydis [125] explored the use of a CNN demodulator and a Variational Auto-Encoder (VAE) transceiver with a focus on error control coding.”
MethodResultDeep Learning-Based Codebook Designs for Generalized Space Shift Keying Systems
Huang, D., Jiang, X., Lee, I., & Hai, H. (2021) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2021.3128693 · Cited by 3
Deng, X., Bian, X., & Li, M. (2021) · Sensors · doi:10.3390/s22010257 · Cited by 4
“On the one hand, various deepearning methods have recently been used to solve the problem of the physicalayer of wireless communication, and valuable results have been obtained [23].”
MethodAutoencoder-based MIMO Communications with Learnable ADCs
Ye, W., Ren, B., Wang, Z., Wu, H., Wu, H., Bai, B. B., & Zhang, G. (2021) · IEEE International Conference on Communication Technology (ICCT) · doi:10.1109/icct52962.2021.9658065 · Cited by 4
Sakano, J., Mori, Y., & Hasegawa, H. (2021) · Photonics in Switching and Computing (PSC) · doi:10.1364/psc.2021.tu5a.6 · Cited by 3
Application of Neural Network to Demodulate SEFDM Signals
Semenova, A. I., & Zavjalov, S. V. (2022) · Lecture notes in computer science · doi:10.1007/978-3-030-97777-1_34 · Cited by 3
Experimental Demonstration of Learned Pulse Shaping Filter for Superchannels
He, Z., Song, J., Häger, C., Amat, A. G. I., Wymeersch, H., Andrekson, P. A., Karlsson, M., & Schröder, J. (2022) · Optical Fiber Communication Conference · doi:10.1364/ofc.2022.w2a.33 · Cited by 3
Kim, H. (2022) · Springer eBooks · doi:10.1007/978-3-030-95041-5_8 · Cited by 3
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.