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 2,001–2,100 of 2,995
LUT-based end-to-end optical communication system for constellation shaping
Jiang, R., Jia, T., Ding, X., Wang, K., Cao, H., Wang, H., Shang, C., Wang, Z., & Wu, C. (2025) · Optics Express · doi:10.1364/oe.563856 · Cited by 2
Automatic Modulation Recognition via Pruned LSTM-GRU with Multi-Head Attention
Zhou, J., & Wan, S. (2025) · International Conference on Communication Networks and Machine Learning (CNML) · doi:10.1145/3728199.3728213 · Cited by 2
“[2] Cheng et al. demonstrated that higher-order statistics of IQ signals enhance recognition performance[3].”
Medusa: A Guide to Build an Agile RF Fingerprinting Testbed [Application Notes]
Nieves, A., & Wyglinski, A. M. (2025) · IEEE Microwave Magazine · doi:10.1109/mmm.2025.3555759 · Cited by 2
“Other researchers implement a protocol-specific testbed that generates fingerprinting datasets where the researcher can easily access the hardware measurements and perform modifications [12], [15], [24].”
Khedhri, N., & Najar, M. (2025) · International Wireless Communications and Mobile Computing Conference (IWCMC) · doi:10.1109/iwcmc65282.2025.11059447 · Cited by 2
Input Optimization in the Composite DNA Storage Channel
Kobovich, A., & Weinberger, N. (2025) · IEEE Journal on Selected Areas in Information Theory · doi:10.1109/jsait.2025.3595005 · Cited by 7
“A prominent method in this domain is end-to-end learning , introduced in [45].”
Reconfigurable Deep Learning Augmented Channel Estimation for Vehicular Networks on System on Chip
Haq, S. A. U., Darak, S. J., & Gizzini, A. K. (2025) · IEEE Transactions on Consumer Electronics · doi:10.1109/tce.2025.3600481 · Cited by 2
“Various research works have explored the complete replacement of wireless PHY with DL [26], [27], [28].”
Semantic Communication via Rate Distortion Perception Bottleneck
Zhao, Z., & Wang, C. (2024) · International Conference on Big Data, Artificial Intelligence and Internet of Things Engineering (ICBAIE) · doi:10.1109/icbaie63306.2024.11116905 · Cited by 3
“During the exponential growth of Artificial Intelligence (AI), existing communication technologies have been significantly enhanced [1], [2], and has greatly accelerated developments in the field of telecommunications [3].”
Deep Multi-Modal Neural Receiver for 6G Vehicular Communication
Saleem, O., Alfaqawi, M., Merdrignac, P., Bensrhair, A., & Ribouh, S. (2025) · International Conference on Communications · doi:10.1109/icc52391.2025.11160713 · Cited by 4
“In contrast, an AI-based solution has the ability to provide efficient end-to-end performance in complex communication scenarios [9].”
Optimizing CRC Polynomial Selection using Reinforcement Learning for Improved GRAND Decoding
Cirpan, A. M., Görçin, A., Hökelek, İ., & Yılmaz, F. (2025) · IEEE International Black Sea Conference on Communications and Networking · doi:10.1109/blackseacom65655.2025.11193897 · Cited by 2
“…and extend-able to Deep Q-Learning for enhanced scalability. improvements in BER performance [1], [4] and decoding reliability [2], thereby paving the way for future high-performance communication systems designed to fulfill the challenging requirements of URLLC and mMTC [1], [2], [4], [12], [13].”
Differential semantic communications for salient object transmission in 6G multimedia systems
Taimori, A., Sellathurai, M., & Ratnarajah, T. (2025) · IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS) · doi:10.1109/infocomwkshps65812.2025.11152869 · Cited by 3
Optimization of transmission signal by artificial intelligent
Rashag, H. F., & Ali, M. H. (2019) · International Journal of Advances in Applied Sciences · doi:10.11591/ijaas.v8.i4.pp290-292 · Cited by 2
A Fiber-Optic Channel Modeled Through BiLSTM Technique
Song, Y., Wang, D., Zhang, M., & Cui, Q. (2019) · International Conference on Optical Communications and Networks (ICOCN) · doi:10.1109/icocn.2019.8933906 · Cited by 2
“There have been some researches put DL into wireless communication to optimize transmitter and receiver [3], but for the model of channel, they just represent it by an additive noise layer.”
BackgroundA Physical Layer Security Communication Mechanism Based on End-to-End Learning
Wu, H., Sun, Z., & Zhou, X. (2019) · IEEE Advanced Information Management, Communicates, Electronic and Automation Control Conference (IMCEC) · doi:10.1109/imcec46724.2019.8984189 · Cited by 2
“In [9], multiple inputs multiple outputs (MIMO) end-to-end communication systems were introduced.”
BackgroundRun-Length Limited Decoding for Visible Light Communications: A Deep Learning Approach
Le, D.-D., Nguyen, D.-P., Tran, T.-H., & Nakashima, Y. (2019) · Asia-Pacific Conference on Communications (APCC) · doi:10.1109/apcc47188.2019.9026486 · Cited by 2
“Furthermore, several studies proposed replacing the structure of conventional communication systems by a deep learning-based end-to-end communication system [12], [17], [18].”
BackgroundOn Training Neural Network Decoders of Rate Compatible Polar Codes via Transfer Learning
Lee, H., Seo, E. Y., Ju, H., & Kim, S. (2020) · Entropy · doi:10.3390/e22050496 · Cited by 2
“There have been a number of DL approaches [5,15,16] to physical layer communications.”
BackgroundImplementation of Deep Learning Based Method for Optimizing Spatial Diversity MIMO Communication
Rohmatillah, M., Pramono, S. H., & Asyari, R. A. I. (2020) · Indonesian Journal of Electrical Engineering and Informatics (IJEEI) · doi:10.11591/ijeei.v8i2.1247 · Cited by 2
“…interesting results of machine learning implementation in a communication system are paper titled “An Introduction to Deep Learning for the Physical Layer” [12] and “Deep Learning-Based Communication Over the Air” [13] which introduce deep learning as an end-to-end system in SISO communication.”
BackgroundOptimization of Fiber Optics Communication Systems via End-to-End Learning
Jovanovic, O., Jones, R. T., Gaiarin, S., Yankov, M. P., Ros, F. D., & Zibar, D. (2020) · International Conference on Transparent Optical Networks · doi:10.1109/icton51198.2020.9203560 · Cited by 2
“In this talk, we review our previous efforts on overcoming the nonlinear Kerr effect occurring in fiber-optic communication using end-to-end learning [21].”
BackgroundDeep Learning Based Secure MISO Transmission
Fulwani, Y., Thapar, S., & Sood, N. (2020) · International Conference on Computing, Communication and Security (ICCCS) · doi:10.1109/icccs49678.2020.9277242 · Cited by 3
Zang, Y., Zhao, Y. K., & Chen, X. (2020) · Asia Communications and Photonics Conference · doi:10.1364/acpc.2020.m4a.276 · Cited by 2
Design of Energy Modulation Massive SIMO Transceivers via Machine Learning
Lan, M., Huang, J., Zhang, H., & Huang, C. Z. (2020) · Globecom · doi:10.1109/globecom42002.2020.9348239 · Cited by 2
“Along a different line of research, the authors in [10] represented the transmitter and the receiver as a deep NN to perform the end-to-end design of the communication systems, and treated the information transmission as a classification problem.”
BackgroundAutoencoder-bank based design for adaptive channel-blind robust transmission
Safi, H., Akbari, M., Vaezpour, E., Parsaeefard, S., & Shubair, R. M. (2021) · EURASIP Journal on Wireless Communications and Networking · doi:10.1186/s13638-021-01929-z · Cited by 2
Wang, D., Zhou, Y., Chen, J., & Zhang, C. (2021) · International Conference on Intelligent Computing and Signal Processing (ICSP) · doi:10.1109/icsp51882.2021.9408757 · Cited by 2
“The input tensor is [1,2,128], which means 128 pairs of I/Q signals sampling points, and the output is 11 signal modulation modes.”
BackgroundData-Driven Precoded MIMO Detection Robust to Channel Estimation Errors
Mayouche, A., Martins, W. A., Chatzinotas, S., & Ottersten, B. (2021) · IEEE Open Journal of the Communications Society · doi:10.1109/ojcoms.2021.3079643 · Cited by 4
“end-to-end link from the Tx to the Rx [7], [8]; the idea is to design a communication system to jointly optimize”
BackgroundMultidimensional Codebook Design Using Deep Learning Techniques for Rayleigh Fading Channels
Fu, X., da Silva, B. F., & Ruyet, D. L. (2021) · IEEE Wireless Communications Letters · doi:10.1109/lwc.2021.3089024 · Cited by 4
“Both MDCs designed by AE techniques, including model-aware [8] and model-free [9], and conventional approaches, lower projection based MDC design [10] and grouped linear constellation-precoding (GLCP) with rotation [5] are studied.”
BackgroundMethodInfluentialDeep Learning Opportunities for Resource Management in Cognitive Radio Networks
Maharaj, B. T., & Awoyemi, B. S. (2020) · Springer eBooks · doi:10.1007/978-3-030-64653-0_10 · Cited by 3
Compression Algorithm for End-to-End Communication using CNN
Liang, Y., Lam, C., & Ng, B. K. (2021) · IEEE International Conference on Computer and Communications (ICCC) · doi:10.1109/iccc54389.2021.9674292 · Cited by 2
Small Sample Electromagnetic Signal Recognition based on Time Series Data Augmentation
Wang, X., & Lin, Y. (2021) · International Conference on Dependable Systems and Their Applications (DSA) · doi:10.1109/dsa52907.2021.00049 · Cited by 2
“O'Shea and hoydis proposed new applications for deep learning of network physical layer, and carried out theoretical verification in [7].”
BackgroundLong Boosted Memory Algorithm for Intelligent Spectrum Sensing in 5G and Beyond Systems
Kansal, P., Gangadharappa, M., & Kumar, A. (2022) · Journal of Network and Systems Management · doi:10.1007/s10922-022-09652-w · Cited by 2
Jiang, L. (2022) · Mobile Information Systems · doi:10.1155/2022/3356213 · Cited by 2
“He applied deep learning to the sending and receiving of the network and achieved great success [3].”
Machine learning for short reach optical fiber systems
Karanov, B., Bayvel, P., & Schmalen, L. (2022) · Elsevier eBooks · doi:10.1016/b978-0-32-385227-2.00010-3 · Cited by 2
Power Intelligent Terminal Intrusion Detection Based on Deep Learning and Cloud Computing
Li, T., Zhao, H., Tao, Y., Huang, D., Yang, C., & Xu, S. (2022) · Computational Intelligence and Neuroscience · doi:10.1155/2022/1415713 · Cited by 4
“There are only edge connections between visible layer units and hidden layer units, as shown in Figure 6 [18].”
BackgroundProbabilistic Shaping for Multidimensional Signals with Autoencoder-based End-to-end Learning
Liu, X., Darwazeh, I., Zein, N., & Sasaki, E. (2022) · IEEE Wireless Communications and Networking Conference · doi:10.1109/wcnc51071.2022.9771910 · Cited by 2
“Deep learning technology has been applied to communication systems by approximating and optimising the transmitter and receiver functions using neural networks [19].”
MethodImproved Non-Uniform Constellations for Non-Binary Codes Through Deep Reinforcement Learning
Klaimi, R., Weithoffer, S., & Nour, C. A. (2022) · IEEE International Workshop on Signal Processing Advances in Wireless Communications (SPAWC) · doi:10.1109/spawc51304.2022.9834022 · Cited by 2
Trainable Communication Systems Based on the Binary Neural Network
Che, B., Li, X., Chen, Z., & He, Q. (2022) · Frontiers in Communications and Networks · doi:10.3389/frcmn.2022.878170 · Cited by 3
“An autoencoder-based communication system regards the entire physical layer, i.e., “ transmitter-channel-receiver, ” as an end-to-end data reconstruction task (O ’ Shea and Hoydis, 2017; Zhu et al., 2019; Ye et al., 2020).”
MethodPolymorphic wireless receivers
Restuccia, F., & Melodia, T. (2022) · Communications of the ACM · doi:10.1145/3547131 · Cited by 3
“6, 11, 12, 18, 19 The seminal work by O’Shea et al. 12 proposed ConvNets-based to address the issue.”
BackgroundModulation Recognition of MIMO Systems Based on Dimensional Interactive Lightweight Network
Aer, S., Zhang, X., Wang, Z., & Wang, K. (2022) · KSII Transactions on Internet and Information Systems · doi:10.3837/tiis.2022.10.014 · Cited by 2
Current Trends on Deep Learning-aided Channel Coding
Dang, T.-M., Cho, E., & Kim, S. (2022) · International Conference on Information and Communication Technology Convergence (ICTC) · doi:10.1109/ictc55196.2022.9952535 · Cited by 2
“Numerous studies have been conducted to enhance the effectiveness of NNs in replacing signal processing tasks [20], especially the proposal of model-driven NNs in [21] dramatically thrives as an alternative method in solving communication system problems.”
BackgroundMachine Learning Based End-to-End Constellation Training for Communication Systems
Lin, P.-C. (2022) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.23919/apsipaasc55919.2022.9980231 · Cited by 2
Compressed Lightweight Network for Edge Intelligence AMC of CPM Signals
Aer, S., Dong, H., Zhang, X., & Wang, K. (2022) · International Conference on 6G Communications and IoT Technologies (6GIoTT) · doi:10.1109/6giott57212.2022.00027 · Cited by 2
“Besides, some efficient network structures are also proposed such as Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) network, and Convolutional Long Short-Term Memory (CLDNN) network.”
MethodInfluentialZhu, J., Bai, C., Zhu, Y., Lu, X., & Wang, K. (2023) · Nano Communication Networks · doi:10.1016/j.nancom.2023.100456 · Cited by 5
Baldini, G., & Bonavitacola, F. (2023) · European Conference on Antennas and Propagation · doi:10.23919/eucap57121.2023.10133651 · Cited by 2
“In recent times, machine learning [1] and deep learning algorithms [2] have been applied with success to the problem of channel classification.”
MethodHuang, T., Xin, F., & Wang, J. (2023) · International Conference on Electronic Technology and Information Science (ICETIS) · doi:10.1117/12.2682404 · Cited by 2
“Classification algorithms are important in modulation recognition [14], and the main classification algorithms include neural network [31-33], support vector machine, decision tree, etc.”
BackgroundBöhm, S., & König, H. (2023) · Computer Communications · doi:10.1016/j.comcom.2023.07.013 · Cited by 2
Exploiting ResNeXt with Convolutional Shortcut for Signal Modulation Classification at Low SNRs
Gu, Y., & Zhou, X. (2023) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn54540.2023.10191354 · Cited by 2
“was proposed in [31]: (1) The small observation window and”
BackgroundEvaluation of Neural Demappers for Trainable Constellation in an End-to-End Communication System
Islam, N., & Shin, S. (2023) · International Conference on Ubiquitous and Future Networks · doi:10.1109/icufn57995.2023.10200350 · Cited by 2
“Owing to that, [2] first proposed physical layer (PHY) of wireless communication system as a Deep Learning (DL) architecture, known as the autoencoder (AE) [3].”
BackgroundMethodXiao, H., Hua, Y., Chen, Q., & Zhou, L. (2023) · International Joint Conference on Information and Communication Engineering (JCICE) · doi:10.1109/jcice59059.2023.00012 · Cited by 2
Tagliapietra, G., & Iannacci, J. (2023) · Lecture notes in electrical engineering · doi:10.1007/978-981-99-4495-8_31 · Cited by 2
Neural Adjusted Min-Sum Decoding for LDPC Codes
Yu, H.-C., Zhao, M.-M., Lei, M., & Zhao, M.-J. (2023) · IEEE Vehicular Technology Conference · doi:10.1109/vtc2023-fall60731.2023.10333372 · Cited by 2
“In recent years, with the vigorous development of machine learning technologies, deep learning has become a popular technique to solve signal processing tasks for communication systems [8].”
BackgroundHuang, C., Li, Z., Jia, J., Xing, S., Wang, C., Dong, B., Shi, J., Chi, N., & Zhang, J. (2023) · Asia Communications and Photonics Conference · doi:10.1109/acp/poem59049.2023.10369802 · Cited by 2
“Unlike traditional approaches where each module is independently optimized, the E2E method employs artificial neural networks (ANN) to globally optimize signal processing steps [4].”
MethodDeep Learning Based Cooperative MIMO Systems for Wireless Body Area Networks
Bui, T., Tran, X. N., & Phan, A. H. (2024) · REV Journal on Electronics and Communications · doi:10.21553/rev-jec.352 · Cited by 2
“Hence, in several studies [11, 36], communication systems based on Autoencoder (AE) propose the adoption of a Radio Transformer Network (RTN) served the purpose of channel estimation and mitigating the impact of fading channels to enhance the system’s over-all performance.”
BackgroundMethodInfluentialChen, S., & Tan, C. W. (2024) · doi:10.1145/3662004.3663552 · Cited by 2
Machine Learning Applied to an RF Communication Channel
McCaskey, M., Kukura, E., Corrigan, R., Bhasin, K. B., & Chelmins, D. (2018) · IEEE National Aerospace and Electronics Conference (NAECON) · doi:10.1109/naecon.2018.8556708 · Cited by 2
“Previous work has shown that this end-to-end communication system can be modeled with a specific type of neural network called an auto-encoder [17].”
BackgroundHigher-order Modulation Signal Detection Scheme Using Sequential Clustering
Jung, K., & Wang, H. (2019) · The Journal of Korean Institute of Information Technology · doi:10.14801/jkiit.2019.17.3.87 · Cited by 2
MIMO Pre-Coder Design Based on Weighted Euclidean Distance Maximization in Interference Channel
Luo, P. H., Pang, B., & Kim, Y. J. D. (2018) · IEEE Vehicular Technology Conference · doi:10.1109/vtcfall.2018.8690722 · Cited by 2
“In [11], authors suggest bringing the deep learning into the communication network, in particular to the interference channel.”
BackgroundSymbol Detection and Channel Estimation using Neural Networks in Optical Communication Systems
Aboutaleb, A., Fatnassi, W., Soltani, M., & Rezki, Z. (2019) · International Conference on Communications · doi:10.1109/icc.2019.8761449 · Cited by 3
“recognition, natural language processing and recommender systems [4], and their ability to capture communication problems [5]–[8], we propose a deep learning architecture for detection and estimation for OWC that does not rely on a channel model.”
BackgroundA Neural Network Simulator of a Nonstationary Medium in an Adaptive Data Transmission System
Nevzorov, A. A., Orlov, A. A., & Stankevich, D. A. (2019) · Technical Physics Letters · doi:10.1134/s1063785019070241 · Cited by 2
Efficient co-design partitioning of WLANs on SoC-based SDRs
Akeela, R., & Elziq, Y. (2019) · Microsystem Technologies · doi:10.1007/s00542-019-04641-7 · Cited by 2
“For example, the authors in O’Shea and Hoydis (2017) propose the use of Convolutional Neural Networks (CNNs) for modulation classifications in order to enhance the performance of wireless communication.”
BackgroundAutomatic Wireless Signal Classification: A Neural-Induced Support Vector Machine-Based Approach
Wahla, A. H., Chen, L., Wang, Y., & Chen, R. (2019) · Information · doi:10.3390/info10110338 · Cited by 2
“Recently, these deep learning-based approaches such as Convolutional Neural Networks (CNNs) [21,22], stack autoencoders [23] and deep belief n tworks [ 4] have b en applied for modulation classificat on as well as performing physical layer ow-level tasks in wireless communication.”
BackgroundLearning to Communicate with Limited Co-design
Sahai, A., Sanz, J., Subramanian, V., Tran, C., & Vodrahalli, K. (2019) · Allerton Conference on Communication, Control, and Computing · doi:10.1109/allerton.2019.8919749 · Cited by 3
“Early work in this area [7], [8], where gradients are shared among agents, demonstrated the success of training a channel auto-encoder using supervised learning when a stochastic model of the channel is known.”
BackgroundMarseet, A. A., & Elganimi, T. Y. (2019) · IEEE Western New York Image and Signal Processing Workshop (WNYISPW) · doi:10.1109/wnyipw.2019.8923057 · Cited by 2
“In [7], the authors proposed a novel deep learning model for the physical layer and applied it to Single Input Single Output (SISO) communication systems.”
BackgroundJoint Estimation and Detection for MIMO-STBC System Based on Deep Neural Network
Zhang, M., Fan, M. Y., & Chen, M. (2019) · IEEE Middle East and North Africa Communications Conference (MENACOMM) · doi:10.1109/menacomm46666.2019.8988569 · Cited by 2
“Recently, the DNN is popular in the design of MIMO physical layer [4], such as channel estimation [5] and signal detection [6-9].”
MethodAuto-encoder Based Analog Network Coding for Free-Space Optical Two-Way Communications
Li, J., & Xiong, J. (2019) · IEEE International Conference on Computer and Communications (ICCC) · doi:10.1109/iccc47050.2019.9064418 · Cited by 2
“In [5], a point to point communication system is designed as an autoencoder network which is optimized for end-to-end performance.”
BackgroundAutomatic Modulation Classification and SNR Estimation Based on CNN in Physical-layer Network Coding
Wang, X., He, Y., Sun, Y., & Zhan, Y. (2020) · Wireless and Optical Communications Conference · doi:10.1109/wocc48579.2020.9114949 · Cited by 2
Multi-Label and Concatenated Neural Block Decoders
Leung, C. T., Motani, M., & Bhat, R. V. (2020) · IEEE International Symposium on Information Theory (ISIT) · doi:10.1109/isit44484.2020.9174106 · Cited by 3
“Moreover, the authors in [21] point out that using the Sigmoid activation at the nodes of the output layer can potentially help in avoiding an exponential increase in the number of nodes in the output layer with k , however, it does not be explored further.”
BackgroundImproving CSI Prediction Accuracy with Deep Echo State Networks in 5G Networks
Pecorella, T., Fantacci, R., & Picano, B. (2020) · Sensors · doi:10.3390/s20226475 · Cited by 4
“Recently, the machine learning (ML) techniques have gained attention for their application in numerous research fields such as channel prediction [10], detection [11], modulation [12], and so on.”
BackgroundA Low-complexity Neural BP Decoder with Network Pruning
Han, S., & Ha, J. (2020) · International Conference on Information and Communication Technology Convergence (ICTC) · doi:10.1109/ictc49870.2020.9289525 · Cited by 2
“In the communication system, inspired by the successes of deep learning, many efforts have been being made to solve long standing problems by taking advantage of deep learning [1].”
BackgroundPhysical Layer Security of Cooperative Network with Machine Learning-based Relay Selection
Liu, F., Wang, Y., & Li, H. (2019) · IEEE International Symposium on Electromagnetic Compatibility · doi:10.1109/isemc48616.2019.9295968 · Cited by 2
“In order to reduce computational burden, machine learning has been gradually introduced into wireless communication [6]–[8], such as resource allocation, modulation recognition and antenna selection.”
BackgroundMethodLi, Z., Wang, Q., Liu, W., Xu, Q., Xing, Z., & Li, Y. (2020) · IET Communications · doi:10.1049/iet-com.2020.0687 · Cited by 2
“Meanwhile, a deep learning structure [11, 12] is introduced into MR, and the convolutional neural network (CNN) model is used for modulation classification without expert feature extraction, resulting in a good performance in terms of efficiency and accuracy [13].”
MethodDeep Learning versus High-order Recurrent Neural Network based Decoding for Convolutional Codes
Teich, W. G., Liu, R., & Belagiannis, V. (2020) · Globecom · doi:10.1109/globecom42002.2020.9348117 · Cited by 2
“O’Shea and Hoydis [1] interpret a communication system as an autoencoder.”
BackgroundDeep-Learning Based Ship-Radiated Noise Suppression for Underwater Acoustic OFDM Systems
Atanackovic, L., Lampe, L., & Diamant, R. (2020) · Global Oceans 2020: Singapore – U.S. Gulf Coast · doi:10.1109/ieeeconf38699.2020.9389436 · Cited by 2
“The field of communications has seen extensive success in using DL-based solutions, demonstrating measurable and advantageous results to traditional analytical approaches in areas such as physical layer transmitter and receiver design [7], channel estimation and signal detection [8], and communication networks [9].”
BackgroundLearning Optimal Lattice Codes for MIMO Communications
Amorós, L., & Pitkänen, M. J. (2021) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp39728.2021.9414237 · Cited by 2
“Applying deep learning (DL) to the physical layer was first considered in [10] in 2017.”
MethodDNN-Based Cross Layer Resource Allocation of NOMA-OFDMA Video Transmissions without Post Processing
Tseng, S., Tsai, C.-H., & Hsu, Y.-T. (2021) · International Conference on Consumer Electronics · doi:10.1109/icce50685.2021.9427663 · Cited by 2
“Applications of deep learning includes radio resource allocation [1] and auto-encoder in the physical (PHY) layer [2].”
BackgroundSelf-Attention-Based Real-Time Signal Detector for Communication Systems With Unknown Channel Models
Chen, L., & Sun, L. (2021) · IEEE Communications Letters · doi:10.1109/lcomm.2021.3082708 · Cited by 4
Alahassa, N. K.-A. (2021) · International Conference on Artificial Intelligence and Computer Science Technology (ICAICST) · doi:10.1109/icaicst53116.2021.9497838 · Cited by 2
“Binary signals, for example, may be interpreted as binary or one-hot vectors, modulated (complex) symbols or integers, and the optimal representation can depend on the architecture of the Neural Network, learning purpose, and loss function, among other variables [23].”
BackgroundMachine Learning for Reliable Communication and Improved Tracking of Dynamical Systems
Vedula, K. (2020) · Cited by 2
An Empirical Analysis of the Progress in Wireless Communication Generations
Luo, K., Dang, S., Zhang, C., Shihada, B., & Alouini, M. (2020) · EAI International Conference on Mobile and Ubiquitous Systems: Computing, Networking and Services (MobiQuitous) · doi:10.1145/3448891.3448894 · Cited by 3
“As a result of these investments, a number of ground-breaking communication technologies were invented and implemented to significantly ease our daily communications nowadays, including multiple-input and multiple-output (MIMO) system [31], orthogonal frequency-division multiplexing (OFDM) [28], cooperative relaying s…”
BackgroundCheng, C., Chen, W., & Li, A. (2021) · Wireless Communications and Mobile Computing · doi:10.1155/2021/8647298 · Cited by 2
“Using distributed technology [6, 7], a cloud computing and big data platform is built based on deep learning-based legal quality and legal knowledge cultivation and promotion analysis [8] to solve the problem of massive big data storage and computing performance [9]], to provide a good research platform for the cultiv…”
Learning the Impedance Entanglement for Wireline Data Communication
Tonello, A. M., Letizia, N. A., & Piante, M. D. (2021) · International Balkan Conference on Communications and Networking (BalkanCom) · doi:10.1109/balkancom53780.2021.9593164 · Cited by 2
“Machine learning and in particular neural network-based receivers have recently found application in communications [8], [9], [10].”
BackgroundA Study of Neural Network Receivers in OFDM Systems Subject to Memoryless Impulse Noise
Tseng, D., & Lin, C.-S. (2021) · Wireless and Optical Communications Conference · doi:10.1109/wocc53213.2021.9603179 · Cited by 2
Scalable Machine Learning Algorithms to Design Massive MIMO Systems
Martí, D. G., Badini, D., Donno, D. D., & Widmer, J. (2021) · ACM International Conference on Modeling, Analysis and Simulation of Wireless and Mobile Systems (MSWiM) · doi:10.1145/3479239.3485672 · Cited by 2
“Recently, the first design of an end-to-end communication system using deep learning models was proposed [10].”
BackgroundMethodInfluentialCohen, A., & Derevyanko, S. (2021) · IEEE International Conference on Microwaves, Communications, Antennas and Electronic Systems (COMCAS) · doi:10.1109/comcas52219.2021.9629004 · Cited by 2
Deep Learning Based Low Complexity Symbol Detection and Modulation Classification Detector
Hao, C., Dang, X., Li, S., & Chenghua, W. A. N. G. (2022) · IEICE Transactions on Communications · doi:10.1587/transcom.2021ebp3148 · Cited by 2
Received Path Power Prediction for Millimeter-wave Using Machine Learning
Yamazaki, T., Shimaoka, K., Ohno, K., Johannsen, U., & de Groot, S. H. (2022) · International Conference on Consumer Electronics · doi:10.1109/icce53296.2022.9730204 · Cited by 2
Wireless Channel Recognition by CNN
Zhu, S., Xiong, J., Yang, C., Yu, P., Ren, W., & Li, Y. (2022) · IEEE International Symposium on Broadband Multimedia Systems and Broadcasting (BMSB) · doi:10.1109/bmsb55706.2022.9828733 · Cited by 2
Wu, F., Wu, J., Kong, Y., Yang, C., Yang, G., Shu, H., Carrault, G., & Senhadji, L. (2022) · Neurocomputing · doi:10.1016/j.neucom.2022.09.088 · Cited by 3
Psychological Motivation of Athletes’ Physical Training Based on Deep Learning Model
Yi-ning, Y. (2022) · International Transactions on Electrical Energy Systems · doi:10.1155/2022/1962461 · Cited by 2
RetractedTurbo AI, Part IV: Estimating Uplink Channels for Ultra High Mobility with Sparse Pilots
Chen, Y., Mohammadi, J. A., Wesemann, S., & Wild, T. (2022) · IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) · doi:10.1109/pimrc54779.2022.9978084 · Cited by 2
“…move, and exhibited certain similarity to the decoding, detection and estimation problems in PHYsical (PHY) layer of wireless communications [4]–[5], pioneers immediately started exploiting AI / ML to create novel concepts, e.g. autoencoders to realize end-to-end wireless communications,…”
BackgroundDeep Learning Aided Multi-Level Transmit Power Recognition in Cognitive Radio Networks
Tan, Z., Wang, D., Liu, Q., Li, Z., Zhang, N., & Abdel‐Raheem, E. (2023) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2023.3235738 · Cited by 2
“To further enhance the performance of traditional communication systems, DL technology is employed in the wireless physical layer [34], [35], [36] and achieve outstanding performance.”
MethodGeometrically shaped multi-dimensional modulation formats designed by deep learning
Naka, A., & Komatsu, M. (2023) · IEICE Communications Express · doi:10.1587/comex.2022xbl0176 · Cited by 2
Overcoming an Evasion Attack on a CNN Model in the MIMO-OFDM Wireless Communication Channel
Komeylian, S., Paolini, C. P., & Sarkar, M. (2023) · Lecture notes in networks and systems · doi:10.1007/978-981-99-1203-2_7 · Cited by 2
Machine Learning-Aided Efficient Decoding of Reed–Muller Subcodes
Jamali, M. V., Liu, X., Makkuva, A. V., Mahdavifar, H., Oh, S., & Viswanath, P. (2023) · IEEE Journal on Selected Areas in Information Theory · doi:10.1109/jsait.2023.3298362 · Cited by 2
“Motivated by these observations, we leverage the recent advances in channel coding via machine/deep learning [18], [19], [20], [21], [22], [23],”
MethodAEVBComm: an intelligent communication system based on $$\beta$$-VAE
Hemadri, R. V., Rayaluru, A., Pandya, R. J., & Iyer, S. (2024) · CSI Transactions on ICT · doi:10.1007/s40012-024-00401-9 · Cited by 2
“The AE’s system [18] has improved significantly in [19] with the introduction of radio transformer networks [14], which helps in actualizing communication knowledge into DL models.”
Joint Quantizer Optimization based on Neural Quantizer for Sum-Product DecoderPreprint
Wadayama, T., & Takabe, S. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1804.06002 · Cited by 2
Machine learning in indoor positioning and channel prediction systems
Zhu, Y. (2018) · UVic’s Research and Learning Repository (University of Victoria) · Cited by 2
Online unsupervised deep unfolding for massive MIMO channel estimationPreprint
Magoarou, L. L., & Paquelet, S. (2020) · HAL Portal Artxiker (Hindustan Aeronautics Limited (India)) · Cited by 2
Configuration Learning in Underwater Optical LinksPreprint
Zhao, X., Qi, Z., & Pompili, D. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2008.01221 · Cited by 2
“e number of novel outputs [16], [17]. The existing works on machine learning and deep learning in physical-layer signal processing research cover on major research problems including signal detection [18]–[22], channel estimation [23]–[25], Channel State Information (CSI) compression [26], precoding [27], and interfer…”
BackgroundOnline unsupervised deep unfolding for MIMO channel estimationPreprint
Magoarou, L. L., & Paquelet, S. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2004.14615 · Cited by 5
“Machine learning holds promise for wireless communications (see [11], [12] for exhaustive surveys).”
BackgroundConclusions and future directions
Hong, Y., Thaj, T., & Viterbo, E. (2022) · Elsevier eBooks · doi:10.1016/b978-0-32-385028-5.00017-7 · Cited by 2
Automatic Modulation Classification for MIMO Systems via Deep Learning and Zero-Forcing EqualizationPreprint
Gui, G., Wang, Y., Yin, Y., Wang, J., Sun, J., Sari, H., Gui, J., Adachi, F., & Gacanin, H. (2020) · doi:10.36227/techrxiv.11634369 · Cited by 2
Channel Estimation using Very Deep Super Resolution Network for Fast Fading Communication SystemsPreprint
Manivannan, G., Jeyakumar, P., Tharanitharan, N. M., Malar, E., & Muthuchidambaranathan, P. (2022) · Research Square · doi:10.21203/rs.3.rs-2396453/v1 · Cited by 2
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.