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 1,601–1,700 of 2,995
Evaluating Deep Neural Networks to Classify Modulated and Coded Radio Signals
Cheong, P. S., Camelo, M., & Latré, S. (2019) · Lecture notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering · doi:10.1007/978-3-030-05490-8_17 · Cited by 5
Deep Learning Decoder for MIMO Communications with Impulsive Noise
Delgado, O., & Labeau, F. (2020) · IEEE Consumer Communications and Networking Conference · doi:10.1109/ccnc46108.2020.9045329 · Cited by 6
“Similarly the authors in [13], [14] proposed the design of an end-to-end communication system and the use of auto-encoders to jointly learn transmitter/receiver implementations and signal encoding/decoding processes.”
BackgroundRecurrent Neural Network Channel Estimation Using Measured Massive MIMO Data
Faghani, T., Shojaeifard, A., Wong, K., & Aghvami, A. H. (2020) · IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) · doi:10.1109/pimrc48278.2020.9217192 · Cited by 9
“Artificial Intelligence (AI) has in recent years entered the classical signal processing layers of wireless communications [1].”
BackgroundHow to choose a neural network architecture? – A modulation classification example
Warrier, A. N., & Amuru, S. (2020) · IEEE 5G World Forum (5GWF) · doi:10.1109/5gwf49715.2020.9221167 · Cited by 5
“some problems in wireless communications such as modulation classification [4], end-to-end transmitter-receiver design [5], low complexity channel decoders [6], among others.”
BackgroundA Deep Learning-based Approach to 5G-New Radio Channel Estimation
Zimaglia, E., Riviello, D. G., Garello, R., & Fantini, R. (2021) · Joint European Conference on Networks and Communications & 6G Summit (EuCNC/6G Summit) · doi:10.1109/eucnc/6gsummit51104.2021.9482426 · Cited by 4
Rana, V., & Chou, R. A. (2021) · IEEE Information Theory Workshop (ITW) · doi:10.1109/itw48936.2021.9611401 · Cited by 5
Stochastic Channel Modeling for Deep Neural Network-aided Sparse Code Multiple Access Communications
Li, D., Jia, M., Zhang, L., Guo, Q., & Gu, X. (2021) · IEEE Vehicular Technology Conference · doi:10.1109/vtc2021-fall52928.2021.9625321 · Cited by 4
Neural Network Generative Models for Radio Frequency Data
Carmack, J. M., Bhatia, A., Robinson, J., Majewski, J., & Kuzdeba, S. (2021) · IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON) · doi:10.1109/uemcon53757.2021.9666725 · Cited by 5
PAPR Reduction Scheme for Localized SC-FDMA Based on Deep Learning
Lü, H., Zhou, Y., Liu, Y., Li, R., & Cao, N. (2022) · Lecture notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering · doi:10.1007/978-3-030-93398-2_60 · Cited by 4
Deep learning Achievements and Opportunities in Domain of Electronic Warfare Applications
Ammar, M. A., Abdel-Latif, M. S., Badran, K. M., & Hassan, H. A. B. (2021) · International Conference on Intelligent Computing and Information Systems (ICICIS) · doi:10.1109/icicis52592.2021.9694245 · Cited by 4
Recognition of Human Body Feature Changes in Sports Health Based on Deep Learning
Jiao, C. (2022) · Computational and Mathematical Methods in Medicine · doi:10.1155/2022/1736350 · Cited by 4
“It can be called a classic of the year [27].”
BackgroundRetractedPerformances prediction in Wireless Sensor Networks: A survey on Deep learning based-approaches
Eljakani, Y., Boulouz, A., Salah, M. B., & Hachemy, S. E. (2022) · ITM Web of Conferences · doi:10.1051/itmconf/20224301010 · Cited by 5
“In [19] the authors train an auto-encoder as a CNN, and they test on a single end-to-end communication system, multipletransmitter/receiver system, and radio transformer networks.”
Wireless Virtual Network Embedding Algorithm Based on Deep Reinforcement Learning
Gao, Q., Lyu, N., Miao, J., & Pan, W. (2022) · Electronics · doi:10.3390/electronics11142243 · Cited by 4
Ma, Y. (2022) · Journal of Environmental and Public Health · doi:10.1155/2022/9714591 · Cited by 4
“Experimental results show that the classification accuracy of deep learning based on convolutional neural networks is more competitive compared to traditional schemes relying on expert features [7].”
BackgroundRetractedA Review: Deep Learning Aided Channel Estimation Techniques for Wireless Communication System
Ahmed, M. H., Jamel, T. M., & Khazaal, H. F. (2022) · Iraqi International Conference on Communication and Information Technologies (IICCIT) · doi:10.1109/iiccit55816.2022.10010434 · Cited by 4
“In [8], it is addressed how DL might be used in physical layers.”
BackgroundDesign and simulation of precision marketing recommendation system based on the NSSVD++ algorithm
Liu, Y., & Zhang, W. (2023) · Neural Computing and Applications · doi:10.1007/s00521-023-08302-9 · Cited by 4
“…to be scored, you can get: sim cos (cid:4) adjust ð x ; y Þ sim pearson ð x ; y Þ When the non-zero items of the commodity item vector are dense, the Pearson correlation similarity calculation is more preferred, and the improved cosine similarity calculation formula is applied on the contrary [14].”
Demonstrating Deep Learning driven BPSK Demodulation using Software-Defined Radios
Ahmad, A., & Agarwal, S. (2023) · International Conference on Communication Systems & Networks (COMSNETS) · doi:10.1109/comsnets56262.2023.10041280 · Cited by 4
“Therefore, we also examine the efficacy ML concepts in signal demodulation to meet our requirements [1].”
BackgroundHybrid Deep Learning for Channel Estimation and Power Allocation for MISO-NOMA System
Gaballa, M., Abbod, M., & Al-nasur, S. J. (2022) · IEEE Future Networks World Forum (FNWF) · doi:10.1109/fnwf55208.2022.00070 · Cited by 6
“In channel approximation phase, the suggested DL scheme based on LSTM includes 4 layers, where every layer is provided with numerous neurons, and the weighted sum will be the input to a nonlinear function [12].”
BackgroundMethodSemantic Communications: A Paradigm Whose Time Has Come
Kutay, E., & Yener, A. (2022) · IEEE International Conference on Collaboration and Internet Computing (CIC) · doi:10.1109/cic56439.2022.00020 · Cited by 5
“In [18], a communication system is modeled as an autoencoder.”
BackgroundMachine Learning Decoder for 5G NR PUCCH Format 0
Yerrapragada, A. K., Jeeva Keshav, S., Gautam, A., & Ganti, R. K. (2023) · National Conference on Communications · doi:10.1109/ncc56989.2023.10067950 · Cited by 4
“The choice of training SNR is far from trivial [15].”
Ma, M., Xu, Y., Wang, Z., Fu, X. G., & Gui, G. (2023) · Drones · doi:10.3390/drones7060391 · Cited by 4
Wang, C., Tao, L., Li, Z., Shi, J., Chi, N., & Zhang, J. (2023) · International Conference on Optical Communications and Networks (ICOCN) · doi:10.1109/icocn59242.2023.10236121 · Cited by 4
“Currently, it is commonly used in end-to-end communication systems as a means of gradient propagation to train transceivers, thereby achieving optimal overall performance from the receiver to the transmitter [2].”
BackgroundAwad, S. D., Sali, A. B., Al-Wani, M. M., Al-Saegh, A. M., Mandeep, J. S., & Abdullah, R. S. A. R. (2022) · Computer Networks · doi:10.1016/j.comnet.2023.110022 · Cited by 4
Rahdari, F., & Sheikh‐Hosseini, M. (2023) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2023.107344 · Cited by 4
Feng, X., El‐Hajjar, M., Xu, C., & Hanzo, L. (2023) · IEEE Open Journal of Vehicular Technology · doi:10.1109/ojvt.2023.3334822 · Cited by 4
“Deep Learning (DL) has been attracting increasing attention in wireless communications [38], [39].”
BackgroundIoT Technology Recognition Using Deep Clustering
Kim, H., Shahid, A., Fontaine, J., Poorter, E. D., Moerman, I., & Nam, H. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3338560 · Cited by 4
“Recently, deep learning has shown superior performance in the fields of signal recognition, image classification, and speech recognition [6], [7].”
BackgroundLiu, Z., Liu, X., Xiao, S., Yang, W., & Hu, W. (2023) · IEEE photonics journal · doi:10.1109/jphot.2023.3344184 · Cited by 4
A Network Coding Scheme Transmitting Specific Data and Universal Data Based on Deep Learning
Zhang, G., Zhang, D., & Li, W. (2024) · IEEE Systems Journal · doi:10.1109/jsyst.2023.3332456 · Cited by 4
“Deep learning method may partly alleviate the above-mentioned problems [9], [10], [11], [12], [13].”
MethodA Comparative Study of Deep Leaning-Aided Relay Selection Protocols in Cooperative Communication
Akdemir, B., Karabulut, M. A., & İlhan, H. (2024) · Wireless Personal Communications · doi:10.1007/s11277-024-11269-y · Cited by 4
“New applications for the physical layer are presented in [16] and proposed as a communication block system as an auto-encoder.”
Deep Learning-Based Radio Estimation Using a Semi-Automatically Created Indoor Building Information
Pyo, C., Sawada, H., & Matsumura, T. (2024) · IEEE Wireless Communications and Networking Conference · doi:10.1109/wcnc57260.2024.10571287 · Cited by 4
“Deep learning techniques have been employed to optimize various physical layer components, such as modulation and coding schemes, power control, beamforming, channel estimation, and equalization [12] − [13].”
MethodChen, C.-H., Karanov, B., van Houtum, W. J., Wu, Y. X., Young, A., & Alvarado, A. (2024) · IEEE Wireless Communications and Networking Conference · doi:10.1109/wcnc57260.2024.10570860 · Cited by 4
“The utilization of machine learning algorithms and neural networks (NN) in communication systems has drawn significant attention, primarily due to their remarkable capacity for universal function approximation [1], [2].”
BackgroundA Machine Learning Approach for Simultaneous Demapping of QAM and APSK Constellations
Gansekoele, A., Balalsoukas-Slimming, A., Brusse, T., Hoogendoorn, M., Bhulai, S., & van der Mei, R. (2024) · IEEE International Conference on Machine Learning for Communication and Networking (ICMLCN) · doi:10.1109/icmlcn59089.2024.10624805 · Cited by 5
“First, one could replace the entire sender-receiver pipeline with a DNN [4]–[6].”
MethodAboutaleb, A., Torabi, M. A., Belzer, B. J., & Sivakumar, K. (2024) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2024.3457014 · Cited by 4
Antón‐Haro, C., & Mestre, X. (2019) · IEEE International Conference on Communications workshops · doi:10.1109/iccw.2019.8756991 · Cited by 4
Deep Learning-Based Decoding for Constrained Sequence Codes
Cao, C., Li, D., & Fair, I. J. (2018) · IEEE Globecom Workshops (GC Wkshps) · doi:10.1109/glocomw.2018.8644185 · Cited by 4
Machine Learning based Detections for mmWave Two-Hop MIMO Systems using One-Bit Transceivers
Kim, D., & Lee, N. (2019) · IEEE International Workshop on Signal Processing Advances in Wireless Communications (SPAWC) · doi:10.1109/spawc.2019.8815577 · Cited by 4
“Beyond the nonlinearity induced by one-bit ADCs, in [9], [10], several deep learning based joint detection-decoding algorithms were also proposed for linear/nonlinear channels.”
BackgroundDeep Learning-based Automatic Modulation Recognition Algorithm in Internet of Things
Wang, Y., Gui, G., Huang, H., Wang, J., Yin, Y., Zhou, T., Zhao, Y., Sheng, H., & Zhu, X. (2019) · IEEE International Conference on Electronic Information and Communication Technology (ICEICT) · doi:10.1109/iceict.2019.8846277 · Cited by 4
“In paper [10], authors propose that CNN, which is trained by in-phase and quadrature component of received signals, can be applied into AMR.”
BackgroundGolay Layer: Limiting Peak-to-Average Power Ratio for OFDM-based Autoencoders
Şahin, A., & Matolak, D. W. (2020) · International Conference on Communications · doi:10.1109/icc40277.2020.9149257 · Cited by 4
Deep Learning-Based Encoder for One-Bit Quantization
Balevi, E., & Andrews, J. G. (2019) · Globecom · doi:10.1109/globecom38437.2019.9013923 · Cited by 4
“The closest study to our paper is [10], which proposes an autoencoder to learn transmit constellations such as M-QAM.”
BackgroundLearning Secured Modulation With Deep Adversarial Neural Networks
Mohammed, H. H., & Saha, D. (2020) · IEEE Vehicular Technology Conference · doi:10.1109/vtc2020-fall49728.2020.9348833 · Cited by 4
“Deep neural network has been used in physical layer communication [9] to learn optimal constellation points without prior mathematical formulation.”
Accelerated Learning-Based MIMO Detection through Weighted Neural Network Design
Mohammad, A., Masouros, C., & Andreopoulos, Y. (2020) · International Conference on Communications · doi:10.1109/icc40277.2020.9148726 · Cited by 4
“One of the earliest attempts is the work of O’Shea et al. [5], who implemented unsupervised learning using an auto-encoder as a continuation of end-to-end learning of previous attempts [3].”
BackgroundMethodCSIscan: Learning CSI for Efficient Access Point Discovery in Dense WiFi Networks
Sankhe, K., Jaisinghani, D., & Chowdhury, K. R. (2020) · IEEE International Conference on Network Protocols (ICNP) · doi:10.1109/icnp49622.2020.9259360 · Cited by 8
“More specific use cases of CNN in the PHY layer of wireless communications has been seen in Modulation Classification [53]–[55], Protocol Identification [56] and RF Fingerprinting [57]–[61].”
MethodSun, Z., Li, J., & Fan, J. (2021) · IEEE Access · doi:10.1109/access.2021.3049950 · Cited by 5
“Among these approaches, many of them are designed by taking radio signals as the input of deep learning network, such as channel decoding, channel estimation and modulation classification [2], [3].”
BackgroundEnd-to-end Physical Layer Communication using Bi-directional GRUs for ISI Channels
Wu, H., Zhang, Y., Zhao, X., Zhu, N., & Coates, M. (2020) · IEEE Globecom Workshops (GC Wkshps) · doi:10.1109/gcwkshps50303.2020.9367429 · Cited by 4
“In order to address the drawbacks of DNN-based autoen-coders [8], such as the curse of codeword dimensionality and high training complexity, fully CNN-based autoencoders [9], [10] were proposed to process information frames of arbitrary lengths.”
BackgroundOpening the Black Box of Deep Neural Networks in Physical Layer Communication
Liu, J., Zhao, H., Ma, D., Mei, K., & Wei, J. (2022) · IEEE Wireless Communications and Networking Conference · doi:10.1109/wcnc51071.2022.9771675 · Cited by 4
Adversarial Learning for Cross Layer Security
Mohammed, H. H., & Saha, D. (2021) · ACM Workshop on Wireless Security and Machine Learning (WiseML) · doi:10.1145/3468218.3469043 · Cited by 4
“An autoencoder (AE) model is used in [19] to learn the different modulation orders without any prior mathematical formulation.”
MethodSecure wireless communication via adversarial machine learning: A Priori vs. A Posteriori
Seo, J., Park, S., & Kang, J. (2021) · ICT Express · doi:10.1016/j.icte.2021.06.005 · Cited by 4
Soleymani, M., Santamarı́a, I., & Schreier, P. J. (2021) · IEEE Access · doi:10.1109/access.2021.3094976 · Cited by 4
“An alternative promising way to obtain fast solutions for complicated optimization problems can be to employ deep neural networks (DNNs) [5], [6], [24], [25].”
BackgroundIntelligent Communication: Application of Deep Learning at the Physical Layer of Communication
Fan, Y., Dan, D., Li, Y., Wang, Z., & Liu, Z. (2021) · IEEE Advanced Information Management, Communicates, Electronic and Automation Control Conference (IMCEC) · doi:10.1109/imcec51613.2021.9482326 · Cited by 4
On Equivalence of Neural Network Receivers
Careem, M. A. A., Dutta, A., & Thawdar, N. (2021) · International Conference on Communications · doi:10.1109/icc42927.2021.9500703 · Cited by 4
“So far, much of the effort has been focused on replacing individual blocks [9], [10] or the end-to-end chain [11], [12] in classical transceivers using “black-box” neural networks to achieve a target error rate for a specific dataset.”
BackgroundFaking Signals to Fool Deep Neural Networks in AMC via Few Data Points
Ma, H., Yang, S., He, G., Wu, R., Hao, X., Li, T., & Feng, Z. (2021) · IEEE Access · doi:10.1109/access.2021.3106704 · Cited by 4
“showed that the end-to-end autoencoder communication systems proposed in [18] are vulnerable to adversarial attacks in an additive white Gaussian noise channel environment.”
BackgroundIntroduction to Deep Learning Possibilities in Communication Systems
Žeger, I., & Šišul, G. (2021) · International Symposium ELMAR (ELMAR) · doi:10.1109/elmar52657.2021.9550825 · Cited by 5
“Therefore, the interest in a joint ( end-to-end ) realization of the communication systems has risen [16–19].”
BackgroundMethodInfluentialDesign of Codebooks Optimized for Noncoherent Transmission over Fast Fading Channel
Sergienko, A. B., & Apalina, P. V. (2021) · International Symposium on Problems of Redundancy in Information and Control Systems (REDUNDANCY) · doi:10.1109/redundancy52534.2021.9606452 · Cited by 4
Measurement-based Indoor Beam Alignment Utilizing Deep Learning
Makara, Á. L., Csathó, B. T., Csurgai‐Horváth, L., & Horváth, B. P. (2021) · International Conference on Electrical, Computer, Communications and Mechatronics Engineering (ICECCME) · doi:10.1109/iceccme52200.2021.9590951 · Cited by 4
Adaptive Uplink Data Compression in Spectrum Crowdsensing Systems
Zeng, Y., Calvo-Palomino, R., Giustiniano, D., Bovet, G., & Banerjee, S. (2021) · IEEE International Symposium on Dynamic Spectrum Access Networks · doi:10.1109/dyspan53946.2021.9677075 · Cited by 4
“Although I and Q readings are usually equally important in traditional modulation schemes, recent advances in deep learning based communication design show that in a learning based modulation scheme, the constellation points can be asymmetric around the origin point [10], so the importance of I and Q readings can diff…”
BackgroundXiong, X., Wu, H., Wu, R., Guo, X., Yang, X., & Luo, R. (2022) · Signal Image and Video Processing · doi:10.1007/s11760-021-02127-6 · Cited by 4
“10a, the distinction between AM-DSB and WBFM is difficult by the small observation window and low information rate with frequent silence between the words of the data [29, 30].”
BackgroundXie, C., Zhang, L., & Zhong, Z. (2022) · Wireless Communications and Mobile Computing · doi:10.1155/2022/6185482 · Cited by 4
Neural Network-Based Transceiver Design for VLC System over ISI Channel
Li, L., Zhu, Z.-R., & Zhang, J. (2022) · Photonics · doi:10.3390/photonics9030190 · Cited by 4
“…parts of communication systems, such as pilot design [7], channel estimation and detection [8], are replaced by a learned efficient neural network (NN); in the other respect, the end-to-end learning of the whole communication system creates a new paradigm for joint optimization of transceivers [9].”
Intelligent Radio Frequency Identification for URLLC in Industrial IoT Networks
Zhang, T., Ren, P., Xu, D., & Ren, Z. (2022) · Symmetry · doi:10.3390/sym14040801 · Cited by 5
“Then, online legitimacy testing is provided to check the information transmission block [35].”
Non-Orthogonal Physical Layer (NOPHY) Design towards 5G Evolution and 6G
Hou, X., Wenjia, L. I. U., Juan, L. I. U., Wang, X., Chen, L., Kishiyama, Y., & Asai, T. (2022) · IEICE Transactions on Communications · doi:10.1587/transcom.2021ebp3192 · Cited by 4
[Retracted] Data Analysis Deep Learning Research on Spatiotemporal Preposition Construction Network
Guo, W., & Jiang, Z. (2022) · Wireless Communications and Mobile Computing · doi:10.1155/2022/9416747 · Cited by 4
“He shows the application of convolutional neural networks on raw samples for modulation classification, which is more accurate than traditional schemes that rely on expert features, but he did not explain why the convolutional neural network is more accurate nor did he verify the accuracy of the convolutional neural n…”
RetractedThe Construction of Sports Health Management Model Based on Deep Learning
Meng, J., & Wang, S. (2022) · Applied Bionics and Biomechanics · doi:10.1155/2022/5194665 · Cited by 4
“However, the new applications they introduced lack representation [4].”
BackgroundRetractedDesign of Financial Risk Control Model Based on Deep Learning Neural Network
Yang, D., Ma, H., Chen, X., Liu, L., & Lang, Y. (2022) · Computational Intelligence and Neuroscience · doi:10.1155/2022/5842039 · Cited by 7
“On this basis, the back-propagation algorithm of the multilayer NNs is improved, that is, the BP NNs model [9, 10].”
MethodRetractedAdaptive Data Augmentation for Deep Receivers
Raviv, T., & Shlezinger, N. (2022) · IEEE International Workshop on Signal Processing Advances in Wireless Communications (SPAWC) · doi:10.1109/spawc51304.2022.9833983 · Cited by 4
Song, Y. S., & Zhao, L. (2022) · Computational Intelligence and Neuroscience · doi:10.1155/2022/7992045 · Cited by 4
“It achieved reliable precision in contrast to conventional scenarios that depend on specialist characteristics [3].”
BackgroundRetractedCommunication Knowledge Aided Neural Network for OFDM Receiver in Terahertz Band
Chakraborty, S., Saha, D., & Thawdar, N. (2022) · International Conference on Communications · doi:10.1109/icc45855.2022.9839246 · Cited by 5
“[27] has represented DNN based physical layer model, with fading channel represented by a generative adversarial network (GAN).”
BackgroundSignal detection and modulation recognition based on convolutional neural networks
Luo, Y., Xu, M., Wu, Z., & Yin, Z. (2022) · IEEE International Conference on Artificial Intelligence and Computer Applications (ICAICA) · doi:10.1109/icaica54878.2022.9844646 · Cited by 4
Frunză, A., Choqueuse, V., Morel, P., & Azou, S. (2022) · IEEE Open Journal of the Communications Society · doi:10.1109/ojcoms.2022.3201130 · Cited by 4
“To benefit from the advantages of both model-based and ML/DL techniques, different approaches that aim to insert the system knowledge into an ML/DL network have been proposed [36], [37], [38], [39].”
MethodLow-Complexity Adaptive DPD: From Online Optimization to Meta-Learning
Falempin, A., Doré, J., Zayani, R., & Strinati, E. C. (2022) · IEEE Transactions on Broadcasting · doi:10.1109/tbc.2022.3204229 · Cited by 5
“Besides, there has been a growing interest in the use of machine learning techniques for non-linear physical layer design of wireless systems [9], [10].”
BackgroundA Novel CNN-based Autoencoder with Channel Feedback for Intelligent Maritime Communications
Lin, B., Han, X., Wu, N., Li, J., Wang, H., & Shao, S. (2022) · IEEE/CIC International Conference on Communications in China (ICCC) · doi:10.1109/iccc55456.2022.9880798 · Cited by 5
“The key idea is to represent transmitter, channel, and receiver as a deep neural network that can be trained as an autoencoder [7].”
BackgroundBoehm, S., & Koenig, H. (2022) · ACM International Conference on Modeling, Analysis and Simulation of Wireless and Mobile Systems (MSWiM) · doi:10.1145/3551659.3559057 · Cited by 4
Low Latency Energy-Efficient Neural Decoders for Block Codes
Leung, C. T., Bhat, R. V., & Motani, M. (2022) · IEEE Transactions on Green Communications and Networking · doi:10.1109/tgcn.2022.3222961 · Cited by 5
“The new wave of deep learning has motivated researchers in designing channel decoders, using the deep learning models [1], [16].”
BackgroundBlack-box Adversarial Examples against Intelligent Beamforming in 5G Networks
Zolotukhin, M., Miraghaie, P., Zhang, D., Hämäläinen, T., Wang, K., & Dunderfelt, M. (2022) · IEEE Conference on Standards for Communications and Networking · doi:10.1109/cscn57023.2022.10051010 · Cited by 4
“In the context of the next-generation wireless networks, potential AI/ML applications proposed in recent scientific studies include but not limited to multiple-input-multipleoutput (MIMO) channel estimation [5], automatic modulation classification [6] and channel coding [7], beamforming [8] and power allocation [9], a…”
BackgroundBoehm, S., & Koenig, H. (2023) · Annual Conference on Wireless On-demand Network Systems and Services (WONS) · doi:10.23919/wons57325.2023.10062069 · Cited by 5
“Specific solutions concern, for example, the channel coding in [26] or the modulation classification based on Neural Networks in [24].”
BackgroundA one-dimension convolutional neural network based interference classification method
Duan, C., Feng, S., Hu, H., & Luo, Z. (2023) · Physical Communication · doi:10.1016/j.phycom.2023.102075 · Cited by 4
A Deep Learning Based Receiver for Wireless Communications Systems With Unknown Channel Models
Li, M., Sun, L., & Wang, Y. (2023) · International Conference on Computer and Communication Systems (ICCCS) · doi:10.1109/icccs57501.2023.10150798 · Cited by 4
“In [12]–[16] , DL techniques were applied in the field of physical layer communications, where DL-based communication systems can achieve extremely competitive performance with acceptable implementation complexity.”
MethodAn End-to-End Trainable Power Line Communication System
Koh, M.-S. (2023) · International Conference on Machine Learning Technologies (ICMLT) · doi:10.1145/3589883.3589922 · Cited by 4
“Ever since the first idea for an end-to-end trainable communication system in [9], it has been successfully extended to optical fiber channels [10], and wireless channels [11-14], where both transmitter and receiver are trained to minimize various special loss functions e.g., a special negative…”
BackgroundMethodInfluentialMachine learning integration in Communication system for efficient selection of signals
Bhardwaj, A., Rebelli, S., Gehlot, A., Pant, K., Gonzáles, J. L. A., & Firos, A. (2023) · International Conference on Advance Computing and Innovative Technologies in Engineering (ICACITE) · doi:10.1109/icacite57410.2023.10182417 · Cited by 5
“The application of machine learning (ML) in communication networks has received a lot of interest recently [8], [9].”
BackgroundA Spatial-Specific Neural Network-Based OFDM Channel Estimation Under Time-Varying Channels
Wei, H., Wang, J., Chen, X., Peng, Q., Zha, Y., & Li, L. (2023) · IEEE Wireless Communications Letters · doi:10.1109/lwc.2023.3313808 · Cited by 5
“Recently, Deep Learning (DL) has been recognized as an effective enhancement for physical layer communication techniques [3], [4], [5], e.g., channel estimation, joint source-channel coding, signal detection, etc.”
BackgroundVNF-Enabled 5G Network Orchestration Framework for Slice Creation, Isolation and Management
Srinivasan, T., Venkatapathy, S., Jo, H., & Ra, I.-H. (2023) · Journal of Sensor and Actuator Networks · doi:10.3390/jsan12050065 · Cited by 4
Security Enhancement of mmWave MIMO Wireless Communication System Using Adversarial Training
Saini, M., & Grewal, S. K. (2024) · IEEJ Transactions on Electrical and Electronic Engineering · doi:10.1002/tee.24025 · Cited by 4
“…been done that proves the applicability of DL models in a lot of MIMO applications, for instance, channel estimation [2], signal classification [3], encoding and decoding [4], non-orthogonal multiple access, millimeter wave (mmWave) and massive MIMO (mMIMO) [5,6], MIMO detection [7], and so forth.”
BackgroundGemeda, M. D., Han, M., Abebe, A. T., & Kang, C. G. (2024) · IEEE Access · doi:10.1109/access.2024.3380203 · Cited by 4
A Novel CNN-based Architecture for Over-the-Air 5G OFDM Channel Estimation
Coutinho, F. D. L., Silva, H. S., Georgieva, P., & Oliveira, A. S. R. (2024) · IEEE MTT-S International Microwave Symposium digest · doi:10.1109/ims40175.2024.10600227 · Cited by 4
“Therefore, the availability of open-access datasets becomes crucial for comparing the performance of different DL-based algorithms and architectures [8].”
AI - Driven Self-Optimizing Receivers for Cognitive Radio Networks
Wang, Y., Tang, X., Mendis, G. J., Wei‐Kocsis, J., Madanayake, A., & Mandal, S. (2019) · IEEE Cognitive Communications for Aerospace Applications Workshop (CCAAW) · doi:10.1109/ccaaw.2019.8904889 · Cited by 4
Deep Learning Transceiver for Terahertz Band Communication System with 1-bit ADC and Oversampling
Gemeda, M. D., Han, M., Abebe, A. T., & Kang, C. G. (2022) · IEEE Consumer Communications and Networking Conference · doi:10.1109/ccnc49033.2022.9700716 · Cited by 4
Deep learning for free space optics in a data center environment
Darwesh, L., & Arnon, S. (2018) · Laser Communication and Propagation through the Atmosphere and Oceans · doi:10.1117/12.2321025 · Cited by 4
“In this network the values of xi in layer i are a function of xi−1 as follows xi = σ(wixi−1 + bi) (2)”
BackgroundDeep Learning for the Degraded Broadcast Channel
Stauffer, E., Wang, A. Y.-D., & Jindal, N. (2019) · Asilomar Conference on Signals, Systems and Computers (ACSSC) · doi:10.1109/ieeeconf44664.2019.9048974 · Cited by 4
“References: [1] Timothy O’Shea, Jakob Hoydis, “An Introduction to Deep Learning for the Physical Layer.”
BackgroundMethodInfluentialMarseet, A. A. (2018) · RIT Scholar Works (Rochester Institute of Technology) · Cited by 5
Restuccia, F., & Melodia, T. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1903.05460 · Cited by 5
“Among other issues, physical-layer modulation recognition through deep learning has received significant attention in the last two years [15, 22–24, 51, 52].”
BackgroundTo Learn or Not to Learn: Deep Learning Assisted Wireless Modem DesignPreprint
Xue, S., Li, A., Wang, J., Yi, N., Ma, Y., Tafazolli, R., & Dodgson, T. E. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1909.07791 · Cited by 4
“It is also possible to replace both the transmitter and receiver with their corresponding ANN black-boxes, so as to form an autoencoder which can be trained end-to-end for joint transmitter and receiver design [4, 5].”
BackgroundMostaani, A., Vu, T. X., Chatzinotas, S., & Ottersten, B. (2020) · arXiv (Cornell University) · Cited by 4
Vameghestahbanati, M. (2020) · doi:10.22215/etd/2020-13955 · Cited by 4
Self-organization for 5G and beyond mobile networks using reinforcement learning
Valente Klaine, P. H. (2019) · Enlighten: Theses (The University of Glasgow) · doi:10.5525/gla.thesis.74295 · Cited by 5
Color space-based autoencoder for optical camera communications
Luna‐Rivera, J. M., Rabadan, J. A., Rufo, J., Guerra, V., Moreno, D., & Pérez‐Jiménez, R. (2024) · Expert Systems with Applications · doi:10.1016/j.eswa.2023.123101 · Cited by 4
Tahenni, A., & Merazka, F. (2024) · International Artificial Intelligence and Data Processing Symposium (IDAP) · doi:10.1109/idap64064.2024.10710961 · Cited by 3
“This comprehensive approach to signal processing at the relay node enhances the decoding performance and ensures reliable communication in PNC-VLC systems [10] [11].”
BackgroundAdaptive signal recognition in mines based on deep learning
Rong, Y., Wang, A., Wang, M., & Zhu, T. (2025) · Scientific Reports · doi:10.1038/s41598-025-95070-2 · Cited by 3
“from IQ data, but their ability to model long-range dependencies in mine signals remains limited 15 .”
Deng, H., Liu, L., Xu, J., & Liu, T. (2025) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2025.3589007 · Cited by 3
“Recently, deep learning (DL) has been introduced into the application of wireless physical layers because of its excellent ability to learn from data [19], [20].”
Mask-Based PAPR Reduction Scheme With Deep Learning for OFDM Systems
He, R., Zhang, X., Liu, Z., Cui, Q., & Tao, X. (2025) · IEEE Transactions on Wireless Communications · doi:10.1109/twc.2025.3593042 · Cited by 4
“The application of DL at the physical layer, focusing on how neural networks (NNs) are trained to optimize the entire communication process, was discussed in [26] and [27].”
Physical Backdoor Attack Against Deep Learning-Based Modulation Classification
Salmi, Y., & Bogucka, H. (2025) · IEEE International Mediterranean Conference on Communications and Networking (MeditCom) · doi:10.1109/meditcom64437.2025.11104329 · Cited by 3
“Deep learning (DL) models have been approved as powerful tools for modeling and managing processes in wireless communications, including signal processing in the physical layer (PHY) [1].”
Harnessing Imperfection: Deep Learning-Based Receiver for Non-Coherent Distributed Transmission
Ahmad, A., Agarwal, S., Kaur, S., Jha, S. S., & Chinni Prabhunath, G. (2025) · IEEE Transactions on Cognitive Communications and Networking · doi:10.1109/tccn.2025.3614634 · Cited by 3
“It is applicable to practical NCDT use cases, such as asynchronous cell-free massive MISO/MIMO systems, autonomous V2X networks, distributed IoT and radar systems, UAV swarms, and beyond-5G/6G networks [40], [41], [42], [43].”
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.