arxiv: 2605.01570 · v1 · submitted 2026-05-02 · 💻 cs.IT · math.IT

Recognition: unknown

Neural Equalisers for Highly Compressed Faster-than-Nyquist Signalling: Design, Performance, Complexity and Robustness

Nambi Sheshadri, R David Koilpillai, Sheetal Kalyani, Shubham Paul

Pith reviewed 2026-05-09 17:38 UTC · model grok-4.3

classification 💻 cs.IT math.IT

keywords neural equalizersfaster-than-Nyquist signalingintersymbol interferencedeep learning receiversspectral efficiencylow complexity detectionrobustness

0 comments

The pith

Deep learning receivers enable reliable detection in Faster-than-Nyquist systems compressed by up to 75 percent.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that neural-network equalizers can manage the severe intersymbol interference created by packing symbols denser than the Nyquist limit, specifically at packing factors that compress spectrum by 75 percent. It introduces a sliding-window approach that supplies short-term context to each detection decision without requiring full-sequence processing, keeping both latency and computational cost low enough for real-time operation. Training on simulated channels and noise profiles produces models that hold performance across a range of conditions, indicating that FTN transmission need not rely on high-complexity classical detectors. A reader would care because this combination directly raises achievable data rates inside fixed bandwidth allocations while remaining practical to implement.

Core claim

A deep-learning framework for FTN signalling, built around a sliding-window detector, delivers reliable symbol recovery even under aggressive spectral compression up to 75 percent; the architecture is tuned for low latency and low complexity and maintains accuracy across varying channel conditions and noise profiles.

What carries the argument

The sliding-window detection method inside deep-learning receivers, which supplies limited temporal context to mitigate the controlled intersymbol interference that FTN deliberately introduces.

If this is right

FTN systems become viable at packing factors that achieve 75 percent spectral compression without catastrophic error rates.
Low-latency, low-complexity neural receivers make FTN suitable for real-time, scalable deployments.
The models exhibit resilience to changes in noise profiles and channel conditions within the tested range.
Aggressive FTN no longer requires high-complexity traditional equalizers to remain practical.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same sliding-window neural structure might be applied to other controlled-interference schemes that exceed classical Nyquist limits.
Hardware deployment would likely require retraining or adaptation steps to capture impairments absent from the simulation environment.
If robustness holds, FTN could increase throughput in spectrum-constrained links such as satellite or dense urban wireless channels.

Load-bearing premise

That detection performance measured under simulated channel and noise conditions will continue to hold when the same networks encounter unmodeled hardware distortions and time-varying propagation effects.

What would settle it

Bit-error-rate measurements of the trained neural equalizers on real RF hardware transmitting FTN waveforms under measured multipath and hardware impairments, compared directly against the simulated results reported in the paper.

Figures

Figures reproduced from arXiv: 2605.01570 by Nambi Sheshadri, R David Koilpillai, Sheetal Kalyani, Shubham Paul.

**Figure 1.** Figure 1: Generic FTN System Model. the performance of the proposed receivers under realworld conditions. 4) Addressing Generalisation Concerns: One of the predominant challenges associated with Deep Learning models is a potential overfitting problem, which undermines their ability to generalise to scenarios that were not represented in the training dataset. Our study demonstrates that the proposed models effect… view at source ↗

**Figure 4.** Figure 4: The operation of the Sliding NN detector view at source ↗

**Figure 3.** Figure 3: Modified System Diagram the reduced channel. The reduced channel allows us to reduce the size of the input window for our models. This allows our models to be more computationally efficient. Refer [40] for more details on the super-minimum phase channel. Other works in Table I usually use a whitening matrix or equalise coloured noise samples. C. Problem Definition In Subsection II-A, we noted that the matr… view at source ↗

**Figure 5.** Figure 5: FTN NN Structure architectures that may be used to design the NN detector in view at source ↗

**Figure 6.** Figure 6: shows the model diagram of the proposed FTNTransformerNet equaliser. We use Tensorflow to build our Transformer systems. We have an Input Layer of size 𝐿 view at source ↗

**Figure 7.** Figure 7: Detailed Structure of the FTN-TransformerNet allocates a single probability value to an output class. This is possible as we have only two possible cases. Namely, 0 and 1 correspond to the symbols -1 and 1. Effectively, the output is the probability of 1. We use a low-complexity model of the transformer that has fewer parameters than the LSTM and BiLSTM models discussed earlier. This ensures swift inferenc… view at source ↗

**Figure 9.** Figure 9: Loss Curves for FTN-TransformerNet iterations, the losses eventually start to decrease significantly after 50 epochs. And finally, the losses do not drop after the 340𝑡 ℎ epoch and thus we stop after the 400𝑡 ℎ epoch. The unidirectional LSTM model is the quickest, followed by the bidirectional LSTM and the transformer takes much longer. The transformer has a larger number of parameters, and so a much large… view at source ↗

**Figure 8.** Figure 8: Training and Validation Loss for FTN-LSTM and FTN-BiLSTM. as the training dataset. This allows for the testing data to be large enough to reliably measure small BER values such as 1.0×10−06 . To ensure that our models are robust to noise, we train them over a range of SNRs varying from 6 dB to 12 dB for 𝜏 = 0.5 and from 8 dB to 20 dB for 𝜏 = 0.35. This is because the impact of ISI is more significant for s… view at source ↗

**Figure 11.** Figure 11: BER Performance across varying packing ratio models view at source ↗

**Figure 12.** Figure 12: BER comparison of FTN Decoders for 𝜏 = 0.35 In this subsection, we study the BER performance of the bidirectional-LSTM models view at source ↗

**Figure 13.** Figure 13: BER comparison of FTN Decoders for 𝜏 = 0.25 -10 -5 0 5 10 15 20 -0.4 -0.2 0 0.2 0.4 0.6 0.8 = 0.50 = 0.35 = 0.25 view at source ↗

**Figure 15.** Figure 15: BER Performance of SNR specific models for 𝜏 = 0.5 with the LSTM model, but similar results can be obtained for other models too. We train the models using a supervised learning approach. During training, we set the observation windows (𝑌𝑘) as the inputs and the corresponding transmitted symbols (𝑥[𝑘]) as the labels. Table IV details the training and testing setup for the FTN-LSTM model. To test our model… view at source ↗

**Figure 16.** Figure 16: BER performance with sampling offset for view at source ↗

**Figure 18.** Figure 18: BER performance with Orthogonal Training and Testing Datasets for view at source ↗

**Figure 19.** Figure 19: BER performance with Varying Training Dataset size for view at source ↗

**Figure 20.** Figure 20: BER performance with Higher modulation for view at source ↗

read the original abstract

Faster-than-Nyquist (FTN) signalling has emerged as a compelling technique for enhancing spectral efficiency in bandwidth-constrained communication systems. By intentionally introducing controlled intersymbol interference (ISI), FTN allows transmission at rates exceeding the traditional Nyquist limit, unlocking new potential in high-speed data communication. However, its practical deployment remains challenged by the need for low-complexity detection strategies that can cope with the induced ISI while maintaining low latency and robust performance. We propose deep learning receivers that are resilient to non-idealities. In this paper, we present a deep learning-based framework for FTN signalling that addresses these challenges through several novel contributions. First, we propose a sliding window detection method that leverages temporal context while preserving computational efficiency. Second, we demonstrate the viability of FTN systems with very low packing factors, showing that reliable performance can be achieved even under aggressive spectral compression (up to 75\%). Our architecture is optimised for low latency and low complexity, making it suitable for real-time applications and scalable deployment. In addition, we assess the robustness of our models across varying channel conditions and noise profiles, providing insights into their generalisability and resilience.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Sliding-window neural equalizers handle 0.25-packing FTN in simulations, but the robustness story stays limited to AWGN and static channels.

read the letter

The main takeaway is that this work applies a sliding-window neural detector to Faster-than-Nyquist signaling at packing factors down to 0.25 and reports workable error rates under the simulated conditions they tested. That specific combination of extreme compression and low-latency architecture is the concrete step forward from earlier neural-receiver papers. The authors also keep the design focused on complexity and latency, which matters for any real deployment talk. They run checks across different noise profiles and claim the models hold up reasonably well, which is a reasonable thing to do even if the gains are incremental. The architecture itself looks like a standard feed-forward or recurrent setup adapted to the ISI pattern, nothing exotic. On the downside, every result comes from Monte Carlo trials over AWGN or fixed multipath with perfect sync. The paper does not inject phase noise, I/Q mismatch, amplifier distortion, or Doppler, so the headline claim that aggressive FTN remains viable in practice rests on an untested assumption about how well the learned equalizer generalizes when the ISI statistics shift. If the full manuscript adds only more simulated curves without those impairments, the practical takeaway shrinks. The citation pattern and derivations appear standard; no obvious self-referential fitting. This is the kind of targeted application paper that a reading group on ML-for-communications would discuss for the concrete numbers and the sliding-window trick. It is worth sending to peer review so referees can check the simulation details, baseline comparisons, and whether the robustness section actually addresses time-varying effects or just varies SNR. The work is coherent on its own terms and the claims are falsifiable, so it clears the bar for referee time even if revisions will be needed.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes a deep learning framework for detection in Faster-than-Nyquist (FTN) signalling, featuring a sliding-window neural equalizer architecture. It claims to demonstrate reliable performance at very low packing factors (corresponding to up to 75% spectral compression), with optimizations for low latency and complexity, plus an assessment of robustness across simulated channel conditions and noise profiles.

Significance. If the empirical results hold under the stated conditions, the work could meaningfully advance practical FTN deployment by showing that neural equalizers can handle aggressive ISI at packing factors down to 0.25 while remaining computationally tractable. The focus on sliding-window processing and robustness evaluation addresses real barriers to FTN adoption in bandwidth-constrained systems.

major comments (2)

[Abstract and robustness section] Abstract and robustness assessment: the central claim that 'reliable performance can be achieved even under aggressive spectral compression (up to 75%)' rests on Monte Carlo simulations over AWGN or static multipath with fixed noise statistics and perfect synchronization. No injection of phase noise, I/Q imbalance, amplifier nonlinearity, or Doppler spread is described, so the reported error rates do not yet substantiate viability under the hardware impairments and time-varying environments that would alter the effective ISI pattern.
[Abstract and results] Performance evaluation: the abstract states performance and robustness claims but supplies no quantitative BER curves, baseline comparisons against conventional equalizers (e.g., BCJR or MMSE), error bars, or training hyper-parameter details. Without these, the viability demonstration at packing factor 0.25 cannot be verified or reproduced.

minor comments (2)

[Design section] The sliding-window method is introduced but the manuscript should explicitly state how window size is selected, its latency-complexity trade-off, and any ablation on window length.
[Throughout] Notation for packing factor and compression ratio should be defined once at first use and used consistently.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback, which helps clarify the scope and presentation of our results. We address each major comment below and indicate the planned revisions.

read point-by-point responses

Referee: [Abstract and robustness section] Abstract and robustness assessment: the central claim that 'reliable performance can be achieved even under aggressive spectral compression (up to 75%)' rests on Monte Carlo simulations over AWGN or static multipath with fixed noise statistics and perfect synchronization. No injection of phase noise, I/Q imbalance, amplifier nonlinearity, or Doppler spread is described, so the reported error rates do not yet substantiate viability under the hardware impairments and time-varying environments that would alter the effective ISI pattern.

Authors: We appreciate the referee highlighting the boundaries of the evaluated conditions. The manuscript reports robustness results under AWGN and static multipath channels with varying noise profiles and perfect synchronization, as described in the robustness assessment section. We agree that phase noise, I/Q imbalance, amplifier nonlinearity, and Doppler spread were not injected into the simulations. We will revise the abstract to state the evaluated conditions more precisely and add a short paragraph in the robustness section acknowledging these unmodeled impairments as a limitation, together with a note that they constitute important directions for future work. revision: partial
Referee: [Abstract and results] Performance evaluation: the abstract states performance and robustness claims but supplies no quantitative BER curves, baseline comparisons against conventional equalizers (e.g., BCJR or MMSE), error bars, or training hyper-parameter details. Without these, the viability demonstration at packing factor 0.25 cannot be verified or reproduced.

Authors: The body of the manuscript contains the requested quantitative elements: BER curves at packing factors down to 0.25, direct comparisons against BCJR and MMSE equalizers, Monte-Carlo error bars, and full training hyper-parameter tables in the experimental setup and results sections. To improve accessibility, we will revise the abstract to include a concise statement of key quantitative outcomes (e.g., achieved BER at packing factor 0.25) and explicit reference to the baseline comparisons. revision: yes

Circularity Check

0 steps flagged

No circularity in derivation chain

full rationale

The paper proposes neural equalizer architectures for FTN signalling and reports simulation results for performance, complexity, and robustness under AWGN and static multipath channels. No equations, uniqueness theorems, or predictions are presented that reduce by construction to fitted inputs, self-citations, or ansatzes from prior author work. Claims of viability at low packing factors rest on empirical Monte Carlo assessments of the proposed models rather than any self-referential derivation.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The central claims rest on standard assumptions of additive white Gaussian noise channels, perfect synchronization, and the universal approximation capability of neural networks; no new free parameters, axioms, or invented entities are introduced in the abstract.

pith-pipeline@v0.9.0 · 5523 in / 1110 out tokens · 37745 ms · 2026-05-09T17:38:38.147040+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

50 extracted references · 2 canonical work pages · 1 internal anchor

[1]

On ”sampling, data transmission, and the nyquist rate

D. Tufts, “On ”sampling, data transmission, and the nyquist rate” and adaptive communication using randomly time-varying channels,” Proceedings of the IEEE, vol. 56, no. 5, pp. 889–889, 1968

1968
[2]

Exploiting faster-than-nyquist signaling,

A. Liveris and C. Georghiades, “Exploiting faster-than-nyquist signaling,” IEEE Transactions on Communications, vol. 51, no. 9, pp. 1502–1511, 2003

2003
[3]

Faster-than-nyquist signaling,

J. E. Mazo, “Faster-than-nyquist signaling,”The Bell System Technical Journal, vol. 54, no. 8, pp. 1451–1462, 1975

1975
[4]

On the minimum distance problem for faster-than-nyquist signaling,

J. Mazo and H. Landau, “On the minimum distance problem for faster-than-nyquist signaling,”IEEE Transactions on Information Theory, vol. 34, no. 6, pp. 1420–1427, 1988

1988
[5]

On computing the minimum distance for faster than nyquist signaling,

D. Hajela, “On computing the minimum distance for faster than nyquist signaling,”IEEE Transactions on Information Theory, vol. 36, no. 2, pp. 289–295, 1990

1990
[6]

Frequency-domain equalization of faster-than-nyquist signal- ing,

S. Sugiura, “Frequency-domain equalization of faster-than-nyquist signal- ing,”IEEE Wireless Communications Letters, vol. 2, no. 5, pp. 555–558, 2013

2013
[7]

Svd-precoded faster-than-nyquist signaling with optimal and truncated power allocation,

T. Ishihara and S. Sugiura, “Svd-precoded faster-than-nyquist signaling with optimal and truncated power allocation,”IEEE Transactions on Wireless Communications, vol. 18, no. 12, pp. 5909–5923, 2019

2019
[8]

Receivers for faster-than- nyquist signaling with and without turbo equalization,

A. Prlja, J. B. Anderson, and F. Rusek, “Receivers for faster-than- nyquist signaling with and without turbo equalization,” in2008 IEEE International Symposium on Information Theory, 2008, pp. 464–468

2008
[9]

Reduced-complexity receivers for strongly narrowband intersymbol interference introduced by faster-than-nyquist signaling,

A. Prlja and J. B. Anderson, “Reduced-complexity receivers for strongly narrowband intersymbol interference introduced by faster-than-nyquist signaling,”IEEE Transactions on Communications, vol. 60, no. 9, pp. 2591–2601, 2012

2012
[10]

Turbo equalization and an m-bcjr algorithm for strongly narrowband intersymbol interference,

J. B. Anderson and A. Prlja, “Turbo equalization and an m-bcjr algorithm for strongly narrowband intersymbol interference,” in2010 International Symposium On Information Theory and Its Applications, 2010, pp. 261– 266

2010
[11]

Faster-than-nyquist signaling,

J. B. Anderson, F. Rusek, and V. ¨Owall, “Faster-than-nyquist signaling,” Proceedings of the IEEE, vol. 101, no. 8, pp. 1817–1830, 2013

2013
[12]

Pre-equalized faster-than- nyquist transmission,

M. Jana, A. Medra, L. Lampe, and J. Mitra, “Pre-equalized faster-than- nyquist transmission,”IEEE Transactions on Communications, vol. 65, no. 10, pp. 4406–4418, 2017

2017
[13]

Faster-than- nyquist signaling: An overview,

J. Fan, S. Guo, X. Zhou, Y. Ren, G. Y. Li, and X. Chen, “Faster-than- nyquist signaling: An overview,”IEEE Access, vol. 5, pp. 1925–1940, 2017

1925
[14]

Eigendecomposition-precoded faster-than-nyquist signaling with index modulation,

P. Chaki, T. Ishihara, and S. Sugiura, “Eigendecomposition-precoded faster-than-nyquist signaling with index modulation,”IEEE Transactions on Communications, vol. 70, no. 7, pp. 4822–4836, 2022

2022
[15]

Optimal decoding of linear codes for minimizing symbol error rate (corresp.),

L. Bahl, J. Cocke, F. Jelinek, and J. Raviv, “Optimal decoding of linear codes for minimizing symbol error rate (corresp.),”IEEE Transactions on Information Theory, vol. 20, no. 2, pp. 284–287, 1974

1974
[16]

Deep-learning based equal- ization of highly compressed faster than nyquist signals,

S. Paul, N. Seshadri, and R. D. Koilpillai, “Deep-learning based equal- ization of highly compressed faster than nyquist signals,” in2023 IEEE International Conference on Advanced Networks and Telecommunica- tions Systems (ANTS), 2023, pp. 396–401

2023
[17]

Learning to communicate and energize: Modulation, coding, and multiple access designs for wireless information-power transmission,

M. Varasteh, J. Hoydis, and B. Clerckx, “Learning to communicate and energize: Modulation, coding, and multiple access designs for wireless information-power transmission,”IEEE Transactions on Communica- tions, vol. 68, no. 11, pp. 6822–6839, 2020

2020
[18]

An introduction to deep learning for the physical layer,

T. O’Shea and J. Hoydis, “An introduction to deep learning for the physical layer,”IEEE Transactions on Cognitive Communications and Networking, vol. 3, no. 4, pp. 563–575, 2017

2017
[19]

Joint learning of geometric and probabilistic constellation shaping,

M. Stark, F. Ait Aoudia, and J. Hoydis, “Joint learning of geometric and probabilistic constellation shaping,” in2019 IEEE Globecom Workshops (GC Wkshps), 2019, pp. 1–6

2019
[20]

Backpropagating through the air: Deep learning at physical layer without channel models,

V. Raj and S. Kalyani, “Backpropagating through the air: Deep learning at physical layer without channel models,”IEEE Communications Letters, vol. 22, no. 11, pp. 2278–2281, 2018

2018
[21]

One-bit ofdm receivers via deep learning,

E. Balevi and J. G. Andrews, “One-bit ofdm receivers via deep learning,” IEEE Transactions on Communications, vol. 67, no. 6, pp. 4326–4336, 2019

2019
[22]

Deep learning based symbol detection for molecular communications,

S. Sharma, D. Dixit, and K. Deka, “Deep learning based symbol detection for molecular communications,” in2020 IEEE International Conference on Advanced Networks and Telecommunications Systems (ANTS), 2020, pp. 1–6

2020
[23]

Viterbinet: A deep learning based viterbi algorithm for symbol detection,

N. Shlezinger, N. Farsad, Y. C. Eldar, and A. J. Goldsmith, “Viterbinet: A deep learning based viterbi algorithm for symbol detection,”IEEE Transactions on Wireless Communications, vol. 19, no. 5, pp. 3319–3331, 2020

2020
[24]

Neural network detection of data sequences in communication systems,

N. Farsad and A. Goldsmith, “Neural network detection of data sequences in communication systems,”IEEE Transactions on Signal Processing, vol. 66, no. 21, pp. 5663–5678, 2018

2018
[25]

Interference cancellation technique for faster-than-nyquist signalling,

S. Nie, M. Guo, and Y. Shen, “Interference cancellation technique for faster-than-nyquist signalling,”Electronics Letters, vol. 52, no. 13, pp. 1126–1128, 2016

2016
[26]

A very low complexity successive symbol-by-symbol sequence estimator for faster-than-nyquist signaling,

E. Bedeer, M. H. Ahmed, and H. Yanikomeroglu, “A very low complexity successive symbol-by-symbol sequence estimator for faster-than-nyquist signaling,”IEEE Access, vol. 5, pp. 7414–7422, 2017

2017
[27]

Low complexity message passing receiver for faster-than-nyquist signaling in nonlinear channels,

X. Wen, W. Yuan, D. Yang, N. Wu, and J. Kuang, “Low complexity message passing receiver for faster-than-nyquist signaling in nonlinear channels,”IEEE Access, vol. 6, pp. 68 233–68 241, 2018

2018
[28]

Joint precoding and pre-equalization for faster-than-nyquist transmission over multipath fading channels,

S. Wen, G. Liu, C. Liu, H. Qu, L. Zhang, and M. A. Imran, “Joint precoding and pre-equalization for faster-than-nyquist transmission over multipath fading channels,”IEEE Transactions on Vehicular Technology, vol. 71, no. 4, pp. 3948–3963, 2022

2022
[29]

Receiver design for faster- than-nyquist signaling: Deep-learning-based architectures,

P. Song, F. Gong, Q. Li, G. Li, and H. Ding, “Receiver design for faster- than-nyquist signaling: Deep-learning-based architectures,”IEEE Access, vol. 8, pp. 68 866–68 873, 2020

2020
[30]

Deep learning-based list sphere decoding for faster-than-nyquist (ftn) signaling detection,

A. Sina and E. Bedeer, “Deep learning-based list sphere decoding for faster-than-nyquist (ftn) signaling detection,” in2022 IEEE 95th Vehicular Technology Conference: (VTC2022-Spring), 2022, pp. 1–6

2022
[31]

A novel sum-product detection algorithm for faster-than-nyquist signaling: A deep learning approach,

B. Liu, S. Li, Y. Xie, and J. Yuan, “A novel sum-product detection algorithm for faster-than-nyquist signaling: A deep learning approach,” IEEE Transactions on Communications, vol. 69, no. 9, pp. 5975–5987, 2021

2021
[32]

Mhsa-cnn-bi- lstm based interference cancellation algorithm for mimo-ftn-owc system,

Q. Yang, M. Cao, G. Zhou, Y. Zhang, R. Yao, and H. Wang, “Mhsa-cnn-bi- lstm based interference cancellation algorithm for mimo-ftn-owc system,” in2024 IEEE 16th International Conference on Advanced Infocomm Technology (ICAIT), 2024, pp. 1–5

2024
[33]

Ftn-based non- orthogonal signal detection technique with machine learning in quasi- static multipath channel,

M.-S. Baek, E.-S. Jung, Y. S. Park, and Y.-T. Lee, “Ftn-based non- orthogonal signal detection technique with machine learning in quasi- static multipath channel,”IEEE Transactions on Broadcasting, vol. 70, no. 1, pp. 78–86, 2024

2024
[34]

Cyclic prefix/suffix-assisted frequency-domain equalization for faster-than-nyquist signaling block transmission,

S.-B. Hong and J.-S. Seo, “Cyclic prefix/suffix-assisted frequency-domain equalization for faster-than-nyquist signaling block transmission,” in2016 IEEE 27th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), 2016, pp. 1–6

2016
[35]

For security and higher spectrum efficiency: A variable packing ratio transmission system based on faster-than-nyquist and deep learning,

P. Song, N. Zhang, L. Cai, G. Li, T. Wu, and F.-K. Gong, “For security and higher spectrum efficiency: A variable packing ratio transmission system based on faster-than-nyquist and deep learning,”IEEE Transactions on Wireless Communications, vol. 22, no. 9, pp. 5898–5913, 2023

2023
[36]

A mimo detec- tor utilizing dnn for faster-than-nyquist optical wireless communications,

M. Cao, R. Yao, Q. Sun, Y. Zhang, Q. Yang, and H. Wang, “A mimo detec- tor utilizing dnn for faster-than-nyquist optical wireless communications,” in2023 IEEE 15th International Conference on Advanced Infocomm Technology (ICAIT), 2023, pp. 27–30

2023
[37]

A mimo detector with deep-neural-network for faster-than-nyquist optical wireless communications,

——, “A mimo detector with deep-neural-network for faster-than-nyquist optical wireless communications,”IEEE Photonics Journal, vol. 16, no. 2, pp. 1–9, 2024

2024
[38]

Deep learning assisted pre-equalization scheme for faster-than- nyquist optical wireless communications,

C. Minghua, W. Zhaoheng, W. Huiqin, X. Jieping, Z. Jiawei, and L. Wen- wen, “Deep learning assisted pre-equalization scheme for faster-than- nyquist optical wireless communications,” in2021 13th International Conference on Advanced Infocomm Technology (ICAIT), 2021, pp. 118– 121

2021
[39]

The evolution of faster-than-nyquist signaling,

T. Ishihara, S. Sugiura, and L. Hanzo, “The evolution of faster-than-nyquist signaling,”IEEE Access, vol. 9, pp. 86 535–86 564, 2021. 14

2021
[40]

Reduced Receivers for Faster-than-Nyquist Signaling and General Linear Channels,

Prlja, Adnan, “Reduced Receivers for Faster-than-Nyquist Signaling and General Linear Channels,” Ph.D. dissertation, Lund University, 2013. [Online]. Available:{https://lup.lub.lu.se/search/files/3525903/3447172. pdf}

work page arXiv 2013
[41]

Sliding bidirectional recurrent neural net- works for sequence detection in communication systems,

N. Farsad and A. Goldsmith, “Sliding bidirectional recurrent neural net- works for sequence detection in communication systems,” in2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2018, pp. 2331–2335

2018
[42]

Imagenet classification with deep convolutional neural networks,

A. Krizhevsky, I. Sutskever, and G. E. Hinton, “Imagenet classification with deep convolutional neural networks,” inAdvances in Neural Information Processing Systems, F. Pereira, C. Burges, L. Bottou, and K. Weinberger, Eds., vol. 25. Curran Associates, Inc., 2012. [Online]. Available: https://proceedings.neurips.cc/paper files/paper/ 2012/file/c399862d3b...

2012
[43]

Learning mid-level features for recognition,

Y.-L. Boureau, F. Bach, Y. LeCun, and J. Ponce, “Learning mid-level features for recognition,” in2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2010, pp. 2559–2566

2010
[44]

Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups,

G. Hinton, L. Deng, D. Yu, G. E. Dahl, A.-r. Mohamed, N. Jaitly, A. Senior, V. Vanhoucke, P. Nguyen, T. N. Sainath, and B. Kingsbury, “Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups,”IEEE Signal Processing Magazine, vol. 29, no. 6, pp. 82–97, 2012

2012
[45]

Deep learning,

Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,”nature, vol. 521, no. 7553, pp. 436–444, 2015

2015
[46]

Ian goodfellow, yoshua bengio, and aaron courville: Deep learning: The mit press, 2016, 800 pp, isbn: 0262035618,

J. Heaton, “Ian goodfellow, yoshua bengio, and aaron courville: Deep learning: The mit press, 2016, 800 pp, isbn: 0262035618,”Genetic programming and evolvable machines, vol. 19, no. 1, pp. 305–307, 2018

2016
[47]

Neural Machine Translation by Jointly Learning to Align and Translate

D. Bahdanau, K. Cho, and Y. Bengio, “Neural machine translation by jointly learning to align and translate,”arXiv preprint arXiv:1409.0473, 2014

work page internal anchor Pith review arXiv 2014
[48]

Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups,

G. Hinton, L. Deng, D. Yu, G. E. Dahl, A.-r. Mohamed, N. Jaitly, A. Senior, V. Vanhoucke, P. Nguyen, T. N. Sainathet al., “Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups,”IEEE Signal processing magazine, vol. 29, no. 6, pp. 82–97, 2012

2012
[49]

Delayed decision-feedback sequence estimation,

A. Duel-Hallen and C. Heegard, “Delayed decision-feedback sequence estimation,”IEEE Transactions on Communications, vol. 37, no. 5, pp. 428–436, 1989

1989
[50]

Decision feedback equalization,

C. Belfiore and J. Park, “Decision feedback equalization,”Proceedings of the IEEE, vol. 67, no. 8, pp. 1143–1156, 1979

1979