pith. sign in

arxiv: 2210.08770 · v1 · submitted 2022-10-17 · 💻 cs.IT · eess.SP· math.IT

Massive MIMO Channel Prediction Via Meta-Learning and Deep Denoising: Is a Small Dataset Enough?

Hwanjin Kim , Junil Choi , David J. Love This is my paper

Pith reviewed 2026-05-24 11:05 UTC · model grok-4.3

classification 💻 cs.IT eess.SPmath.IT
keywords massive MIMOchannel predictionmeta-learningMAMLdeep image priordenoisingsmall datasetwireless communications
0
0 comments X

The pith

Meta-learning allows massive MIMO channel predictors to adapt to new environments with only a few fine-tuning samples.

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

The paper aims to establish that a meta-learning approach can train a channel predictor that adapts quickly to new wireless environments with minimal additional data. This matters because conventional machine learning methods for channel prediction require large training sets for each new setting, creating high overhead in massive MIMO systems. The authors combine model-agnostic meta-learning for rapid adaptation with deep image prior denoising of the training data to boost accuracy. If the claims hold, the technique would cut the labeled data needed while delivering better predictions, especially under low signal-to-noise ratio conditions.

Core claim

The authors claim that applying the model-agnostic meta-learning algorithm produces a channel predictor that reaches higher accuracy in new environments after fine-tuning on only a few samples, and that denoising the training data via deep image prior yields further gains particularly in low signal-to-noise ratio regimes.

What carries the argument

Model-agnostic meta-learning (MAML) optimizes an initial model so that a small number of gradient steps on new labeled samples produces good performance, paired with deep image prior (DIP) denoising that reconstructs clean channel matrices from noisy observations without requiring external training data.

If this is right

  • Prediction accuracy improves in new environments after fine-tuning the meta-trained model with only a few labeled samples.
  • The DIP denoising step supplies additional accuracy gains, especially in low signal-to-noise ratio regimes.
  • Training overhead for channel prediction drops relative to standard machine learning methods that start from scratch in each environment.
  • The predictor adapts to changing conditions without collecting large new datasets for every shift.

Where Pith is reading between the lines

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

  • The approach could support real-time operation in mobile settings where channel statistics change on short timescales.
  • If the distribution-family premise holds in practice, similar meta-learning could apply to prediction of other quantities such as interference or beamforming vectors.
  • Testing the method on measured outdoor channels rather than simulated ones would reveal whether the reported gains survive hardware impairments and unmodeled propagation effects.

Load-bearing premise

New environments share the same statistical distribution family as the meta-training tasks so that a small number of labeled samples suffices for effective fine-tuning.

What would settle it

In a new environment whose channel statistics fall outside the meta-training distribution family, fine-tuning the MAML model on a few samples yields no accuracy gain or lower accuracy than a non-meta-trained baseline.

Figures

Figures reproduced from arXiv: 2210.08770 by David J. Love, Hwanjin Kim, Junil Choi.

Figure 1
Figure 1. Figure 1: Massive MIMO system consisting of a BS with [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: MAML structure: meta-training, meta-adaptation, a [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: MAML datasets consist of the meta-training, meta-ad [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: DIP-based denoising process architecture. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: NMSE vs. number of source tasks per UE with [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: NMSE vs. complexity order with Tad = 10, Nad = 20, and SNR = 20 dB. For the DIP architecture, we adopt the CNN with the number of iterations Niter = 2000 and Ld = 4 hidden-layers including Mi = 64 for all i. We also set the number of BS antennas M = 64, the number of source tasks per UE Tu = 1024, the complexity order no = 3, the number of epochs Nepoch = 20, and the batch size V = 64. The number of sample… view at source ↗
Figure 8
Figure 8. Figure 8: NMSE vs. number of adaptation samples with SNR [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: NMSE vs. number of adaptation samples with SNR [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
read the original abstract

Accurate channel knowledge is critical in massive multiple-input multiple-output (MIMO), which motivates the use of channel prediction. Machine learning techniques for channel prediction hold much promise, but current schemes are limited in their ability to adapt to changes in the environment because they require large training overheads. To accurately predict wireless channels for new environments with reduced training overhead, we propose a fast adaptive channel prediction technique based on a meta-learning algorithm for massive MIMO communications. We exploit the model-agnostic meta-learning (MAML) algorithm to achieve quick adaptation with a small amount of labeled data. Also, to improve the prediction accuracy, we adopt the denoising process for the training data by using deep image prior (DIP). Numerical results show that the proposed MAML-based channel predictor can improve the prediction accuracy with only a few fine-tuning samples. The DIP-based denoising process gives an additional gain in channel prediction, especially in low signal-to-noise ratio regimes.

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.

Referee Report

1 major / 1 minor

Summary. The paper proposes a meta-learning framework using model-agnostic meta-learning (MAML) for rapid adaptation of massive MIMO channel predictors to new environments with limited fine-tuning samples, augmented by deep image prior (DIP) denoising of training data. Numerical simulations are presented to show that the MAML approach yields higher prediction accuracy than standard training with few samples, and that DIP provides further improvement especially in low-SNR regimes.

Significance. If the empirical gains hold under broader validation, the work would address a practical bottleneck in ML-based channel prediction by reducing adaptation overhead in time-varying wireless environments. The combination of meta-learning initialization with denoising is a targeted contribution to the intersection of meta-learning and wireless communications.

major comments (1)
  1. [Numerical Results / Simulation Setup] The central claim that MAML enables quick adaptation with small labeled samples from new environments rests on the assumption that meta-training tasks and target tasks are drawn from the same distribution family. The reported numerical results vary parameters (user speed, angle spread) only inside a single channel model family; no cross-model experiments or measurement data are described to test whether the observed few-shot gains persist when this assumption is relaxed. This directly limits support for the headline result on reduced training overhead in general new environments.
minor comments (1)
  1. [Abstract] The abstract states numerical gains but omits any mention of network architectures, training/validation splits, baseline details, or error bars; these must be explicitly summarized in the main text (e.g., § on experimental setup) for reproducibility.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We provide a point-by-point response to the major comment below.

read point-by-point responses

  1. Referee: The central claim that MAML enables quick adaptation with small labeled samples from new environments rests on the assumption that meta-training tasks and target tasks are drawn from the same distribution family. The reported numerical results vary parameters (user speed, angle spread) only inside a single channel model family; no cross-model experiments or measurement data are described to test whether the observed few-shot gains persist when this assumption is relaxed. This directly limits support for the headline result on reduced training overhead in general new environments.

    Authors: We appreciate the referee pointing out the scope of our experimental validation. Our simulations are performed within one standard channel model family (with variations in user speed and angle spread to emulate different environments), which is a common evaluation setting in the wireless communications literature for meta-learning methods. This tests rapid adaptation when tasks share the same underlying distribution family but differ in parameters, directly supporting the claim of reduced overhead for such new environments. We agree that cross-model experiments or measurement-based validation would provide broader evidence for generalization to arbitrary environments; however, these are outside the current scope. We will revise the manuscript to explicitly articulate this assumption and the targeted scope of the results. revision: partial

Circularity Check

0 steps flagged

No circularity; results are empirical simulation outcomes

full rationale

The paper proposes an MAML-based predictor with DIP denoising and evaluates it via numerical simulations on channel models. No equations, predictions, or central claims reduce by construction to fitted parameters or self-defined quantities inside the same derivation. The reported accuracy gains with few fine-tuning samples are presented as simulation results rather than tautological re-statements of inputs. The shared-distribution-family premise for adaptation is a modeling assumption tested inside one simulation family, but this is an empirical scope limitation rather than a circular reduction of the derivation chain itself.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available, so the ledger is necessarily incomplete; the central claim rests on the unstated assumption that simulated channel realizations adequately represent real deployment distributions and that the meta-training task distribution matches future test environments.

pith-pipeline@v0.9.0 · 5699 in / 1048 out tokens · 18329 ms · 2026-05-24T11:05:27.547062+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

46 extracted references · 46 canonical work pages · 2 internal anchors

  1. [1]

    Noncooperative cellular wireless with unlimited num- bers of base station antennas,

    T. L. Marzetta, “Noncooperative cellular wireless with unlimited num- bers of base station antennas,” IEEE Transactions on Wireless Commu- nications, vol. 9, no. 11, pp. 3590–3600, Nov. 2010

    work page 2010

  2. [2]

    Impact of general channel agi ng conditions on the downlink performance of massive MIMO,

    A. K. Papazafeiropoulos, “Impact of general channel agi ng conditions on the downlink performance of massive MIMO,” IEEE Transactions on V ehicular Technology, vol. 66, no. 2, pp. 1428–1442, Feb. 2017

    work page 2017

  3. [3]

    Effects of channel aging in m assive MIMO systems,

    K. T. Truong and R. W. Heath, “Effects of channel aging in m assive MIMO systems,” Journal of Communications and Networks , vol. 15, no. 4, pp. 338–351, Aug. 2013

    work page 2013

  4. [4]

    Downlink training t echniques for FDD massive MIMO systems: Open-loop and closed-loop traini ng with memory,

    J. Choi, D. J. Love, and P . Bidigare, “Downlink training t echniques for FDD massive MIMO systems: Open-loop and closed-loop traini ng with memory,” IEEE Journal of Selected Topics in Signal Processing , vol. 8, no. 5, pp. 802–814, Oct. 2014

    work page 2014

  5. [5]

    Sum-rate and power scaling of massive MIMO system s with channel aging,

    C. Kong, C. Zhong, A. K. Papazafeiropoulos, M. Matthaiou , and Z. Zhang, “Sum-rate and power scaling of massive MIMO system s with channel aging,” IEEE Transactions on Communications , vol. 63, no. 12, pp. 4879–4893, Dec. 2015

    work page 2015

  6. [6]

    Adaptive beam tracking with th e unscented Kalman filter for millimeter wave communication,

    S. G. Larew and D. J. Love, “Adaptive beam tracking with th e unscented Kalman filter for millimeter wave communication,” IEEE Signal Pro- cessing Letters , vol. 26, no. 11, pp. 1658–1662, Nov. 2019

    work page 2019

  7. [7]

    Fast position-aided MIMO beam training via noisy tensor complet ion,

    T.-H. Chou, N. Michelusi, D. J. Love, and J. V . Krogmeier, “Fast position-aided MIMO beam training via noisy tensor complet ion,” IEEE Journal of Selected Topics in Signal Processing , vol. 15, no. 3, pp. 774– 788, Apr. 2021

    work page 2021

  8. [8]

    Study on Artificial Intelligence (AI)/Machine Learning (ML ) for NR Air Interface, 3GPP TR 38.843 Std., Dec. 2021

    work page 2021

  9. [9]

    Deep CNN for wideband mmwave massive MIMO channel estimati on using frequency correlation,

    P . Dong, H. Zhang, G. Y . Li, N. NaderiAlizadeh, and I. S. Ga spar, “Deep CNN for wideband mmwave massive MIMO channel estimati on using frequency correlation,” in 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , May 2019, pp. 4529–4533

    work page 2019

  10. [10]

    Long-rang e MIMO channel prediction using recurrent neural networks,

    W. Jiang, M. Strufe, and H. Dieter Schotten, “Long-rang e MIMO channel prediction using recurrent neural networks,” in 2020 IEEE 17th Annual Consumer Communications Networking Conference (CC NC), 2020, pp. 1–6

    work page 2020

  11. [11]

    Machine learning- based channel prediction in massive MIMO with channel aging,

    J. Y uan, H. Q. Ngo, and M. Matthaiou, “Machine learning- based channel prediction in massive MIMO with channel aging,” IEEE Transactions on Wireless Communications, vol. 19, no. 5, pp. 2960–2973, May 2020

    work page 2020

  12. [12]

    Adaptive channel pre diction, beamforming and scheduling design for 5G V2I network: Analy tical and machine learning approaches,

    T. E. Bogale, X. Wang, and L. B. Le, “Adaptive channel pre diction, beamforming and scheduling design for 5G V2I network: Analy tical and machine learning approaches,” IEEE Transactions on V ehicular Technology, vol. 69, no. 5, pp. 5055–5067, May 2020

    work page 2020

  13. [13]

    Mass ive MIMO channel prediction: Kalman filtering vs. machine learn ing,

    H. Kim, S. Kim, H. Lee, C. Jang, Y . Choi, and J. Choi, “Mass ive MIMO channel prediction: Kalman filtering vs. machine learn ing,” IEEE Transactions on Communications, vol. 69, no. 1, pp. 518–528, Jan. 2021

    work page 2021

  14. [14]

    Channel prediction in high-mobility massive MIMO: From spatio-temporal autor egression to deep learning,

    C. Wu, X. Yi, Y . Zhu, W. Wang, L. Y ou, and X. Gao, “Channel prediction in high-mobility massive MIMO: From spatio-temporal autor egression to deep learning,” IEEE Journal on Selected Areas in Communications , vol. 39, no. 7, pp. 1915–1930, Jul. 2021

    work page 1915

  15. [15]

    Learning to l earn by gradient descent by gradient descent,

    M. Andrychowicz, M. Denil, S. G´ omez, M. W. Hoffman, D. P fau, T. Schaul, B. Shillingford, and N. de Freitas, “Learning to l earn by gradient descent by gradient descent,” in Advances in Neural Information Processing Systems, vol. 29, 2016

    work page 2016

  16. [16]

    Optimization as a model for f ew-shot learning,

    S. Ravi and H. Larochelle, “Optimization as a model for f ew-shot learning,” in International Conference on Learning Representations , 2017

    work page 2017

  17. [17]

    Learn to sense: A meta-learning-based sensing and fusion framework for wire less sensor networks,

    H. Wu, Z. Zhang, C. Jiao, C. Li, and T. Q. S. Quek, “Learn to sense: A meta-learning-based sensing and fusion framework for wire less sensor networks,” IEEE Internet of Things Journal , vol. 6, no. 5, pp. 8215– 8227, Oct. 2019

    work page 2019

  18. [18]

    Roemnet: Robust meta lea rning based channel estimation in OFDM systems,

    H. Mao, H. Lu, Y . Lu, and D. Zhu, “Roemnet: Robust meta lea rning based channel estimation in OFDM systems,” in ICC 2019 - 2019 IEEE International Conference on Communications (ICC) , 2019, pp. 1–6

    work page 2019

  19. [19]

    Meta lea rning- based MIMO detectors: Design, simulation, and experimenta l test,

    J. Zhang, Y . He, Y .-W. Li, C.-K. Wen, and S. Jin, “Meta lea rning- based MIMO detectors: Design, simulation, and experimenta l test,” IEEE Transactions on Wireless Communications , vol. 20, no. 2, pp. 1122–1137, Feb. 2021

    work page 2021

  20. [20]

    Learning to de modulate from few pilots via offline and online meta-learning,

    S. Park, H. Jang, O. Simeone, and J. Kang, “Learning to de modulate from few pilots via offline and online meta-learning,” IEEE Transactions on Signal Processing , vol. 69, pp. 226–239, Jan. 2021

    work page 2021

  21. [21]

    Transfer learning and meta learning-based fast downlink beamformin g adapta- tion,

    Y . Y uan, G. Zheng, K.-K. Wong, B. Ottersten, and Z.-Q. Lu o, “Transfer learning and meta learning-based fast downlink beamformin g adapta- tion,” IEEE Transactions on Wireless Communications , vol. 20, no. 3, pp. 1742–1755, Mar. 2021

    work page 2021

  22. [22]

    Meta-learning based beamforming design for MISO downlink,

    J. Xia and D. Gunduz, “Meta-learning based beamforming design for MISO downlink,” in 2021 IEEE International Symposium on Informa- tion Theory (ISIT) , Jul. 2021, pp. 2954–2959

    work page 2021

  23. [23]

    Few-shot learning based hybrid b eamforming under birth-death process of scattering paths,

    Y . Long and S. Murphy, “Few-shot learning based hybrid b eamforming under birth-death process of scattering paths,” IEEE Communications Letters, vol. 25, no. 5, pp. 1687–1691, May 2021

    work page 2021

  24. [24]

    Embedding model-based fast meta learning for downlink beamforming ad aptation,

    J. Zhang, Y . Y uan, G. Zheng, I. Krikidis, and K.-K. Wong, “Embedding model-based fast meta learning for downlink beamforming ad aptation,” IEEE Transactions on Wireless Communications, vol. 21, no. 1, pp. 149– 162, Jan. 2022

    work page 2022

  25. [25]

    Deep t ransfer learning-based downlink channel prediction for FDD massiv e MIMO systems,

    Y . Y ang, F. Gao, Z. Zhong, B. Ai, and A. Alkhateeb, “Deep t ransfer learning-based downlink channel prediction for FDD massiv e MIMO systems,” IEEE Transactions on Communications , vol. 68, no. 12, pp. 7485–7497, Dec. 2020

    work page 2020

  26. [26]

    Downlink CSI feedback algorithm with deep transfe r learning for FDD massive MIMO systems,

    J. Zeng, J. Sun, G. Gui, B. Adebisi, T. Ohtsuki, H. Gacani n, and H. Sari, “Downlink CSI feedback algorithm with deep transfe r learning for FDD massive MIMO systems,” IEEE Transactions on Cognitive Communications and Networking , vol. 7, no. 4, pp. 1253–1265, Dec. 2021

    work page 2021

  27. [27]

    Model-agnostic meta -learning for fast adaptation of deep networks,

    C. Finn, P . Abbeel, and S. Levine, “Model-agnostic meta -learning for fast adaptation of deep networks,” in Proceedings of the 34th International Conference on Machine Learning , vol. 70, Aug. 2017, pp. 1126–1135

    work page 2017

  28. [28]

    Massive MIMO channel prediction: Machine learning versus Kalman filtering,

    H. Kim, S. Kim, H. Lee, and J. Choi, “Massive MIMO channel prediction: Machine learning versus Kalman filtering,” in 2020 IEEE Globecom W orkshops (GC Wkshps) , Dec. 2020, pp. 1–6

    work page 2020

  29. [29]

    Channel estimation f or cell-free mmwave massive MIMO through deep learning,

    Y . Jin, J. Zhang, S. Jin, and B. Ai, “Channel estimation f or cell-free mmwave massive MIMO through deep learning,” IEEE Transactions on V ehicular Technology, vol. 68, no. 10, pp. 10 325–10 329, Oct. 2019

    work page 2019

  30. [30]

    Deep learning-based channel estimation,

    M. Soltani, V . Pourahmadi, A. Mirzaei, and H. Sheikhzad eh, “Deep learning-based channel estimation,” IEEE Communications Letters , vol. 23, no. 4, pp. 652–655, Apr. 2019

    work page 2019

  31. [31]

    Massive MIMO cha nnel estimation with an untrained deep neural network,

    E. Balevi, A. Doshi, and J. G. Andrews, “Massive MIMO cha nnel estimation with an untrained deep neural network,” IEEE Transactions on Wireless Communications, vol. 19, no. 3, pp. 2079–2090, Mar. 2020

    work page 2079

  32. [32]

    Beyond a Gaussian denoiser: Residual learning of deep CNN for image d enoising,

    K. Zhang, W. Zuo, Y . Chen, D. Meng, and L. Zhang, “Beyond a Gaussian denoiser: Residual learning of deep CNN for image d enoising,” IEEE Transactions on Image Processing , vol. 26, no. 7, pp. 3142–3155, Jul. 2017

    work page 2017

  33. [33]

    Deep learning-bas ed channel estimation for beamspace mmwave massive MIMO systems,

    H. He, C.-K. Wen, S. Jin, and G. Y . Li, “Deep learning-bas ed channel estimation for beamspace mmwave massive MIMO systems,” IEEE Wireless Communications Letters, vol. 7, no. 5, pp. 852–855, Oct. 2018

    work page 2018

  34. [34]

    Deep learn ing-based beamspace channel estimation in mmwave massive MIMO system s,

    Y . Zhang, Y . Mu, Y . Liu, T. Zhang, and Y . Qian, “Deep learn ing-based beamspace channel estimation in mmwave massive MIMO system s,” IEEE Wireless Communications Letters , vol. 9, no. 12, pp. 2212–2215, Dec. 2020

    work page 2020

  35. [35]

    Deep learnin g-based denoise network for CSI feedback in FDD massive MIMO systems ,

    H. Y e, F. Gao, J. Qian, H. Wang, and G. Y . Li, “Deep learnin g-based denoise network for CSI feedback in FDD massive MIMO systems ,” IEEE Communications Letters , vol. 24, no. 8, pp. 1742–1746, Aug. 2020

    work page 2020

  36. [36]

    Deep image pr ior,

    D. Ulyanov, A. V edaldi, and V . Lempitsky, “Deep image pr ior,” in Proceedings of the IEEE Conference on Computer Vision and Pa ttern Recognition (CVPR) , Jun. 2018

    work page 2018

  37. [37]

    Pilot-based LMMSE channel est imation for OFDM systems with power–delay profile approximation,

    K.-C. Hung and D. W. Lin, “Pilot-based LMMSE channel est imation for OFDM systems with power–delay profile approximation,” IEEE 11 Transactions on V ehicular Technology, vol. 59, no. 1, pp. 150–159, Jan. 2010

    work page 2010

  38. [38]

    Meta- learning in neural networks: A survey,

    T. M. Hospedales, A. Antoniou, P . Micaelli, and A. J. Sto rkey, “Meta- learning in neural networks: A survey,” IEEE Transactions on Pattern Analysis and Machine Intelligence , pp. 1–1, 2021

    work page 2021

  39. [39]

    Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour

    P . Goyal, P . Doll´ ar, R. Girshick, P . Noordhuis, L. Weso lowski, A. Kyrola, A. Tulloch, Y . Jia, and K. He, “Accurate, large min ibatch SGD: Training imagenet in 1 hour,” 2017. [Online]. Availabl e: https://arxiv.org/abs/1706.02677

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  40. [40]

    Adam: A Method for Stochastic Optimization

    D. P . Kingma and J. Ba, “Adam: A method for stochastic optimization,” in ICLR (Poster) , 2015. [Online]. Available: http://arxiv.org/abs/1412.6980

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  41. [41]

    Solving inverse problems using data-driven models,

    S. Arridge, P . Maass, O. ¨Oktem, and C.-B. Sch¨ onlieb, “Solving inverse problems using data-driven models,” Acta Numerica , vol. 28, p. 1–174, 2019

    work page 2019

  42. [42]

    Learning-based signal detection for MIMO systems with unk nown noise statistics,

    K. He, L. He, L. Fan, Y . Deng, G. K. Karagiannidis, and A. N allanathan, “Learning-based signal detection for MIMO systems with unk nown noise statistics,” IEEE Transactions on Communications , vol. 69, no. 5, pp. 3025–3038, May 2021

    work page 2021

  43. [43]

    Hunger, Floating point operations in matrix-vector calculus

    R. Hunger, Floating point operations in matrix-vector calculus . Munich University of Technology, Inst. for Circuit Theory and Sign al, 2005

    work page 2005

  44. [44]

    On complexity analysis o f super- vised MLP-learning for algorithmic comparisons,

    E. Mizutani and S. E. Dreyfus, “On complexity analysis o f super- vised MLP-learning for algorithmic comparisons,” in IJCNN’01. In- ternational Joint Conference on Neural Networks. Proceedi ngs (Cat. No.01CH37222), vol. 1, Jul. 2001, pp. 347–352

    work page 2001

  45. [45]

    Hardware complexity analys is of deep neural networks and decision tree ensembles for real-time n eural data classification,

    M. Taghavi and M. Shoaran, “Hardware complexity analys is of deep neural networks and decision tree ensembles for real-time n eural data classification,” in 2019 9th International IEEE/EMBS Conference on Neural Engineering (NER) , Mar. 2019, pp. 407–410

    work page 2019

  46. [46]

    Study on 3D channel model for LTE , 3GPP TR 36.873 V12.7.0 Std., Jan. 2018

    work page 2018