Professional Documents
Culture Documents
Convolutional Neural Network For Behavioral
Convolutional Neural Network For Behavioral
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
To address the above issues and inspired by the emer- band low complexity model, a CNN-based architecture
gence of artificial intelligence (AI) in the broadband com- for extracting the PA behavioral model is proposed to
munications area, advanced NN-based models have been improve nonlinear modeling performance.
investigated [20]–[24]. Deep learning [25]–[28] in the AI 2) The CNN-based architectures were commonly used in
field has shown excellent performance in discovering com- image processing, where the input data are in a 2-D
plex nonlinear relationships using labeled data. In particular, format. However, in PA modeling and DPD, the input
convolutional NNs (CNNs) [29], [30] and recurrent NNs data have a 1-D format (varies only versus time). One
(RNNs) [31], [32] in deep learning have been proven to of the contributions of this article is in proposing a way
be effective in many fields, including wireless communi- to map the 1-D data to 2-D data to make the use of a
cation [33]–[38]. However, according to the results of our CNN-based architecture possible.
research, the work on the use of deep learning to build the 3) If RNN or CNN is used in the modeling and linearization
behavior models and linearize PAs [14], [39]–[42] is limited. of PAs, they have high computational complexity for
One of the important reasons is that the regression algorithm parameters training [43]. To reduce the computational
that is based on RNNs is often utilized for natural speech and complexity, a training methodology for PA modeling
time series processing tasks. If they are used for modeling based on transfer learning [44] is proposed to accelerate
and linearization of PAs, although they have fewer parameters the training of the PA model.
compared with feedforward NNs due to their characteristics of 4) As a result of all the above contributions, this work
weight sharing, the complex training algorithm seems to make achieved the best PA modeling and linearization per-
the method complicated [43]. In addition, CNN is usually used formance (normalized mean square error (NMSE) and
as a classifier, and the output layer makes a discrete decision ACPR) for nonlinear PAs driven with broadband signals
rather than outputting a continuous signal. However, in this with bandwidths higher than 100 MHz. In addition,
presented work, it will demonstrate for the first time that CNN the proposed modeling technique achieved the lowest
can be adapted and used in the fields of behavior modeling complexity and the fastest convergence time among all
and DPD synthesis of the PAs. The NN model’s complexity the other models used for the same application (PA
reduction will mainly result from the characteristic of weight modeling and DPD).
sharing in CNN structures [29], [30]. The aspect of increasing The remainder of this article is organized as follows.
the input dimension without changing the network structure In Section II, the existing NN models for PA modeling,
has attracted our attention. including shallow NN models and deep NN (DNN) models,
We first apply CNN to PA modeling and propose a real- are briefly reviewed. Section III proposes the structure of
valued time-delay CNN (RVTDCNN) behavior model for the RVTDCNN model and describes it in a detailed manner.
wideband wireless PA. As CNN cannot be directly used to Section IV discusses the training process of the RVTDCNN
build the PA model, since the input signal is in a 1-D format model and analyzes the model complexity of the RVTDCNN
that changes with time, the input data are sorted and arranged model. Section V extends the RVTDCNN model to DPD
as the graph composed of the in-phase and quadrature (I /Q) design. Section VI describes the platform for experimental
components and envelope-dependent terms of current and past validation. Section VII reports the measurement and validation
signals. Next, this model constructs a predesigned filter using results and compares the proposed model with other models.
the convolutional layer to extract the basis functions required Finally, Section VIII gives the conclusions.
for PA forward or reverse modeling. Finally, the extracted basis
functions are inputted into a simple, fully connected (FC) layer II. ANN S FOR PA M ODELING
to build the model. The model complexity of RVTDCNN is
significantly reduced due to the weight sharing characteris- A. Shallow Neural Networks for PA Modeling
tics of the convolution structure. Meanwhile, the extraction The shallow NNs, with fewer hidden layers, are used to
effect of the predesigned filter also reduces the training express the output characteristics of the PA due to its rela-
complexity of the model. In order to evaluate the model tively simple network structure and training process, as shown
performance of the RVTDCNN model, we compared the in Fig. 1(a) and (b) [15], [19].
RVTDCNN model with other existing models (including NN A commonly used shallow NN structure includes an input
and multilayer NN model) by experiment and simulation. The layer, a hidden layer structure with one or two layers, and an
results show that, compared with the existing state-of-the-art output layer. The model in Fig. 1(a) considers injecting the
models, the performance of the RVTDCNN model, especially I /Q components of the input signal and embeds their corre-
its complexity, is reduced in terms of the number of the sponding time-delayed values into the spatial structure of the
model’s coefficients. input layer of the network to reflect the corresponding memory
The contributions of this article are as follows. effects, such as real-valued time-delay NN (RVTDNN) model
1) As the signal bandwidth increases, PA exhibits compli- in [15]. However, hidden and related information, such as
cated nonlinear characteristics and memory effects. It is envelope-dependent terms, requires further network compu-
difficult to achieve good modeling performance with low tation capability, which leads to a complex network struc-
complexity-based traditional behavioral models [14]. ture and additional hidden layers. To this end, the structure
To address the problem of how to derive the broad- in Fig. 1(b) is proposed to simplify the network structure,
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
TABLE II
M ODELING P ERFORMANCE U NDER D IFFERENT C ONVOLUTION K ERNEL S CALES
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Leaky− ReLU(x) = max(λx, x) where K a and L a are the index for aligned signal and envelope;
exp(2x) − 1 K b , L b , and Mb are the index for signal and lagging envelope;
tanh(x) = . (10) and K c , L c , and Mc are the index for signal and leading
exp(2x) + 1
envelope.
The ARVTDNN in [19], the RVTDNN in [15], and the DNN
B. Complexity Analysis of RVTDCNN
model in [14] are all FC networks. The coefficient number of
The complexity analysis aims to evaluate the capability the FC networks can be obtained as follows:
of different models to assess if the training procedure of I1
RVTDCNN is simpler than the training of other typical PMNNs = Ni1 −1 + 1 × Ni1 (16)
models. In terms of the model complexity, it refers to both i 1 =2
the number of coefficients and the argument floating-point
operations (FLOPs). The comparison of complexity, including where Ni1 means the neurons of the i 1 th layer, and I1 means
the total number of coefficients of the network structure and the number of layers (I1 ≥ 3, including the input and output
layers).
the FLOPs, is shown in Table IV. Based on the theory and
experimental data, the proposed RVTDCNN in this article For the long short-term memory (LSTM) model in [41],
has superior performance than the traditional models for its the number of model coefficients of the LSTM layer is
convolution calculation. The following is a specific calculation PLSTM = 4I (Nin + I + 1) (17)
process for the complexity of RVTDCNN.
Based on RVTDCNN, it can be stated that the convolution where Nin means the input number of the LSTM layer at each
structure decreases the model size, which makes the extraction moment; I is the number of neurons in the LSTM layer.
of the model’s features more efficient. Compared to the Except for the total number of the coefficients, the argument
standard feedforward network, the coefficient number of the FLOPs are also introduced to assess the network complexity.
convolutional structure to generate the same feature is much For the convolutional process, considering the complexity of
less due to the weight sharing, which reduces the coefficient the activation function, the formula for calculating the number
complexity of the RVTDCNN model. The total number of of FLOPs can be derived as follows:
coefficients is equal to the sum of the weight number and the FLOPsconv = 2r sz × BCL + 13BCL. (18)
bias number between layers. The number of coefficients of the
predesigned filter layer can be calculated as follows: For the FC layer in the RVTDCNN model, the FLOPs can
be calculated as follows:
Pconv = Wconv + Bconv
= r ×s ×z ×L +L (11) FLOPsfc = 2(B × C × L × T ) + 13T. (19)
where the kernel size is r × s × z, and the number of kernels For the output layer in the RVTDCNN model, the FLOPs
is L. can be calculated as follows:
The coefficient number of the FC layer can be calculated FLOPsout = 2T Tout . (20)
as follows:
The FLOPs of the FC networks can be obtained as follows:
Pfc = Wfc + Bfc I1
= B×C ×L ×T +T (12) FLOPsMNNs = 2Ni1 −1 Ni1 + K i1 Ni1 (21)
i 1 =2
where B × C × L denotes the size of the output tensor of the
predesigned filter, and T is the number of neurons of the FC where K i1 is the FLOPs that calculate the activation function
layer. of the i 1 th layer.
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
TABLE IV
C OMPLEXITY C ALCULATIONS FOR D IFFERENT M ODELS
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
TABLE VI
C OMPARISON OF THE P ERFORMANCE B ETWEEN THE T RADITIONAL
T RAINING M ETHOD AND THE T RANSFER L EARNING M ETHOD
TABLE VII
M ODELING P ERFORMANCE AND C OMPLEXITY OF VARIOUS M ETHODS
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
TABLE VIII
D IFFERENT C ASES OF S IGNAL D ISTORTION AT T RANSMITTER
TABLE IX Fig. 12. Gain characteristics and phase characteristics of the PA output under
NMSE AND ACPR P ERFORMANCES OF THE RVTDCNN different bandwidths. (a) Gain characteristics. (b) Phase characteristics.
M ODEL U NDER D IFFERENT C ASES
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
TABLE X
M ODELING P ERFORMANCE AND C OMPLEXITY OF THE RVTDCNN M ODEL U NDER D IFFERENT I NPUT S IGNAL BANDWIDTHS
563. For the proposed model, the memory depth increases To simplify the derivation process, the coefficients of the
from 2 to 5, and the number of the model’s coefficients model, including ak ’s and ckq ’s, have been omitted in (25).
is 266, which can be considered as reasonable, and is half Thus, (25) can be expanded to be as follows:
the number of model coefficients of the ANN model. The
y(n) = (Iin (n) + j Q in (n)) + (Iin (n) + j Q in (n))|x(n)|
coefficient number of the LSTM model does not increase with
the signal bandwidth, but it is still about twice the coefficient + (Iin (n) + j Q in (n))|x(n)|2 + · · ·
number of the RVTDCNN model. + (Iin (n) + j Q in (n))|x(n − 1)|
Fig. 13 shows the output spectrum after the linearization + (Iin (n) + j Q in (n))|x(n − 1)|2 + · · ·
of the PA using the RVTDCNN model at the 200-MHz LTE + (Iin (n) + j Q in (n))|x(n − 2)|
signal. The results show that, under a wide signal bandwidth, + (Iin (n) + j Q in (n))|x(n − 2)|2 + · · ·
the RVTDCNN model still has a significant linearization effect = Iin (n) + Iin (n)|x(n)|+ Iin (n)|x(n)|2 + Iin (n)|x(n −1)|
on the PA. + Iin (n)|x(n − 1)|2 + Iin (n)|x(n − 2)|
+ Iin (n)|x(n − 2)|2
VIII. C ONCLUSION
+ j Q in (n) + Q in (n)|x(n)| + Q in (n)|x(n)|2
In this article, the RVTDCNN is proposed for modeling the + Q in (n)|x(n − 1)| + Q in (n)|x(n − 1)|2
nonlinear and memory effects of wideband PA. RVTDCNN + Q in (n)|x(n − 2)| + Q in (n)|x(n − 2)|2
extracts the effective features of the 2-D input graph data with
+ other terms. (26)
a convolutional structure. Doherty PA, with the LTE signals
from 40 to 200 MHz, is tested to verify the effectiveness of If these items in (26) can be fitted in the predesigned filter,
the RVTDCNN model. For the PA with 100-MHz input signal the predesigned filter can generate rich basis functions, and the
under different cases, the NMSE can reach about −36 dB, modeling performance of the proposed model can be achieved.
with an ACPR around −46 dBc with DPD. The results show In the predesigned filter, the 3*3 convolution kernels are used
that the RVTDCNN still has a good modeling effect when to extract features from the input signal. In the convolution
there are I /Q imbalance and dc offset, which verifies that process, the stride is set to be 1 until the input tensors are all
the proposed model has strong adaptability. Compared with convolved. The convolution process is shown in Fig. 4. The
the existing shallow NN and DNN, in terms of the number convolution output under different convolution steps represents
of model coefficients and FLOPs, the proposed RVTDCNN is the capture of different local features of the input. We took
verified to reduce the number of model coefficients by more out the local input convoluted in the first step
than 50% under different signal bandwidths.
h 11 = Iin (n) + Q in (n) + |x(n)| + Iin (n − 1)
A PPENDIX + Q in (n − 1) + |x(n − 1)| + Iin (n − 2)
E XTRACTION E FFECT OF P REDESIGNED + Q in (n − 2) + |x(n − 2)| + b (27)
F ILTER ON BASIS F UNCTION
where b is the bias. For the convenience of calculation,
Because the predesigned filter can completely capture the
the corresponding coefficient is ignored in the formula.
basis function features required for modeling, the number of
This article deduced the item in (26) through the above
neurons in the FC layer required is significantly reduced,
local input, and the other input term with memory effect can
thereby reducing the complexity of the model. To verify
be deduced through the same method. The convolution output
that the predesigned filter can generate rich basis functions,
described above is input into a nonlinear activation function
we introduce a baseband PA model [2], which is expressed as
to obtain the output of the predesigned filter. The output u 11
follows:
can be expressed as follows:
K −1 K −1 Q−1
y(n) = ak x(n)|x(n)| +k
ckq x(n)|x(n − q)|k (25) u 11 = tanh(h 11 )
k=0 k=1 q=1 = tanh(Iin (n) + Q in (n) + |x(n)| + Iin (n − 1)
where K and Q are the nonlinearity order and the lagging + Q in (n − 1) + |x(n − 1)| + Iin (n − 2)
cross terms index, respectively; ak and ckq are the coefficients
+ Q in (n − 2) + |x(n − 2)| + b) (28)
of the model. For simplicity, the memory terms in (25) have
been omitted. where tanh(·) is the activation function given.
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
The hyperbolic tangent sigmoid function can be expanded [2] Y. Liu, W. Pan, S. Shao, and Y. Tang, “A general digital predistor-
by the Taylor series, and the approximate output is described tion architecture using constrained feedback bandwidth for wideband
power amplifiers,” IEEE Trans. Microw. Theory Techn., vol. 63, no. 5,
as follows: pp. 1544–1555, May 2015.
1 2 π [3] A. Katz, J. Wood, and D. Chokola, “The evolution of PA linearization:
tanh(x) = x − x 3 + x 5 + − · · · |x| < . (29) From classic feedforward and feedback through analog and digital
3 15 2 predistortion,” IEEE Microw. Mag., vol. 17, no. 2, pp. 32–40, Feb. 2016.
Bring (27) into x 3 , while considering the critical correlation [4] Z. Popovic, “Amping up the PA for 5G: Efficient GaN power ampli-
items related to the justification process, ignored the irrelevant, fiers with dynamic supplies,” IEEE Microw. Mag., vol. 18, no. 3,
pp. 137–149, May 2017.
redundant terms. The result will be shown in the following: [5] X. Hu, T. Liu, Z. Liu, W. Wang, and F. M. Ghannouchi, “A novel single
feedback architecture with time-interleaved sampling for multi-band
h 311 = (Iin (n) + Q in (n) + |x(n)| + Iin (n − 1) + Q in (n − 1) DPD,” IEEE Commun. Lett., vol. 23, no. 6, pp. 1033–1036, Jun. 2019.
+|x(n −1)|)+ Iin (n −2)+ Q in (n −2)+|x(n −2)|+b)3 [6] J. Reina-Tosina, M. Allegue-Martinez, C. Crespo-Cadenas, C. Yu, and
S. Cruces, “Behavioral modeling and predistortion of power amplifiers
= 6b Iin (n)|x(n)| + 3Iin (n)|x(n)|2 + 6b Iin (n)|x(n − 1)| under sparsity hypothesis,” IEEE Trans. Microw. Theory Techn., vol. 63,
+ 3Iin (n)|x(n − 1)|2 + 6b Iin (n)|x(n − 2)| no. 2, pp. 745–753, Feb. 2015.
[7] Q. Zhang, W. Chen, and Z. Feng, “Reduced cost digital predistortion
+ 3Iin (n)|x(n − 2)|2 +6b Q in (n)|x(n)|+3Q in (n)|x(n)|2 only with in-phase feedback signal,” IEEE Microw. Wireless Compon.
Lett., vol. 28, no. 3, pp. 257–259, Mar. 2018.
+ 6b Q in (n)|x(n − 1)| + 3Q in (n)|x(n − 1)|2
[8] J. Joung, C. K. Ho, K. Adachi, and S. Sun, “A survey on power-
+ 6b Q in (n)|x(n − 2)| + 3Q in (n)|x(n − 2)|2 amplifier-centric techniques for spectrum- and energy-efficient wire-
less communications,” IEEE Commun. Surveys Tuts., vol. 17, no. 1,
+ other terms. (30) pp. 315–333, 1st Quart., 2015.
[9] A. Cheaito, M. Crussière, J.-F. Hélard, and Y. Louët, “Quantifying
In summary, we combine (28)–(30) while omitting both the the memory effects of power amplifiers: EVM closed-form derivations
irrelevant, redundant terms and multiplication factors. The out- of multicarrier signals,” IEEE Wireless Commun. Lett., vol. 6, no. 1,
put of the predesigned filter can be expressed approximately pp. 34–37, Feb. 2017.
[10] S. Wang, M. Roger, and C. Lelandais-Perrault, “Impacts of crest factor
using by the following: reduction and digital predistortion on linearity and power efficiency of
power amplifiers,” IEEE Trans. Circuits Syst. II, Exp. Briefs, vol. 66,
u 11 = tanh(h 11 ) = h 11 + h 311 + other terms no. 3, pp. 407–411, Mar. 2019.
= (Iin (n) + Q in (n) + |x(n)| + Iin (n − 1) + Q in (n − 1) [11] J. Cai, C. Yu, L. Sun, S. Chen, and J. B. King, “Dynamic behav-
ioral modeling of RF power amplifier based on time-delay support
+ |x(n −1)|)+ Iin (n −2)+ Q in (n − 2) + |x(n − 2)|) vector regression,” IEEE Trans. Microw. Theory Techn., vol. 67, no. 2,
+ b + Iin (n)|x(n)| + Iin (n)|x(n)|2 + Iin (n)|x(n − 1)| pp. 533–543, Feb. 2019.
[12] R. N. Braithwaite, “Digital predistortion of an RF power amplifier using
+ Iin (n)|x(n − 1)|2 + Iin (n)|x(n − 2)| a reduced volterra series model with a memory polynomial estimator,”
IEEE Trans. Microw. Theory Techn., vol. 65, no. 10, pp. 3613–3623,
+ Iin (n)|x(n − 2)|2 + Q in (n)|x(n)| Oct. 2017.
+ Q in (n)|x(n)|2 + Q in (n)|x(n − 1)| [13] A. Rahati Belabad, S. A. Motamedi, and S. Sharifian, “A novel gen-
eralized parallel two-box structure for behavior modeling and digital
+ Q in (n)|x(n − 1)|2 + Q in (n)|x(n − 2)| predistortion of RF power amplifiers at LTE applications,” Circuits, Syst.,
Signal Process., vol. 37, no. 7, pp. 2714–2735, Jul. 2018.
+ Q in (n)|x(n − 2)|2
[14] R. Hongyo, Y. Egashira, T. M. Hone, and K. Yamaguchi, “Deep neural
+ other terms. (31) network-based digital predistorter for Doherty power amplifiers,” IEEE
Microw. Wireless Compon. Lett., vol. 29, no. 2, pp. 146–148, Feb. 2019.
By arranging (31), (31) can be rewritten as follows: [15] T. Liu, S. Boumaiza, and F. M. Ghannouchi, “Dynamic behav-
ioral modeling of 3G power amplifiers using real-valued time-delay
u 11 neural networks,” IEEE Trans. Microw. Theory Techn., vol. 52, no. 3,
pp. 1025–1033, Mar. 2004.
= Iin (n) + Iin (n)|x(n)| + Iin (n)|x(n)|2
[16] M. Rawat, K. Rawat, and F. M. Ghannouchi, “Adaptive digital predistor-
+ Iin (n)|x(n − 1)| + Iin (n)|x(n − 1)|2 + Iin (n)|x(n − 2)| tion of wireless power amplifiers/transmitters using dynamic real-valued
focused time-delay line neural networks,” IEEE Trans. Microw. Theory
+ Iin (n)|x(n − 2)|2 Techn., vol. 58, no. 1, pp. 95–104, Jan. 2010.
+ Q in (n)+ Q in (n)|x(n)|+ Q in (n)|x(n)|2 + Q in (n)|x(n −1)| [17] M. Rawat and F. M. Ghannouchi, “A mutual distortion and impair-
ment compensator for wideband direct-conversion transmitters using
+ Q in (n)|x(n − 1)|2 + Q in (n)|x(n − 2)| neural networks,” IEEE Trans. Broadcast., vol. 58, no. 2, pp. 168–177,
Jun. 2012.
+ Q in (n)|x(n − 2)|2 + other terms. (32) [18] M. Hui, T. Liu, M. Zhang, Y. Ye, D. Shen, and X. Ying, “Aug-
mented radial basis function neural network predistorter for linearisa-
Comparing (32) and (26), we can find that the linear term, tion of wideband power amplifiers,” Electron. Lett., vol. 50, no. 12,
the nonlinear term, and the lagging cross-terms in (26) both pp. 877–879, Jun. 2014.
exist in (32). Therefore, the terms produced by the predesigned [19] D. Wang, M. Aziz, M. Helaoui, and F. M. Ghannouchi, “Augmented
filter correspond to the terms of the polynomials, and more real-valued time-delay neural network for compensation of distortions
and impairments in wireless transmitters,” IEEE Trans. Neural Netw.
terms can be provided. Namely, the predesigned filter can Learn. Syst., vol. 30, no. 1, pp. 242–254, Jan. 2019.
generate enough rich basis set to get an excellent performance. [20] X. You, C. Zhang, X. Tan, S. Jin, and H. Wu, “AI for 5G: Research
directions and paradigms,” Sci. China Inf. Sci., vol. 62, no. 2, p. 21301,
Feb. 2019.
R EFERENCES
[21] Z. M. Fadlullah, B. Mao, F. Tang, and N. Kato, “Value iteration
[1] V. Camarchia, M. Pirola, R. Quaglia, S. Jee, Y. Cho, and B. Kim, “The architecture based deep learning for intelligent routing exploiting het-
Doherty power amplifier: Review of recent solutions and trends,” IEEE erogeneous computing platforms,” IEEE Trans. Comput., vol. 68, no. 6,
Trans. Microw. Theory Techn., vol. 63, no. 2, pp. 559–571, Feb. 2015. pp. 939–950, Jun. 2019.
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
[22] Y. Wang, M. Liu, J. Yang, and G. Gui, “Data-driven deep learning for [45] D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,”
automatic modulation recognition in cognitive radios,” IEEE Trans. Veh. in Proc. 3rd Int. Conf. Learn. Represent. (ICLR), San Diego, CA, USA,
Technol., vol. 68, no. 4, pp. 4074–4077, Apr. 2019. 2015, pp. 7–9.
[23] Z. M. Fadlullah et al., “State-of-the-art deep learning: Evolving machine [46] S. Haykin, Neural Networks: A Comprehensive Foundation.
intelligence toward tomorrow’s intelligent network traffic control sys- Upper Saddle River, NJ, USA: Prentice-Hall, 1999.
tems,” IEEE Commun. Surveys Tuts., vol. 19, no. 4, pp. 2432–2455, [47] D. R. Morgan, Z. Ma, J. Kim, M. G. Zierdt, and J. Pastalan, “A gener-
4th Quart., 2017. alized memory polynomial model for digital predistortion of RF power
[24] N. Kato et al., “The deep learning vision for heterogeneous network amplifiers,” IEEE Trans. Signal Process., vol. 54, no. 10, pp. 3852–3860,
traffic control: Proposal, challenges, and future perspective,” IEEE Oct. 2006.
Wireless Commun., vol. 24, no. 3, pp. 146–153, Jun. 2017.
[25] J. Schmidhuber, “Deep learning in neural networks: An overview,”
Neural Netw., vol. 61, pp. 85–117, Jan. 2015.
[26] W. Liu, Z. Wang, X. Liu, N. Zeng, Y. Liu, and F. E. Alsaadi, “A
survey of deep neural network architectures and their applications,”
Neurocomputing, vol. 234, pp. 11–26, Apr. 2017.
[27] M. Robnik-Sikonja, “Data generators for learning systems based on
RBF networks,” IEEE Trans. Neural Netw. Learn. Syst., vol. 27, no. 5,
pp. 926–938, May 2016. Xin Hu (Senior Member, IEEE) received the B.Sc.
[28] Z. Liu, X. Hu, T. Liu, X. Li, W. Wang, and F. M. Ghannouchi, degree in electrical information engineering from the
“Attention-based deep neural network behavioral model for wideband Huazhong University of Science and Technology,
wireless power amplifiers,” IEEE Microw. Wireless Compon. Lett., Wuhan, China, in 2007, and the Ph.D. degree in
vol. 30, no. 1, pp. 82–85, Jan. 2020. electrical information engineering from the Institute
[29] Y. Lecun, L. Bottou, Y. Bengio, and P. Haffner, “Gradient-based learn- of Electrics, Chinese Academy of Sciences, Beijing,
ing applied to document recognition,” Proc. IEEE, vol. 86, no. 11, China, in 2012.
pp. 2278–2324, Nov. 1998. From 2012 to 2016, he was a Senior Engineer with
[30] C. Szegedy et al., “Going deeper with convolutions,” in Proc. IEEE Aerospace Science and Technology Corporation.
Conf. Comput. Vis. Pattern Recognit. (CVPR), Jun. 2015, pp. 1–9. In 2016, he joined the Electronics Technology Lab-
[31] A. Graves, A.-R. Mohamed, and G. Hinton, “Speech recognition with oratory, Beijing University of Posts and Telecommu-
deep recurrent neural networks,” in Proc. IEEE Int. Conf. Acoust., nications, Beijing, where he is currently an Associate Professor. From 2019 to
Speech Signal Process., May 2013, pp. 6645–6649. 2020, he was a Visiting Scholar with the iRadio Laboratory, University of
[32] M. Schuster and K. K. Paliwal, “Bidirectional recurrent neural net- Calgary, Calgary, AB, Canada, where he has been an Adjunct Researcher,
works,” IEEE Trans. Signal Process., vol. 45, no. 11, pp. 2673–2681, since 2019. His current research interests include digital predistortion of
Nov. 1997. nonlinear power amplifiers and the application of signal processing techniques
[33] G. Gui, H. Huang, Y. Song, and H. Sari, “Deep learning for an effective to RF and microwave problems, dynamic wireless resource management.
nonorthogonal multiple access scheme,” IEEE Trans. Veh. Technol., Dr. Hu received the National Natural Science Foundation of China in 2018.
vol. 67, no. 9, pp. 8440–8450, Sep. 2018. He was involved in several projects funded by the National High Technology
[34] H. Huang, Y. Song, J. Yang, G. Gui, and F. Adachi, “Deep-learning- Research and Development Program of China and the National Natural
based millimeter-wave massive MIMO for hybrid precoding,” IEEE Science Foundation of China.
Trans. Veh. Technol., vol. 68, no. 3, pp. 3027–3032, Mar. 2019.
[35] F. Meng, P. Chen, L. Wu, and X. Wang, “Automatic modulation
classification: A deep learning enabled approach,” IEEE Trans. Veh.
Technol., vol. 67, no. 11, pp. 10760–10772, Nov. 2018.
[36] F. Tang, B. Mao, Z. M. Fadlullah, and N. Kato, “On a novel deep-
learning-based intelligent partially overlapping channel assignment in
SDN-IoT,” IEEE Commun. Mag., vol. 56, no. 9, pp. 80–86, Sep. 2018.
[37] Z. Md. Fadlullah, F. Tang, B. Mao, J. Liu, and N. Kato, “On intelligent
traffic control for large-scale heterogeneous networks: A value matrix- Zhijun Liu received the B.Sc. degree in electron-
based deep learning approach,” IEEE Commun. Lett., vol. 22, no. 12, ics science and technology from the Nanjing Uni-
pp. 2479–2482, Dec. 2018. versity of Posts and Telecommunications, Nanjing,
China, in 2017. He is currently pursuing the Ph.D.
[38] M. Liu, T. Song, J. Hu, J. Yang, and G. Gui, “Deep learning-inspired
degree in electronics science and technology with
message passing algorithm for efficient resource allocation in cognitive
the Beijing University of Posts and Telecommunica-
radio networks,” IEEE Trans. Veh. Technol., vol. 68, no. 1, pp. 641–653,
tions, Beijing, China.
Jan. 2019.
His research interests include digital predistortion
[39] P. Chen, S. Alsahali, A. Alt, J. Lees, and P. J. Tasker, “Behavioral
of nonlinear power amplifiers and the application
modeling of GaN power amplifiers using long short-term memory
of deep learning techniques to RF and microwave
networks,” in Proc. Int. Workshop Integr. Nonlinear Microw. Millimetre-
problems.
Wave Circuits (INMMIC), Brive La Gaillarde, France, Jul. 2018, pp. 1–3.
[40] J. Sun, W. Shi, Z. Yang, J. Yang, and G. Gui, “Behavioral modeling and
linearization of wideband RF power amplifiers using BiLSTM networks
for 5G wireless systems,” IEEE Trans. Veh. Technol., vol. 68, no. 11,
pp. 10348–10356, Nov. 2019, doi: 10.1109/TVT.2019.2925562.
[41] D. Phartiyal and M. Rawat, “LSTM-deep neural networks based pre-
distortion linearizer for high power amplifiers,” in Proc. Nat. Conf.
Commun. (NCC), Bangalore, India, Feb. 2019, pp. 1–5.
[42] T. J. Liu et al., “Digital predistortion linearization with deep neural Xiaofei Yu received the B.Sc. degree in electronics
networks for 5G power amplifiers,” in Proc. Eur. Microw. Conf. Central science and technology from the Beijing University
Eur. (EuMCE), May 2019, pp. 216–219. of Posts and Telecommunications, Beijing, China,
[43] F. Mkadem and S. Boumaiza, “Physically inspired neural network model in 2018, where she is currently pursuing the M.Sc.
for RF power amplifier behavioral modeling and digital predistortion,” degree in electronics science and technology.
IEEE Trans. Microw. Theory Techn., vol. 59, no. 4, pp. 913–923, Her research interests include digital predistortion
Apr. 2011. of nonlinear power amplifiers and the application of
[44] H. B. Ammar, E. Eaton, J. M. Luna, and P. Ruvolo, “Autonomous cross- deep learning techniques to RF and crowdsensing.
domain knowledge transfer in lifelong policy gradient reinforcement
learning,” in Proc. Int. Joint Conf. Artif. Intell. (AAAI), Jan. 2015,
pp. 3345–3351.
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Yulong Zhao was born in Henan, China, in 1983. Xiang Li (Member, IEEE) received the B.Sc.
He received the B.S. degree in applied physics and M.Sc. degrees in electronic engineering from
and the M.S. degree in electronic engineering from Tsinghua University, Beijing, China, in 2009 and
Xidian University, Xi’an, China, in 2006 and 2009, 2012, respectively, and the Ph.D. degree in electrical
respectively. He is currently pursuing the Ph.D. and computer engineering from the University of
degree with the University of Calgary, Calgary, AB, Calgary, Calgary, AB, Canada, in 2018.
Canada. He is currently working as a Post-Doctoral Fel-
From 2009 to 2015, he was a Microwave Engi- low with the University of Calgary. His research
neer and Project Leader with the ZTE Corporation. interests include multiband/wide-band power ampli-
He was with the Radio Remote Unit Development fier design, outphasing transmitter design, and
Department, where he designed high-power ampli- MMIC power amplifiers for wireless and satellite
fiers for 3G and 4G wireless communications. He is currently with the communication.
Intelligent RF Radio Technology Laboratory, University of Calgary. His Dr. Li was the recipient of the Student Paper Award of the 2010 Asia-Pacific
current research interests include high-efficiency and wideband RF PAs, Microwave Conference (APMC).
MMIC PAs, microwave passive components, and digital predistortion.
Wenhua Chen (Senior Member, IEEE) received Mohamed Helaoui (Senior Member, IEEE)
the B.S. degree in microwave engineering from the received the M.Sc. degree in communications and
University of Electronic Science and Technology information technology from école Supérieure des
of China (UESTC), Chengdu, China, in 2001, and Communications de Tunis, Tunisia, in 2003, and
the Ph.D. degree in electronic engineering from the Ph.D. degree in electrical engineering from
Tsinghua University, Beijing, China, in 2006. the University of Calgary, Calgary, AB, Canada,
From 2010 to 2011, he was a Post-Doctoral Fellow in 2008.
with the Intelligent RF Radio Laboratory(iRadio He is currently an Associate Professor with
Lab), University of Calgary, Calgary, AB, Canada. the Department of Electrical And Computer
He is currently an Associate Professor with the Engineering, University of Calgary. His research
Department of Electronic Engineering, Tsinghua activities have led to over 60 publications. He holds
University. He has authored or coauthored more than 200 journals and seven patents (pending). His current research interests include digital signal
conference papers. His current research interests include energy-efficient processing, power efficiency enhancement for wireless transmitters, switching
power amplifier (PA) design and linearization, millimeter-wave and terahertz mode power amplifiers, and advanced transceiver design for software-defined
integrated circuits and systems. radio and millimeter-wave applications.
Dr. Chen was a recipient of the 2015 Outstanding Youth Science Foundation Dr. Helaoui is a member of the COMMTTAP Chapter in the IEEE
of NSFC, the 2014 URSI Young Scientist Award, and the Student Paper Southern Alberta Section.
Awards of several international conferences. He is an Associate Editor of
the IEEE T RANSACTIONS ON M ICROWAVE T HEORY AND T ECHNIQUES .
Authorized licensed use limited to: Syracuse University Libraries. Downloaded on June 15,2021 at 13:38:50 UTC from IEEE Xplore. Restrictions apply.