Hyperparameter Optimization of Two-Hidden-Layer Neural Networks for Power Amplifiers Behavioral Modeling Using Genetic Algorithms

Wang, Siqi; Roger, Michel; Sarrazin, Julien; Lelandais-Perrault, Caroline

doi:10.1109/lmwc.2019.2950801

Cited by 33 publications

(27 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The procedure of this combined system is complex to explain for estimating TEC data that is not an issue in this study. A number of relevant studies exist (Edwards, 1989;Lin et al, 2001;Liu et al, 2004;Tsai and Tsai, 2004;Nishihashi et al, 2009;Pulinets and Boyarchuk, 2005;Liu et al, 2006;Klimenko et al, 2012;Lin, 2013;Chuo, 2014, Hayakawa, 2016Chatterjee et al, 2014;Nenovski et al, 2015;Thomas et al, 2017;Wang et al, 2017). TEC maps for three time periods are used in this study.…”

Section: Data Sourcementioning

confidence: 99%

Detecting Total Electron Content Precursors Before Earthquakes by Examining Total Electron Content Images Based on Butterworth Filter in Convolutional Neural Networks

Lin

Chiou

2020

IEEE Access

View full text Add to dashboard Cite

Taiwan Standard Time [TST]) as training images (inputs) were used to create two convolutional neural network (CNN) models. However, splitting the TEC maps of the three time periods into daily TEC images caused wedge effects. The wedge effects were reduced using a low-pass filter called the Butterworth filter. This resulted in clearer TEC precursors for earthquakes, facilitating the identification of earthquakes of magnitude M w ≥ 5.0 that exhibited associated TEC precursors during three periods, particularly for the Chi-Chi earthquake of September 21, 1999. The results of this study were compared with those of Lin et al. (2001) and Lin (2010) associated with the Chi-Chi earthquake. Simultaneously, two CNN models that were developed were verified to be rational due to the high accuracy of their predictions. These two models were used to verify each other's accuracies and to demonstrate the reliability of the method in this study. Therefore, statistical analysis was not the aim. The final outputs of the two CNN model were defined as similarities. Similarities, which are larger than 0.5, were defined as TEC precursors of earthquakes. TEC precursors described as temporal TEC multi-precursors (TTMPs) by Zoran et al. (2018) were detectable on the 1st, 3rd, and 4th days (that is, on September 17, 18, and 20, 1999, respectively) prior to the Chi-Chi earthquake of September 21, 1999. These results are consistent with those of Liu et al. (2001) and Lin (2010). A TEC precursor on May 13, 2003, (TST) was detectable 2 days prior to the earthquake on May 15, 2003, (TST) with the magnitude (M w) of 5.52. The low standard deviation (STD) and mean square error (MSE) confirm the reliability of both CNN models. Regarding mechanical principles, the TTMPs related to the Chi-Chi earthquake were caused by an electric field (Liu et al., 2001). The cause of the TEC precursor on May 13, 2003, prior to the earthquake on May 15, 2003, was an argument without any corresponding study for comparison.

show abstract

Section: Data Sourcementioning

confidence: 99%

Detecting Total Electron Content Precursors Before Earthquakes by Examining Total Electron Content Images Based on Butterworth Filter in Convolutional Neural Networks

Lin

Chiou

2020

IEEE Access

View full text Add to dashboard Cite

show abstract

“…According to [10] [14], the joint CFR/DPD has obvious advantage on complexity and linearization performance as well compared with the traditional CFR algorithm which is implemented independently of the DPD. Therefore we mainly focus on comparing the complexities of GMP-based CFRDPD in [21] and the proposed NN-CFRDPD.…”

Section: Complexity Analysismentioning

confidence: 99%

“…We set NN model with time delay L = 6 and N 1 = N 2 = 30 neurons in each hidden layer for all tests. These hyperparameters are determined by the algorithm proposed in [21]. The proposed NN-CFRDPD is then validated on the test bench with another segment of signal with 40000 samples.…”

Section: Nn-cfrdpdmentioning

confidence: 99%

“…That is because the GMP model structure is determined by the algorithm in [41] which optimizes both the model modeling accuracy and model complexity. On the other hand, the hyperparameters of NN model are determined by the algorithm in [21] which considers only average modeling accuracy since NN training signals are randomly partitioned and the corresponding performance is not a constant value for the same group of hyperparameters. Furthermore, we can see that the proposed AILC-based NN-CFRDPD reaches at least 1 dB better than the GMP-CFRDPD.…”

Section: Nn-cfrdpdmentioning

confidence: 99%

“…In [19], a DPD model based on real-valued focused time-delay neural network (RVFTDNN) has been experimentally validated for PA linearization considering memory effects. Different NN models have been proposed to model the nonlinear behavior of wideband PAs and transmitters, such as two hidden layers artificial neural network (2HLANN) [20] [21], distributed spatiotemporal NN based model (DSTNN) [22], complex-Chebyshev functional link neural network (CCFLNN) [23]. The RVFTDNN has also been proposed in [24] to compensate simultaneously for the nonlinearity of the transmitters and the I/Q imbalance of the modulator.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Augmented Iterative Learning Control for Neural-Network-Based Joint Crest Factor Reduction and Digital Predistortion of Power Amplifiers

Wang

Roger

Sarrazin

et al. 2020

IEEE Trans. Microwave Theory Techn.

Self Cite

View full text Add to dashboard Cite

Digital predistorsion (DPD) is a commonly used approach to compensate for the power amplifiers (PA) nonlinearities and memory effects as well as to improve its power efficiency. To alleviate the restriction brought by high peak-to-average power ratio (PAPR) of input signals, crest factor reduction (CFR) is needed for higher efficiency. In modern communication systems, power of transmitted signals gets lower, which makes complexities of CFR and DPD become non-negligible. This paper proposes a new approach to realize a joint CFR and DPD model using neural networks (NN). The modeling accuracy is guaranteed by a new proposed augmented iterative learning control (AILC) algorithm for the NN training signals. Compared with conventional ILC, the proposed AILC is shown more robust according to simulation and experimental results. The proposed AILC-based NN-CFRDPD is experimentally evaluated on different test benches.

show abstract

A genetic‐algorithm‐based synthesis of microwave integrated distributed amplifiers

Metel

Добуш

Kalentyev

et al. 2023

Int J Numerical Modelling

View full text Add to dashboard Cite

This article presents a genetic algorithm‐based technique for microwave integrated distributed amplifier synthesis. To speed‐up synthesis, only linear characteristics are used in the goal function. The technique uses MMIC foundry process design kit models providing manufacturing‐ready schematics to be found. The models from the process design kit are presented as S‐parameters again to enhance algorithm performance. The technique allows searching among most of the existing schematics of the microwave distributed amplifier but is not limited to typical topologies. The section number is theoretically unlimited and makes it possible to find principally new circuit topology. The synthesis shortens and simplifies the distributed amplifier design time. Experimental validation of the technique is carried out by synthesizing the broadband driver stage for the 20–30 GHz buffer amplifier. The technique produces several amplifier topologies meeting the requirements. The most appropriate circuit was manufactured using a commercial 0.25 μm GaAs pHEMT process. The manufactured buffer amplifier demonstrates a gain of 10 dB and 20 dBm output power at a 1 dB compression point.

show abstract

Hyperparameter Optimization of Two-Hidden-Layer Neural Networks for Power Amplifiers Behavioral Modeling Using Genetic Algorithms

Cited by 33 publications

References 14 publications

Detecting Total Electron Content Precursors Before Earthquakes by Examining Total Electron Content Images Based on Butterworth Filter in Convolutional Neural Networks

Detecting Total Electron Content Precursors Before Earthquakes by Examining Total Electron Content Images Based on Butterworth Filter in Convolutional Neural Networks

Augmented Iterative Learning Control for Neural-Network-Based Joint Crest Factor Reduction and Digital Predistortion of Power Amplifiers

A genetic‐algorithm‐based synthesis of microwave integrated distributed amplifiers

Contact Info

Product

Resources

About