%PDF-1.4
%
1 0 obj
<<
/Subtype/Type1
/Type/Font
/BaseFont/Helvetica
/Encoding/WinAnsiEncoding
>>
endobj
2 0 obj
<<
/Keywords(Behavioral modeling, Power amplifiers, Radio Frequency amplifiers, )
/ModDate(D:20240328170750+00'00')
/Creator(InControl Productions, Inc.)
/CreationDate(D:20200912062154-08'00')
/Subject(Tracking the nonlinear behavior of an RF power amplifier \(PA\) is challenging. To tackle this problem, we build a connection between residual learning and the PA nonlinearity, and propose a novel residual neural network structure, referred to as the residual real-valued time-delay neural network \(R2TDNN\). Instead of learning the whole behavior of the PA, the R2TDNN focuses on learning its nonlinear behavior by adding identity shortcut connections between the input and output layer. In particular, we apply the R2TDNN to digital predistortion and measure experimental results on a real PA. Compared with neural networks recently proposed by Liu et at. and Wang et at., the R2TDNN achieves the best linearization performance in terms of normalized mean square error and adjacent channel power ratio with less or similar computational complexity. Furthermore, the R2TDNN exhibits significantly faster training speed and lower training error.)
/Producer(PDFsharp 1.50.5147 \(www.pdfsharp.com\) \(Original: InControl Productions, Inc.; modified using iText 5.5.6 2000-2015 iText Group NV \(AGPL-version\)\))
/Author(Wu, Y., Gustavsson, U., Graell I Amat, A. et al)
/Title(Residual Neural Networks for Digital Predistortion)
>>
endobj
3 0 obj
<<
/Subtype/XML
/Type/Metadata
/Length 3739
>>
stream
SAC MLC4: Machine Learning for Communications
Residual Neural Networks for Digital Predistortion
Residual Neural Networks for Digital Predistortion
Yibo Wu, Ulf Gustavsson, Alexandre Graell i Amat and Henk Wymeersch
endstream
endobj
4 0 obj
<<
/Filter/FlateDecode
/Length 10
>>
stream
x+ |
endstream
endobj
5 0 obj
<<
/Filter/FlateDecode
/Length 188
>>
stream
xN0Ds
AV8RQi9`Tq=wKOg1XTp*lMQr'9$)suܒ e={ym걝8GS
mBej7C M<;
\ S.vR`etŽB
endstream
endobj
6 0 obj
<<
/Filter/FlateDecode
/Length 10
>>
stream
x+ |
endstream
endobj
7 0 obj
<<
/Filter/FlateDecode
/Length 187
>>
stream
xn0Ds,JcSz5J6ĕMmJü@?ib0=x&\Iۦ(Wm-
/wpe3"gmʜnh!=M3v:}&7Xo
@B
endstream
endobj
8 0 obj
<<
/Filter/FlateDecode
/Length 10
>>
stream
x+ |
endstream
endobj
9 0 obj
<<
/Filter/FlateDecode
/Length 187
>>
stream
xn0Ds,JSz5J6ĕMmJü@?ib0=)<TCMVm-
/wpe3"gmʜnh!=M3v:}&7Xo
@B
endstream
endobj
10 0 obj
<<
/Filter/FlateDecode
/Length 10
>>
stream
x+ |
endstream
endobj
11 0 obj
<<
/Filter/FlateDecode
/Length 187
>>
stream
xN0Ds
AV8RQi9`T1=wKOgyv PeSm5{8Y&I؍8D䖴@XFN/7;m|T>Th*Su]dwNhi䘂tIG+E#ƍB
endstream
endobj
12 0 obj
<<
/Filter/FlateDecode
/Length 10
>>
stream
x+ |
endstream
endobj
13 0 obj
<<
/Filter/FlateDecode
/Length 188
>>
stream
xN0Ds Tr+mK](4ojTq={Ї瓬=L?TX}asMN:15rIR5c&G/?%-ѫÍ{vzN>c3q0O4ڄf]o8->iv 8 ]<ҁ]B
endstream
endobj
14 0 obj
<<
/Filter/FlateDecode
/Length 10
>>
stream
x+ |
endstream
endobj
15 0 obj
<<
/Filter/FlateDecode
/Length 188
>>
stream
xMO0D+
(֖8RQi9v¯Es4LKGy
vUP(nDwK5/pұ.L2'[7q8x-i܍^ܳ/WQ=s`JSMLmVub:igc
%Qγ[
oj,iG-B
endstream
endobj
16 0 obj
<<
/Filter/FlateDecode
/Length 354
>>
stream
x]Kn E"h@֓O}H'8\4:*ME)gޟi\Ds NDI~հ,u\axee @L.J"+$dǪn?KlǢmj4u@ᶣVN 53D
cJ;@ݡcOmcQhS1+)I6sݓB;a!&U)Rbo}lە~Vj2ބeL.F:@8PN7H/
PT-R,J GuDጱ"9%DKF*ʃ3_vq5ۃ*I~ҿ
endstream
endobj
17 0 obj
<<
/Filter/FlateDecode
/Length 543
>>
stream
xMTI0{j0_"htAhe{m|D>*r)eU}?W;c{RAk'^tE\wJj8
=>88'hk[|O}:b
Guw b