As an agricultural innovation, low-temperature plasma technology is an environmentally friendly green technology that increases crop quality and productivity. However, there is a lack of research on the identification of plasma-treated rice growth. Although traditional convolutional neural networks (CNN) can automatically share convolution kernels and extract features, the outputs are only suitable for entry-level categorization. Indeed, shortcuts from the bottom layers to fully connected layers can be established feasibly in order to utilize spatial and local information from the bottom layers, which contain small distinctions necessary for fine-grain identification. In this work, 5000 original images which contain the basic growth information of rice (including plasma treated rice and the control rice) at the tillering stage were collected. An efficient multiscale shortcut CNN (MSCNN) model utilizing key information and cross-layer features was proposed. The results show that MSCNN outperforms the mainstream models in terms of accuracy, recall, precision and F1 score with 92.64%, 90.87%, 92.88% and 92.69%, respectively. Finally, the ablation experiment, comparing the average precision of MSCNN with and without shortcuts, revealed that the MSCNN with three shortcuts achieved the best performance with the highest precision.
Citation: Wenzhuo Chen, Yuan Wang, Xiaojiang Tang, Pengfei Yan, Xin Liu, Lianfeng Lin, Guannan Shi, Eric Robert, Feng Huang. A specific fine-grained identification model for plasma-treated rice growth using multiscale shortcut convolutional neural network[J]. Mathematical Biosciences and Engineering, 2023, 20(6): 10223-10243. doi: 10.3934/mbe.2023448
[1] | Lina Liu, Huiting Zhang, Yinlan Chen . The generalized inverse eigenvalue problem of Hamiltonian matrices and its approximation. AIMS Mathematics, 2021, 6(9): 9886-9898. doi: 10.3934/math.2021574 |
[2] | Shixian Ren, Yu Zhang, Ziqiang Wang . An efficient spectral-Galerkin method for a new Steklov eigenvalue problem in inverse scattering. AIMS Mathematics, 2022, 7(5): 7528-7551. doi: 10.3934/math.2022423 |
[3] | Yalçın Güldü, Ebru Mişe . On Dirac operator with boundary and transmission conditions depending Herglotz-Nevanlinna type function. AIMS Mathematics, 2021, 6(4): 3686-3702. doi: 10.3934/math.2021219 |
[4] | Batirkhan Turmetov, Valery Karachik . On solvability of some inverse problems for a nonlocal fourth-order parabolic equation with multiple involution. AIMS Mathematics, 2024, 9(3): 6832-6849. doi: 10.3934/math.2024333 |
[5] | Wei Ma, Zhenhao Li, Yuxin Zhang . A two-step Ulm-Chebyshev-like Cayley transform method for inverse eigenvalue problems with multiple eigenvalues. AIMS Mathematics, 2024, 9(8): 22986-23011. doi: 10.3934/math.20241117 |
[6] | Liangkun Xu, Hai Bi . A multigrid discretization scheme of discontinuous Galerkin method for the Steklov-Lamé eigenproblem. AIMS Mathematics, 2023, 8(6): 14207-14231. doi: 10.3934/math.2023727 |
[7] | Lingling Sun, Hai Bi, Yidu Yang . A posteriori error estimates of mixed discontinuous Galerkin method for a class of Stokes eigenvalue problems. AIMS Mathematics, 2023, 8(9): 21270-21297. doi: 10.3934/math.20231084 |
[8] | Jia Tang, Yajun Xie . The generalized conjugate direction method for solving quadratic inverse eigenvalue problems over generalized skew Hamiltonian matrices with a submatrix constraint. AIMS Mathematics, 2020, 5(4): 3664-3681. doi: 10.3934/math.2020237 |
[9] | Phakhinkon Phunphayap, Prapanpong Pongsriiam . Extremal orders and races between palindromes in different bases. AIMS Mathematics, 2022, 7(2): 2237-2254. doi: 10.3934/math.2022127 |
[10] | Hannah Blasiyus, D. K. Sheena Christy . Two-dimensional array grammars in palindromic languages. AIMS Mathematics, 2024, 9(7): 17305-17318. doi: 10.3934/math.2024841 |
As an agricultural innovation, low-temperature plasma technology is an environmentally friendly green technology that increases crop quality and productivity. However, there is a lack of research on the identification of plasma-treated rice growth. Although traditional convolutional neural networks (CNN) can automatically share convolution kernels and extract features, the outputs are only suitable for entry-level categorization. Indeed, shortcuts from the bottom layers to fully connected layers can be established feasibly in order to utilize spatial and local information from the bottom layers, which contain small distinctions necessary for fine-grain identification. In this work, 5000 original images which contain the basic growth information of rice (including plasma treated rice and the control rice) at the tillering stage were collected. An efficient multiscale shortcut CNN (MSCNN) model utilizing key information and cross-layer features was proposed. The results show that MSCNN outperforms the mainstream models in terms of accuracy, recall, precision and F1 score with 92.64%, 90.87%, 92.88% and 92.69%, respectively. Finally, the ablation experiment, comparing the average precision of MSCNN with and without shortcuts, revealed that the MSCNN with three shortcuts achieved the best performance with the highest precision.
Since the 1960s, the rapid development of high-speed rail has made it a very important means of transportation. However, the vibration will be caused because of the contact between the wheels of the train and the train tracks during the operation of the high-speed train. Therefore, the analytical vibration model can be mathematically summarized as a quadratic palindromic eigenvalue problem (QPEP) (see [1,2])
(λ2A1+λA0+A⊤1)x=0, |
with Ai∈Rn×n, i=0,1 and A⊤0=A0. The eigenvalues λ, the corresponding eigenvectors x are relevant to the vibration frequencies and the shapes of the vibration, respectively. Many scholars have put forward many effective methods to solve QPEP [3,4,5,6,7]. In addition, under mild assumptions, the quadratic palindromic eigenvalue problem can be converted to the following linear palindromic eigenvalue problem (see [8])
Ax=λA⊤x, | (1) |
with A∈Rn×n is a given matrix, λ∈C and nonzero vectors x∈Cn are the wanted eigenvalues and eigenvectors of the vibration model. We can obtain 1λx⊤A⊤=x⊤A by transposing the equation (1). Thus, λ and 1λ always come in pairs. Many methods have been proposed to solve the palindromic eigenvalue problem such as URV-decomposition based structured method [9], QR-like algorithm [10], structure-preserving methods [11], and palindromic doubling algorithm [12].
On the other hand, the modal data obtained by the mathematical model are often evidently different from the relevant experimental ones because of the complexity of the structure and inevitable factors of the actual model. Therefore, the coefficient matrices need to be modified so that the updated model satisfies the dynamic equation and closely matches the experimental data. Al-Ammari [13] considered the inverse quadratic palindromic eigenvalue problem. Batzke and Mehl [14] studied the inverse eigenvalue problem for T-palindromic matrix polynomials excluding the case that both +1 and −1 are eigenvalues. Zhao et al. [15] updated ∗-palindromic quadratic systems with no spill-over. However, the linear inverse palindromic eigenvalue problem has not been extensively considered in recent years.
In this work, we just consider the linear inverse palindromic eigenvalue problem (IPEP). It can be stated as the following problem:
Problem IPEP. Given a pair of matrices (Λ,X) in the form
Λ=diag{λ1,⋯,λp}∈Cp×p, |
and
X=[x1,⋯,xp]∈Cn×p, |
where diagonal elements of Λ are all distinct, X is of full column rank p, and both Λ and X are closed under complex conjugation in the sense that λ2i=ˉλ2i−1∈C, x2i=ˉx2i−1∈Cn for i=1,⋯,m, and λj∈R, xj∈Rn for j=2m+1,⋯,p, find a real-valued matrix A that satisfy the equation
AX=A⊤XΛ. | (2) |
Namely, each pair (λt,xt), t=1,⋯,p, is an eigenpair of the matrix pencil
P(λ)=Ax−λA⊤x. |
It is known that the mathematical model is a "good" representation of the system, we hope to find a model that is closest to the original model. Therefore, we consider the following best approximation problem:
Problem BAP. Given ˜A∈Rn×n, find ˆA∈SA such that
‖ˆA−˜A‖=minA∈SA‖A−˜A‖, | (3) |
where ‖⋅‖ is the Frobenius norm, and SA is the solution set of Problem IPEP.
In this paper, we will put forward a new direct method to solve Problem IPEP and Problem BAP. By partitioning the matrix Λ and using the QR-decomposition, the expression of the general solution of Problem IPEP is derived. Also, we show that the best approximation solution ˆA of Problem BAP is unique and derive an explicit formula for it.
We first rearrange the matrix Λ as
Λ=[10 00Λ1 000 Λ2]t2s2(k+2l) t 2s2(k+2l), | (4) |
where t+2s+2(k+2l)=p, t=0 or 1,
Λ1=diag{λ1,λ2,⋯,λ2s−1,λ2s},λi∈R, λ−12i−1=λ2i, 1≤i≤s,Λ2=diag{δ1,⋯,δk,δk+1,δk+2,⋯,δk+2l−1,δk+2l}, δj∈C2×2, |
with
δj=[αj+βji00αj−βji], i=√−1, 1≤j≤k+2l,δ−1j=ˉδj, 1≤j≤k,δ−1k+2j−1=δk+2j, 1≤j≤l, |
and the adjustment of the column vectors of X corresponds to those of Λ.
Define Tp as
Tp=diag{It+2s,1√2[1−i1i],⋯,1√2[1−i1i]}∈Cp×p, | (5) |
where i=√−1. It is easy to verify that THpTp=Ip. Using this matrix of (5), we obtain
˜Λ=THpΛTp=[1000Λ1000˜Λ2], | (6) |
˜X=XTp=[xt,⋯,xt+2s,√2yt+2s+1,√2zt+2s+1,⋯,√2yp−1,√2zp−1], | (7) |
where
˜Λ2=diag{[α1β1−β1α1],⋯,[αk+2lβk+2l−βk+2lαk+2l]}≜diag{˜δ1,⋯,˜δk+2l}, |
and ˜Λ2∈R2(k+2l)×2(k+2l), ˜X∈Rn×p. yt+2s+j and zt+2s+j are, respectively, the real part and imaginary part of the complex vector xt+2s+j for j=1,3,⋯,2(k+2l)−1. Using (6) and (7), the matrix equation (2) is equivalent to
A˜X=A⊤˜X˜Λ. | (8) |
Since rank(X)=rank(˜X)=p. Now, let the QR-decomposition of ˜X be
˜X=Q[R0], | (9) |
where Q=[Q1,Q2]∈Rn×n is an orthogonal matrix and R∈Rp×p is nonsingular. Let
Q⊤AQ=[A11A12A21A22]pn−p p n−p. | (10) |
Using (9) and (10), then the equation of (8) is equivalent to
A11R=A⊤11R˜Λ, | (11) |
A21R=A⊤12R˜Λ. | (12) |
Write
R⊤A11R≜F=[f11F12 F13F21F22 F23F31F32 F33]t2s2(k+2l) t 2s 2(k+2l), | (13) |
then the equation of (11) is equivalent to
F12=F⊤21Λ1, F21=F⊤12, | (14) |
F13=F⊤31˜Λ2, F31=F⊤13, | (15) |
F23=F⊤32˜Λ2, F32=F⊤23Λ1, | (16) |
F22=F⊤22Λ1, | (17) |
F33=F⊤33˜Λ2. | (18) |
Because the elements of Λ1,˜Λ2 are distinct, we can obtain the following relations by Eqs (14)-(18)
F12=0, F21=0, F13=0, F31=0, F23=0, F32=0, | (19) |
F22=diag{[0h1λ1h10],⋯,[0hsλ2s−1hs0]}, | (20) |
F33=diag{G1,⋯,Gk,[0Gk+1G⊤k+1˜δk+10],⋯,[0Gk+lG⊤k+l˜δk+2l−10]}, | (21) |
where
Gi=aiBi, Gk+j=ak+2j−1D1+ak+2jD2, G⊤k+j=Gk+j,Bi=[11−αiβi−1−αiβi1], D1=[100−1], D2=[0110], |
and 1≤i≤k,1≤j≤l. h1,⋯,hs,a1,⋯,ak+2l are arbitrary real numbers. It follows from Eq (12) that
A21=A⊤12E, | (22) |
where E=R˜ΛR−1.
Theorem 1. Suppose that Λ=diag{λ1,⋯,λp}∈Cp×p, X=[x1,⋯,xp]∈Cn×p, where diagonal elements of Λ are all distinct, X is of full column rank p, and both Λ and X are closed under complex conjugation in the sense that λ2i=ˉλ2i−1∈C, x2i=ˉx2i−1∈Cn for i=1,⋯,m, and λj∈R, xj∈Rn for j=2m+1,⋯,p. Rearrange the matrix Λ as (4), and adjust the column vectors of X with corresponding to those of Λ. Let Λ,X transform into ˜Λ,˜X by (6)−(7) and QR-decomposition of the matrix ˜X be given by (9). Then the general solution of (2) can be expressed as
SA={A|A=Q[R−⊤[f11000F22000F33]R−1A12A⊤12EA22]Q⊤}, | (23) |
where E=R˜ΛR−1, f11 is arbitrary real number, A12∈Rp×(n−p),A22∈R(n−p)×(n−p) are arbitrary real-valued matrices and F22,F33 are given by (20)−(21).
In order to solve Problem BAP, we need the following lemma.
Lemma 1. [16] Let A,B be two real matrices, and X be an unknown variable matrix. Then
∂tr(BX)∂X=B⊤, ∂tr(X⊤B⊤)∂X=B⊤, ∂tr(AXBX)∂X=(BXA+AXB)⊤,∂tr(AX⊤BX⊤)∂X=BX⊤A+AX⊤B, ∂tr(AXBX⊤)∂X=AXB+A⊤XB⊤. |
By Theorem 1, we can obtain the explicit representation of the solution set SA. It is easy to verify that SA is a closed convex subset of Rn×n×Rn×n. By the best approximation theorem (see Ref. [17]), we know that there exists a unique solution of Problem BAP. In the following we will seek the unique solution ˆA in SA. For the given matrix ˜A∈Rn×n, write
Q⊤˜AQ=[˜A11˜A12˜A21˜A22]pn−p p n−p, | (24) |
then
‖A−˜A‖2=‖[R−⊤[f11000F22000F33]R−1−˜A11A12−˜A12A⊤12E−˜A21A22−˜A22]‖2=‖R−⊤[f11000F22000F33]R−1−˜A11‖2+‖A12−˜A12‖2+‖A⊤12E−˜A21‖2+‖A22−˜A22‖2. |
Therefore, ‖A−˜A‖=min if and only if
‖R−⊤[f11000F22000F33]R−1−˜A11‖2=min, | (25) |
‖A12−˜A12‖2+‖A⊤12E−˜A21‖2=min, | (26) |
A22=˜A22. | (27) |
Let
R−1=[R1R2R3], | (28) |
then the relation of (25) is equivalent to
‖R⊤1f11R1+R⊤2F22R2+R⊤3F33R3−˜A11‖2=min. | (29) |
Write
R1=[r1,t], R2=[r2,1⋮r2,2s], R3=[r3,1⋮r3,k+2l], | (30) |
where r1,t∈Rt×p,r2,i∈R1×p,r3,j∈R2×p, i=1,⋯,2s, j=1,⋯,k+2l.
Let
{Jt=r⊤1,tr1,t,Jt+i=λ2i−1r⊤2,2ir2,2i−1+r⊤2,2i−1r2,2i (1≤i≤s),Jr+i=r⊤3,iBir3,i (1≤i≤k),Jr+k+2i−1=r⊤3,k+2iD1˜δk+2i−1r3,k+2i−1+r⊤3,k+2i−1D1r3,k+2i (1≤i≤l),Jr+k+2i=r⊤3,k+2iD2˜δk+2i−1r3,k+2i−1+r⊤3,k+2i−1D2r3,k+2i (1≤i≤l), | (31) |
with r=t+s,q=t+s+k+2l. Then the relation of (29) is equivalent to
g(f11,h1,⋯,hs,a1,⋯,ak+2l)=‖f11Jt+h1Jt+1+⋯+hsJr+a1Jr+1+⋯+ak+2lJq−˜A11‖2=min, |
that is,
g(f11,h1,⋯,hs,a1,⋯,ak+2l)=tr[(f11Jt+h1Jt+1+⋯+hsJr+a1Jr+1+⋯+ak+2lJq−˜A11)⊤(f11Jt+h1Jt+1+⋯+hsJr+a1Jr+1+⋯+ak+2lJq−˜A11)]=f211ct,t+2f11h1ct,t+1+⋯+2f11hsct,r+2f11a1ct,r+1+⋯+2f11ak+2lct,q−2f11et+h21ct+1,t+1+⋯+2h1hsct+1,r+2h1a1ct+1,r+1+⋯+2h1ak+2lct+1,q−2h1et+1+⋯+h2scr,r+2hsa1cr,r+1+⋯+2hsak+2lcr,q−2hser+a21cr+1,r+1+⋯+2a1ak+2lcr+1,q−2a1er+1+⋯+a2k+2lcq,q−2ak+2leq+tr(˜A⊤11˜A11), |
where ci,j=tr(J⊤iJj),ei=tr(J⊤i˜A11)(i,j=t,⋯,t+s+k+2l) and ci,j=cj,i.
Consequently,
∂g(f11,h1,⋯,hs,a1,⋯,ak+2l)∂f11=2f11ct,t+2h1ct,t+1+⋯+2hsct,r+2a1ct,r+1+⋯+2ak+2lct,q−2et,∂g(f11,h1,⋯,hs,a1,⋯,ak+2l)∂h1=2f11ct+1,t+2h1ct+1,t+1+⋯+2hsct+1,r+2a1ct+1,r+1+⋯+2ak+2lct+1,q−2et+1,⋯∂g(f11,h1,⋯,hs,a1,⋯,ak+2l)∂hs=2f11cr,t+2h1cr,t+1+⋯+2hscr,r+2a1cr,r+1+⋯+2ak+2lcr,q−2er,∂g(f11,h1,⋯,hs,a1,⋯,ak+2l)∂a1=2f11cr+1,t+2h1cr+1,t+1+⋯+2hscr+1,r+2a1cr+1,r+1+⋯+2ak+2lcr+1,q−2er+1,⋯∂g(f11,h1,⋯,hs,a1,⋯,ak+2l)∂ak+2l=2f11cq,t+2h1cq,t+1+⋯+2hscq,r+2a1cq,r+1+⋯+2ak+2lcq,q−2eq. |
Clearly, g(f11,h1,⋯,hs,a1,⋯,ak+2l)=min if and only if
∂g(f11,h1,⋯,hs,a1,⋯,ak+2l)∂f11=0,⋯,∂g(f11,h1,⋯,hs,a1,⋯,ak+2l)∂ak+2l=0. |
Therefore,
f11ct,t+h1ct,t+1+⋯+hsct,r+a1ct,r+1+⋯+ak+2lct,q=et,f11ct+1,t+h1ct+1,t+1+⋯+hsct+1,r+a1ct+1,r+1+⋯+ak+2lct+1,q=et+1,⋯f11cr,t+h1cr,t+1+⋯+hscr,r+a1cr,r+1+⋯+ak+2lcr,q=er,f11cr+1,t+h1cr+1,t+1+⋯+hscr+1,r+a1cr+1,r+1+⋯+ak+2lcr+1,q=er+1,⋯f11cq,t+h1cq,t+1+⋯+hscq,r+a1cq,r+1+⋯+ak+2lcq,q=eq. | (32) |
If let
C=[ct,tct,t+1⋯ct,rct,r+1⋯ct,qct+1,tct+1,t+1⋯ct+1,rct+1,r+1⋯ct+1,q⋮⋮⋮⋮⋮cr,tcr,t+1⋯cr,rcr,r+1⋯cr,qcr+1,tcr+1,t+1⋯cr+1,rcr+1,r+1⋯cr+1,q⋮⋮⋮⋮⋮cq,tcq,t+1⋯cq,rcq,r+1⋯cq,q], h=[f11h1⋮hsa1⋮ak+2l], e=[etet+1⋮erer+1⋮eq], |
where C is symmetric matrix. Then the equation (32) is equivalent to
Ch=e, | (33) |
and the solution of the equation (33) is
h=C−1e. | (34) |
Substituting (34) into (20)-(21), we can obtain f11,F22 and F33 explicitly. Similarly, the equation of (26) is equivalent to
g(A12)=tr(A⊤12A12)+tr(˜A⊤12˜A12)−2tr(A⊤12˜A12)+tr(E⊤A12A⊤12E)+tr(˜A⊤21˜A21)−2tr(E⊤A12˜A21). |
Applying Lemma 1, we obtain
∂g(A12)∂A12=2A12−2˜A12+2EE⊤A12−2E˜A⊤21, |
setting ∂g(A12)∂A12=0, we obtain
A12=(Ip+EE⊤)−1(˜A12+E˜A⊤21), | (35) |
Theorem 2. Given ˜A∈Rn×n, then the Problem BAP has a unique solution and the unique solution of Problem BAP is
ˆA=Q[R−⊤[f11000F22000F33]R−1A12A⊤12E˜A22]Q⊤, | (36) |
where E=R˜ΛR−1, F22,F33,A12,˜A22 are given by (20),(21),(35),(24) and f11,h1,⋯,hs,a1,⋯,ak+2l are given by (34).
Based on Theorems 1 and 2, we can describe an algorithm for solving Problem BAP as follows.
Algorithm 1.
1) Input matrices Λ, X and ˜A;
2) Rearrange Λ as (4), and adjust the column vectors of X with corresponding to those of Λ;
3) Form the unitary transformation matrix Tp by (5);
4) Compute real-valued matrices ˜Λ,˜X by (6) and (7);
5) Compute the QR-decomposition of ˜X by (9);
6) F12=0,F21=0,F13=0,F31=0,F23=0,F32=0 by (19) and E=R˜ΛR−1;
7) Compute ˜Aij=Q⊤i˜AQj,i,j=1,2;
8) Compute R−1 by (28) to form R1,R2,R3;
9) Divide matrices R1,R2,R3 by (30) to form r1,t,r2,i,r3,j, i=1,⋯,2s,j=1,⋯,k+2l;
10) Compute Ji, i=t,⋯,t+s+k+2l, by (31);
11) Compute ci,j=tr(J⊤iJj),ei=tr(J⊤i˜A11), i,j=t,⋯,t+s+k+2l;
12) Compute f11,h1,⋯,hs,a1,⋯,ak+2l by (34);
13) Compute F22,F33 by (20), (21) and A22=˜A22;
14) Compute A12 by (35) and A21 by (22);
15) Compute the matrix ˆA by (36).
Example 1. Consider a 11-DOF system, where
˜A=[96.189818.184751.325049.086413.197364.911562.561981.762858.704531.110226.22120.463426.380340.180848.925394.205173.172278.022779.483120.774292.338060.284377.491014.55397.596733.771995.613564.77468.112664.431830.124643.020771.121681.730313.606923.991690.005457.520945.092492.938637.860947.092318.481622.174786.869586.929212.331936.92475.978054.700977.571381.158023.048890.488111.74188.443657.970518.390811.120323.478029.632148.679253.282684.430997.974829.667639.978354.986023.995378.025235.315974.469343.585935.072719.476443.887031.877825.987014.495541.726738.973982.119418.895544.678493.900222.592211.111942.416780.006885.30314.965424.16911.540368.677530.634987.594317.070825.806550.785843.141462.205590.271640.39124.302418.351150.850955.015622.766440.87208.551691.064835.095294.47879.645516.899036.848551.077262.247543.569959.489626.2482], |
the measured eigenvalue and eigenvector matrices Λ and X are given by
Λ=diag{1.0000, −1.8969, −0.5272, −0.1131+0.9936i,−0.1131−0.9936i,1.9228+2.7256i, 1.9228−2.7256i, 0.1728−0.2450i, 0.1728+0.2450i}, |
and
X=[−0.0132−1.00000.17530.0840+0.4722i0.0840−0.4722i−0.09550.39370.1196−0.3302−0.1892i−0.3302+0.1892i−0.19920.5220−0.04010.3930−0.2908i0.3930+0.2908i0.07400.02870.6295−0.3587−0.3507i−0.3587+0.3507i0.4425−0.3609−0.57450.4544−0.3119i0.4544+0.3119i0.4544−0.3192−0.2461−0.3002−0.1267i−0.3002+0.1267i0.25970.33630.9046−0.2398−0.0134i−0.2398+0.0134i0.11400.09660.08710.1508+0.0275i0.1508−0.0275i−0.0914−0.0356−0.2387−0.1890−0.0492i−0.1890+0.0492i0.24310.5428−1.00000.6652+0.3348i0.6652−0.3348i1.0000−0.24580.2430−0.2434+0.6061i−0.2434−0.6061i0.6669+0.2418i0.6669−0.2418i0.2556−0.1080i0.2556+0.1080i−0.1172−0.0674i−0.1172+0.0674i−0.5506−0.1209i−0.5506+0.1209i0.5597−0.2765i0.5597+0.2765i−0.3308+0.1936i−0.3308−0.1936i−0.7217−0.0566i−0.7217+0.0566i−0.7306−0.2136i−0.7306+0.2136i0.0909+0.0713i0.0909−0.0713i0.5577+0.1291i0.5577−0.1291i0.1867+0.0254i0.1867−0.0254i0.2866+0.1427i0.2866−0.1427i−0.5311−0.1165i−0.5311+0.1165i−0.3873−0.1096i−0.3873+0.1096i0.2624+0.0114i0.2624−0.0114i−0.6438+0.2188i−0.6438−0.2188i−0.0619−0.1504i−0.0619+0.1504i0.2787−0.2166i0.2787+0.2166i0.3294−0.1718i0.3294+0.1718i0.9333+0.0667i0.9333−0.0667i−0.4812+0.5188i−0.4812−0.5188i0.6483−0.1950i0.6483+0.1950i]. |
Using Algorithm 1, we obtain the unique solution of Problem BAP as follows:
ˆA=[34.256341.782433.357333.629823.806442.077050.064137.570531.090848.616919.097218.856135.225235.959244.350231.991855.292055.305254.379331.390960.834516.954029.63597.680519.124917.718316.708240.063618.291649.943737.691315.60274.960358.878251.490647.897435.698545.688956.043453.090856.540255.512038.344735.889433.408746.96359.776741.421551.446652.105865.672460.12935.806162.013916.523131.658051.235924.797865.556761.784062.549458.936374.709952.210555.853244.392519.296151.233322.428056.934042.634845.845356.372961.555531.683667.952540.201241.279671.382134.414033.281777.439360.894432.1411108.505649.607819.835185.743464.089057.652419.128025.039439.052466.774020.902348.851214.469518.928424.834837.255032.325438.353459.735833.590254.026550.777070.201165.415958.072040.065228.130114.76388.950720.096325.590759.694030.855866.878130.480723.610712.9984], |
and
‖ˆAX−ˆA⊤XΛ‖=8.2431×10−13. |
Therefore, the new model ˆAX=ˆA⊤XΛ reproduces the prescribed eigenvalues (the diagonal elements of the matrix Λ) and eigenvectors (the column vectors of the matrix X).
Example 2. (Example 4.1 of [12]) Given α=cos(θ), β=sin(θ) with θ=0.62 and λ1=0.2,λ2=0.3,λ3=0.4. Let
J0=[02ΓI2I2], Js=[03diag{λ1,λ2,λ3}I303], |
where Γ=[α−ββα]. We construct
˜A=[J000Js], |
the measured eigenvalue and eigenvector matrices Λ and X are given by
Λ=diag{5,0.2,0.8139+0.5810i,0.8139−0.5810i}, |
and
X=[−0.41550.6875−0.2157−0.4824i−0.2157+0.4824i−0.4224−0.3148−0.3752+0.1610i−0.3752−0.1610i−0.0703−0.6302−0.5950−0.4050i−0.5950+0.4050i−1.0000−0.46670.2293−0.1045i0.2293+0.1045i0.26500.3051−0.2253+0.7115i−0.2253−0.7115i0.9030−0.23270.4862−0.3311i0.4862+0.3311i−0.67420.31320.5521−0.0430i0.5521+0.0430i0.63580.1172−0.0623−0.0341i−0.0623+0.0341i−0.4119−0.27680.1575+0.4333i0.1575−0.4333i−0.20621.0000−0.1779−0.0784i−0.1779+0.0784i]. |
Using Algorithm 1, we obtain the unique solution of Problem BAP as follows:
ˆA=[−0.1169−0.23660.6172−0.7195−0.08360.28840.0092−0.0490−0.02020.0171−0.0114−0.09570.14620.61940.3738−0.16370.1291−0.00710.09720.12470.7607−0.04970.5803−0.03460.09790.29590.0937−0.10600.1323−0.0339−0.01090.6740−0.30130.73400.1942−0.08720.00540.00510.02970.08140.17830.22830.26430.03870.0986−0.3125−0.02920.2926−0.0717−0.05460.09530.10270.03600.2668−0.24180.12060.1406−0.05510.30710.2097−0.0106−0.23190.1946−0.0298−0.19350.0158−0.08860.0216−0.05600.24840.10440.12850.19020.22770.69610.16570.0728−0.0262−0.0831−0.00010.09060.00210.0764−0.12640.21440.6703−0.08500.0764−0.0104−0.0149−0.12450.08130.1952−0.07840.0760−0.08750.7978−0.00930.0206−0.1182], |
and
‖ˆAX−ˆA⊤XΛ‖=1.7538×10−8. |
Therefore, the new model ˆAX=ˆA⊤XΛ reproduces the prescribed eigenvalues (the diagonal elements of the matrix Λ) and eigenvectors (the column vectors of the matrix X).
In this paper, we have developed a direct method to solve the linear inverse palindromic eigenvalue problem by partitioning the matrix Λ and using the QR-decomposition. The explicit best approximation solution is given. The numerical examples show that the proposed method is straightforward and easy to implement.
The authors declare no conflict of interest.
[1] |
A. M. Khaneghah, L. M. Martins, A. M. Von Hertwig, R. Bertoldo, A. S. Sant'Ana, Deoxynivalenol and its masked forms: Characteristics, incidence, control and fate during wheat and wheat-based products processing - A review, Trends Food Sci. Technol., 71 (2018), 13−24. https://doi.org/10.1016/j.tifs.2017.10.012 doi: 10.1016/j.tifs.2017.10.012
![]() |
[2] |
N.S. Poluxeni, M. Sotirios, K. Chrysanthi, S. Panagiotis, H. Luc, Chemical pesticides and human health: the urgent need for a new concept in agriculture, Front. Public Health, 4 (2016) 148−148. https://doi.org/10.3389/fpubh.2016.00148 doi: 10.3389/fpubh.2016.00148
![]() |
[3] |
X. Lei, R. Qiu, Evaluation of food security in China based entropy TOPSIS model and the diagnosis of its obstacle factors, J. China Agric. Univ., 27 (2022), 1−14. https://doi.org/10.11841/j.issn.1007-4333.2022.12.01 doi: 10.11841/j.issn.1007-4333.2022.12.01
![]() |
[4] |
Y. T. Hui, D. C. Wang, Y. You, C. Y. Shao, C. S. Zhong, H. D. Wang, Effect of low temperature plasma treatment on biological characteristics and yield components of wheat seeds (Triticum aestivum L.), Plasma Chem. Plasma Process., 40 (2020), 1555−1570. https://doi.org/10.1007/s11090-020-10104-z doi: 10.1007/s11090-020-10104-z
![]() |
[5] |
H. Liu, Y. H. Zhang, H. Yin, W. X. Wang, X. M. Zhao, Y. G. Du, Alginate oligosaccharides enhanced triticum aestivum L. tolerance to drought stress, Plant Physiol. Biochem., 62 (2013), 33−40. https://doi.org/10.1016/j.plaphy.2012.10.012 doi: 10.1016/j.plaphy.2012.10.012
![]() |
[6] |
B. Šerá, P. r Špatenka, M. l Šerý, N. Vrchotová, I. a Hrušková, Influence of plasma treatment on wheat and oat germination and early growth, IEEE Trans. Plasma Sci., 38 (2010), 2963−2968. https://doi.org/10.1109/TPS.2010.2060728 doi: 10.1109/TPS.2010.2060728
![]() |
[7] |
R. Thirumdas, A. Kothakota, U. Annapure, K. Siliveru, R. Blundell, R. Gatt, et al., Plasma activated water (PAW) Chemistry, physico-chemical properties, applications in food and agriculture, Trends Food Sci. Technol., 77 (2018), 21−31. https://doi.org/10.1016/j.tifs.2018.05.007 doi: 10.1016/j.tifs.2018.05.007
![]() |
[8] | L. Tonks, Oscillations in ionized gases, Plasma and Oscillations, Elsevier, 1961,122−139. https://doi.org/10.1016/B978-1-4831-9913-9.50014-5 |
[9] |
B. Zhao, J. S. Feng, X. Wu, S. C. Yan, A survey on deep learning-based fine-grained object classification and semantic segmentation, Int. J. Autom. Comput., 14 (2017), 119−135. https://doi.org/10.1007/s11633-017-1053-3 doi: 10.1007/s11633-017-1053-3
![]() |
[10] | A. Srivastava, E. Han, V. Kumar, V. Singh, Parallel formulations of decision-tree classification algorithms, High Performance Data Mining, Springer, Boston, 1999,237−261. https://doi.org/10.1007/0-306-47011-X_2 |
[11] | G. D. Guo, H. Wang, D. Bell, Y. X. Bi, KNN model-based approach in classification, in OTM Confederated International Conferences CoopIS, DOA, and ODBASE, (2003), 986−996. https://doi.org/10.1007/978-3-540-39964-3_62 |
[12] |
A. Tharwat, A. E. Hassanien, B. E. Elnaghi, A BA-based algorithm for parameter optimization of Support Vector Machine, Pattern Recognit. Lett., 93 (2017), 13−22. https://doi.org/10.1016/j.patrec.2016.10.007 doi: 10.1016/j.patrec.2016.10.007
![]() |
[13] | N. Coskun, T. Yildirim, The effects of training algorithms in MLP network on image classification, in Proceedings of the International Joint Conference on Neural Networks, (2003), 1223−1226. |
[14] | J. Deng, J. Krause, F. F. Li, Fine-grained crowdsourcing for fine-grained recognition, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, (2013), 580−587. https://doi.org/10.1109/CVPR.2013.81 |
[15] | E. Gavves, B. Fernando, C. G. Snoek, A. W. Smeulders, T. Tuytelaars, Fine-grained categorization by alignments, in Proceedings of the IEEE International Conference on Computer Vision, (2013), 1713−1720. https://doi.org/10.1109/ICCV.2013.215 |
[16] | K. M. He, X. Y. Zhang, S. Q. Ren, J. Sun, Deep residual learning for image recognition, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, (2016), 770−778. https://doi.org/10.1109/CVPR.2016.90 |
[17] | G. Huang, Z. Liu, L. Van Der Maaten, K. Q. Weinberger, Densely connected convolutional networks, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, (2017), 4700−4708. https://doi.org/10.1109/cvpr.2017.243 |
[18] | S. Ioffe, C. Szegedy, Batch normalization: Accelerating deep network training by reducing internal covariate shift, in Proceedings of the 32nd International Conference on Machine Learning, (2015), 448−456. |
[19] |
A. Krizhevsky, I. Sutskever, G. E. Hinton, ImageNet classification with deep convolutional neural networks, Commun. ACM, 60 (2017), 84-90. https://doi.org/10.1145/3065386 doi: 10.1145/3065386
![]() |
[20] |
S. Jin, H. X. Yao, X. S. Sun, S. C. Zhou, L. Zhang, X. S. Hua, Deep saliency hashing for fine-grained retrieval, IEEE Trans. Image Process., 29 (2020), 5336−5351. https://doi.org/10.1109/TIP.2020.2971105 doi: 10.1109/TIP.2020.2971105
![]() |
[21] |
Y. Jing, W. Wang, L. Wang, T. N. Tan, Learning aligned image-text representations using graph attentive relational net-work, IEEE Trans. Image Process., 30 (2021), 1840−1852. https://doi.org/10.1109/TIP.2020.3048627 doi: 10.1109/TIP.2020.3048627
![]() |
[22] |
L. L. Zhang, J. Liu, M. N. Luo, X. J. Chang, Q. H. Zheng, Deep semisupervised zero-shot learning with maximum mean discrepancy, Neural Comput., 30 (2018), 1426−1447. https://doi.org/10.1162/neco_a_01071 doi: 10.1162/neco_a_01071
![]() |
[23] |
K. Liu, D. Liu, L. Li, N. Yan, H. Q. Li, Semantics-to-signal scalable image compression with learned revertible representations, Int. J. Comput. Vis., 129 (2021), 2605−2621. https://doi.org/10.1007/s11263-021-01491-7 doi: 10.1007/s11263-021-01491-7
![]() |
[24] |
L. Qi, X. Q. Lu, X. L. Li, Exploiting spatial relation for fine-grained image classification, Pattern Recognit., 91 (2019), 47−55. https://doi.org/10.1016/j.patcog.2019.02.007 doi: 10.1016/j.patcog.2019.02.007
![]() |
[25] |
L. Wang, K. He, X. Feng, X. T. Ma, Multilayer feature fusion with parallel convolutional block for fine-grained image classification, Appl. Intell., 52 (2022), 2872−2883. https://doi.org/10.1007/s10489-021-02573-2 doi: 10.1007/s10489-021-02573-2
![]() |
[26] | M. Srinivas, Y. Y. Lin, H. Y. M. Liao, Deep dictionary learning for fine-grained image classification, in 2017 IEEE International Conference on Image Processing, (2017), 835−839. https://doi.org/10.1109/ICIP.2017.8296398 |
[27] | L. Liao, R. M. Hu, J. Xiao, Q. Wang, J. Xiao, J. Chen, Exploiting effects of parts in fine-grained categorization of vehicles, in 2015 IEEE international conference on image processing, (2015), 745−749. https://doi.org/10.1109/ICIP.2015.7350898 |
[28] |
K. Wang, M. Z. Liu, YOLOv3-MTis A YOLOv3 using multi-target tracking for vehicle visual detection, Appl. Intell., 52 (2022), 2070−2091. https://doi.org/10.1007/s10489-021-02491-3 doi: 10.1007/s10489-021-02491-3
![]() |
[29] |
S. M. Pan, W. Q. Feng, Y. W. Chong, Attribute-guided global and part-level identity network for person re-identification, Int. J. Pattern Recognit. Artif. Intell., 36 (2022), 2250011. https://doi.org/S0218001422500112 doi: 10.1142/S0218001422500112
![]() |
[30] |
C. Wang, J. Y. Sun, S. W. Ma, Y. Q. Lu, W. Liu, Multi-stream network for human-object interaction detection, Int. J. Pattern Recognit. Artif. Intell., 35 (2021), 2150025. https://doi.org/10.1142/S0218001421500257 doi: 10.1142/S0218001421500257
![]() |
[31] |
Z. Q. Lin, S. M. Mu, F. Huang, K. A. Mateen, M. J. Wang, W. L. Gao, et al., A unified matrix-based convolutional neural network for fine-grained image classification of wheat leaf diseases, IEEE Access, 7 (2019), 11570−11590. https://doi.org/10.1109/ACCESS.2019.2891739 doi: 10.1109/ACCESS.2019.2891739
![]() |
[32] |
Z. Q. Lin, S. M. Mu, A. J. Shi, C. Pang, X. X. Sun, A novel method of maize leaf disease image identification based on a multichannel convolutional neural network, Trans. ASABE, 61 (2018), 1461−1474. https://doi.org/10.13031/trans.12440 doi: 10.13031/trans.12440
![]() |
[33] | H. Lu, Z. G. Cao, Y. Xiao, Z. W. Fang, Y. J. Zhu, Fine-grained maize cultivar identification using filter-specific convolutional activations, in 2016 IEEE International Conference on Image Processing, (2016), 3718−3722. https://doi.org/10.1109/ICIP.2016.7533054 |
[34] |
X. P. Zhang, H. K. Xiong, W. G. Zhou, Q. Tian, Fused one-vs-all features with semantic alignments for fine-grained visual categorization, IEEE Trans. Image Process., 25 (2015), 878−892. https://doi.org/10.1109/TIP.2015.2509425 doi: 10.1109/TIP.2015.2509425
![]() |
[35] |
X. S. Wei, C. W. Xie, J. X. Wu, C. H. Shen, Mask-CNN is Localizing parts and selecting descriptors for fine-grained bird species categorization, Pattern Recognit., 76 (2018), 704−714. https://doi.org/10.1016/j.patcog.2017.10.002 doi: 10.1016/j.patcog.2017.10.002
![]() |
[36] |
L. Qi, X. Q. Lu, X. L. Li, Exploiting spatial relation for fine-grained image classification, Pattern Recognit., 91 (2019), 47−55. https://doi.org/10.1016/j.patcog.2019.02.007 doi: 10.1016/j.patcog.2019.02.007
![]() |
[37] |
Y. Zhang, X. S. Wei, J. X. Wu, J. F. Cai, J. B. Lu, V. A. Nguyen, et al., Weakly supervised fine-grained categorization with part-based image representation, IEEE Trans. Image Process., 25 (2016), 1713−1725. https://doi.org/10.1109/TIP.2016.2531289 doi: 10.1109/TIP.2016.2531289
![]() |
[38] | S. L. Huang, Z. Xu, D. C. Tao, Y. Zhang, Part-Stacked CNN for fine-grained visual categorization, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, (2016), 1173−1182. https://doi.org/10.1109/CVPR.2016.132 |
[39] |
S. H. Lee, C. S. Chan, S. J. Mayo, P. Remagnino, How deep learning extracts and learns leaf features for plant classification, Pattern Recognit., 71 (2017), 1−13. https://doi.org/10.1016/j.patcog.2017.05.015 doi: 10.1016/j.patcog.2017.05.015
![]() |
[40] |
M. Rohrbach, A. Rohrbach, M. Regneri, S. Amin, M. Andriluka, M. Pinkal, et al., Recognizing fine-grained and composite activities using hand-centric features and script data, Int. J. Comput. Vision., 119 (2016), 346−373. https://doi.org/10.1007/s11263-015-0851-8 doi: 10.1007/s11263-015-0851-8
![]() |
[41] | S. Cai, W. Zuo, Z. Lei, Higher-order integration of hierarchical convolutional activations for fine-grained visual categorization, in Proceedings of the IEEE International Conference on Computer Vision, (2017), 511−520. |
[42] |
Q. Hu, H. Wang, T. Li, C. Shen, Deep CNNs with spatially weighted pooling for fine-grained car recognition, IEEE. Intell Transp., 91 (2019), 47−55. https://doi.org/10.1016/j.patcog.2019.02.007 doi: 10.1016/j.patcog.2019.02.007
![]() |
[43] | P. J. Burt, E. H. Adelson, Readings in computer vision, Elsevier, Piscataway, 1987,671−679. https://doi.org/10.1016/B978-0-08-051581-6.50065-9 |
[44] |
C. Farabet, C. Couprie, L. Najman, Y. LeCun, Learning hierarchical features for scene labeling, IEEE Trans. Pattern Anal. Mach. Intell., 35 (2012), 1915−1929. https://doi.org/10.1109/TPAMI.2012.231 doi: 10.1109/TPAMI.2012.231
![]() |
[45] | B. Hariharan, P. Arbeláez, R. Girshick, J. Malik, Hyper columns for object segmentation and fine-grained localization, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, (2015), 447−456. https://doi.org/10.1109/CVPR.2015.7298642 |
[46] |
J. Weber, J. Malik, Robust computation of optical flow in a multi-scale differential framework, Int. J. Comput. Vis., 14 (1995), 67−81. https://doi.org/10.1007/BF01421489 doi: 10.1007/BF01421489
![]() |
[47] | H. L. Zheng, J. L. Fu, T. Mei, J.B. Luo, Learning multi-attention convolutional neural network for fine-grained image recognition, in Proceedings of the IEEE international conference on computer vision, (2017), 5209−5217. https://doi.org/10.1109/ICCV.2017.557 |
[48] | T. Y. Lin, A. RoyChowdhury, S. Maji, Bilinear CNN models for fine-grained visual recognition, in Proceedings of the IEEE International Conference on Computer Vision, (2015), 1449−1457. https://doi.org/10.1109/ICCV.2015.170 |
[49] | A. Fawzi, H. Samulowitz, D. Turaga, P. Frossard, Adaptive data augmentation for image classification, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, (2013), 580−587. https://doi.org/10.1109/ICIP.2016.7533048 |
[50] | R. Dellana, K. Roy, Data augmentation in CNN-based periocular authentication, in 2016 6th International Conference on Information Communication and Management, (2016), 141−145. https://doi.org/10.1109/INFOCOMAN.2016.7784231 |
[51] | J. Johnson, A. Karpathy, F. F. Li, Densecap is Fully convolutional localization networks for dense captioning, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, (2016), 4565−4574. https://doi.org/10.1109/CVPR.2016.494 |
[52] | N. S. Keskar, D. Mudigere, J. Nocedal, M. Smelyanskiy, P. T. P. Tang, On large-batch training for deep learning is generalization gap and sharp minima, (2016). https://doi.org/10.48550/arXiv.1609.04836 |
[53] | H. Li, Z. Xu, G. Taylor, T. Goldstein, Visualizing the loss landscape of neural nets, in 32nd Conference on Neural Information Processing Systems, 31 (2018). |
[54] | P. Goyal, P. Dollár, R. Girshick, P. Noordhuis, L. Wesolowski, A. Kyrola, et al., Accurate, large minibatch SGD is Training ImageNet in 1 hour, (2017). https://doi.org/10.48550/arXiv.1706.02677 |
[55] | K. Simonyan, A. Zisserman, Very deep convolutional networks for large-scale image recognition, in Proceedings of International Conference on Learning Representations, (2015). https://doi.org/10.48550/arXiv.1409.1556 |
1. | Jiajie Luo, Lina Liu, Sisi Li, Yongxin Yuan, A direct method for the simultaneous updating of finite element mass, damping and stiffness matrices, 2022, 0308-1087, 1, 10.1080/03081087.2022.2092047 |