Related
Kindly please help me with the problem as I need to use nlinfit function for fitting unknown parameters but it is showing some error. Although yesterday I was getting some values for parameters to be fitted but now if I am running it is having some issue for the function output to be used in fitted with NaN answer for last iteration only. X data is a concatenated matrix of three columns as independent variable and yk is dependent variable, taua is a matrix of initial guesses of number of parameters to be fitted.
function [yk]=activity_coefficientE(taua,x)
T=523;
alpha12=0.3; alpha13=0.3; alpha21=0.3; alpha23=0.3; alpha31=0.3; alpha32=0.3;
alpha18=0.2; alpha81=0.2; alpha28=0.2; alpha82=0.2; alpha38=0.2; alpha83=0.3;
alpha19=0.2; alpha91=0.2; alpha29=0.2; alpha92=0.2; alpha39=0.2; alpha93=0.2;
alpha110=0.2;alpha101=0.2;alpha210=0.2;alpha102=0.2;alpha310=0.2;alpha103=0.2;
alpha113=0.2;alpha131=0.2;alpha213=0.2;alpha132=0.2;alpha313=0.2;alpha133=0.2;
alpha114=0.2;alpha141=0.2;alpha214=0.2;alpha142=0.2;alpha314=0.2;alpha143=0.2;
alpha115=0.2;alpha151=0.2;alpha215=0.2;alpha152=0.2;alpha315=0.2;alpha153=0.2;
alpha117=0.2;alpha171=0.2;alpha217=0.2;alpha172=0.2;alpha317=0.2;alpha173=0.2;
alpha118=0.2;alpha181=0.2;alpha218=0.2;alpha182=0.2;alpha318=0.2;alpha183=0.2;
alpha810=0.2;alpha915=0.2;alpha1314=0.2;alpha108=0.2;alpha159=0.2;alpha1413=0.2;
alpha1718=0.2;alpha1817=0.2;
tau12=0; tau13=0; tau21=0; tau23=0; tau31=0; tau32=0;
%taua=randi([-5,5],1,112)
tau18=taua(1)+taua(57)/T;
tau81=taua(2)+taua(58)/T;
tau28=taua(3)+taua(59)/T;
tau82=taua(4)+taua(60)/T;
tau38=taua(5)+taua(61)/T;
tau83=taua(6)+taua(62)/T;
tau19=taua(7)+taua(63)/T;
tau91=taua(8)+taua(64)/T;
tau29=taua(9)+taua(65)/T;
tau92=taua(10)+taua(66)/T;
tau39=taua(11)+taua(67)/T;
tau93=taua(12)+taua(68)/T;
tau110=taua(13)+taua(69)/T;
tau101=taua(14)+taua(70)/T;
tau210=taua(15)+taua(71)/T;
tau102=taua(16)+taua(72)/T;
tau310=taua(17)+taua(73)/T;
tau103=taua(18)+taua(74)/T;
tau113=taua(19)+taua(75)/T;
tau131=taua(20)+taua(76)/T;
tau213=taua(21)+taua(77)/T;
tau132=taua(22)+taua(78)/T;
tau313=taua(23)+taua(79)/T;
tau133=taua(24)+taua(80)/T;
tau114=taua(25)+taua(81)/T;
tau141=taua(26)+taua(82)/T;
tau214=taua(27)+taua(83)/T;
tau142=taua(28)+taua(84)/T;
tau314=taua(29)+taua(85)/T;
tau143=taua(30)+taua(86)/T;
tau115=taua(31)+taua(87)/T;
tau151=taua(32)+taua(88)/T;
tau215=taua(33)+taua(89)/T;
tau152=taua(34)+taua(90)/T;
tau315=taua(35)+taua(91)/T;
tau153=taua(36)+taua(92)/T;
tau117=taua(37)+taua(93)/T;
tau171=taua(38)+taua(94)/T;
tau217=taua(39)+taua(95)/T;
tau172=taua(40)+taua(96)/T;
tau317=taua(41)+taua(97)/T;
tau173=taua(42)+taua(98)/T;
tau118=taua(43)+taua(99)/T;
tau181=taua(44)+taua(100)/T;
tau218=taua(45)+taua(101)/T;
tau182=taua(46)+taua(102)/T;
tau318=taua(47)+taua(103)/T;
tau183=taua(48)+taua(104)/T;
tau810=taua(49)+taua(105)/T;
tau108=taua(50)+taua(106)/T;
tau915=taua(51)+taua(107)/T;
tau159=taua(52)+taua(108)/T;
tau1314=taua(53)+taua(109)/T;
tau1413=taua(54)+taua(110)/T;
tau1718=taua(55)+taua(111)/T;
tau1817=taua(56)+taua(112)/T;
G12=exp(-(tau12*alpha12));
G21=exp(-(tau21*alpha21));
G13=exp(-(tau13*alpha13));
G31=exp(-(tau31*alpha31));
G23=exp(-(tau23*alpha23));
G32=exp(-(tau32*alpha32));
G18=exp(-(tau18*alpha18));
G81=exp(-(tau81*alpha81));
G28=exp(-(tau28*alpha28));
G82=exp(-(tau82*alpha82));
G38=exp(-(tau38*alpha83));
G83=exp(-(tau83*alpha83));
G19=exp(-(tau19*alpha19));
G91=exp(-(tau91*alpha91));
G29=exp(-(tau29*alpha29));
G92=exp(-(tau92*alpha92));
G39=exp(-(tau39*alpha39));
G93=exp(-(tau93*alpha93));
G110=exp(-(tau110*alpha110));
G101=exp(-(tau101*alpha101));
G210=exp(-(tau210*alpha210));
G102=exp(-(tau102*alpha102));
G310=exp(-(tau310*alpha310));
G103=exp(-(tau103*alpha103));
G113=exp(-(tau113*alpha113));
G131=exp(-(tau131*alpha131));
G213=exp(-(tau213*alpha213));
G132=exp(-(tau132*alpha132));
G313=exp(-(tau313*alpha313));
G133=exp(-(tau133*alpha133));
G114=exp(-(tau114*alpha114));
G141=exp(-(tau141*alpha141));
G214=exp(-(tau214*alpha214));
G142=exp(-(tau142*alpha142));
G314=exp(-(tau314*alpha314));
G143=exp(-(tau143*alpha143));
G115=exp(-(tau115*alpha115));
G151=exp(-(tau151*alpha151));
G215=exp(-(tau215*alpha215));
G152=exp(-(tau152*alpha152));
G315=exp(-(tau315*alpha315));
G153=exp(-(tau153*alpha153));
G117=exp(-(tau117*alpha117));
G171=exp(-(tau171*alpha171));
G217=exp(-(tau217*alpha217));
G172=exp(-(tau172*alpha172));
G317=exp(-(tau317*alpha317));
G173=exp(-(tau173*alpha173));
G118=exp(-(tau118*alpha118));
G181=exp(-(tau181*alpha181));
G218=exp(-(tau218*alpha218));
G182=exp(-(tau182*alpha182));
G318=exp(-(tau318*alpha318));
G183=exp(-(tau183*alpha183));
G810=exp(-(tau810*alpha810));
G108=exp(-(tau108*alpha108));
G915=exp(-(tau915*alpha915));
G159=exp(-(tau159*alpha159));
G1314=exp(-(tau1314*alpha1314));
G1413=exp(-(tau1413*alpha1413));
G1718=exp(-(tau1718*alpha1718));
G1817=exp(-(tau1817*alpha1817));
%calculating mole fractions of ionic species
x1=x(:,1);
x2=x(:,2);
x3=x(:,3);
%x1=[0.1577 0.1492 0.1462 0.1366 0.1299 0.1180 0.0863 0.0761 0.0550 ];
%x2=[0.8278 0.7945 0.7678 0.7450 0.6979 0.6309 0.4611 0.4114 0.2952 ];
%x3=[0.0145 0.0563 0.0860 0.1184 0.1722 0.2511 0.4526 0.5125 0.6498 ];
A=[0.0674243 0.0773881 0.0843400 0.0865343 0.0899223 0.0882858 0.0715087 0.0643867 0.0483658];
B=[0.0141081 0.0479814 0.0643151 0.0737477 0.0820756 0.0838701 0.0701576 0.0634457 0.0479639];
C=[0.0565665 0.0450072 0.0387724 0.0313828 0.02506094 0.0186280 0.0092734 0.0073438 0.0041595 ];
D=[0.0336447 0.0267694 0.0230611 0.0186659 0.0149058 0.0110795 0.0055157 0.0043679 0.0024739 ];
E=[0.0008148 0.0008756 0.00087131 0.0008794 0.0008711 0.0008441 0.0007384 0.0006997 0.0005980 ];
N=length(A);
x1n=zeros(N,1);x2n=zeros(N,1);x3n=zeros(N,1);
X1=zeros(N,1);X2=zeros(N,1);X3=zeros(N,1);X4=zeros(N,1);X5=zeros(N,1);X6=zeros(N,1);X7=zeros(N,1);
X12=zeros(N,1);X16=zeros(N,1);
for i=1:N
x1n(i)=(x1(i)-A(i)-D(i)-2*E(i)-C(i)+3*B(i))
x2n(i)=(x2(i)-A(i)-C(i)-D(i))
x3n(i)=(x3(i)-B(i))
X1(i)=(x1n(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)))
X2(i)=(x2n(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)))
X3(i)=(x3n(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)))
X4(i)=(A(i)+D(i)+E(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)))
X5(i)=(C(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)))
X6(i)=(A(i)-B(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)))
X7(i)=(B(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)))
X12(i)=(E(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)))
X16(i)=(C(i)+D(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)))
end
yc4=X4./(X4+X5);
yc5=X5./(X4+X5);
yc6=X6./(X6+X7+X12+X16);
yc7=X7./(X6+X7+X12+X16);
yc12=X12./(X6+X7+X12+X16);
yc16=X16./(X6+X7+X12+X16);
alpha14=yc6.*alpha18+yc7.*alpha19+yc12.*alpha113+yc16.*alpha117;
%alpha41=alpha14;
alpha24=yc6.*alpha28+yc7.*alpha29+yc12.*alpha213+yc16.*alpha217;
%alpha42=alpha24;
alpha34=yc6.*alpha38+yc7.*alpha39+yc12.*alpha313+yc16.*alpha317;
%alpha43=alpha34;
alpha15=yc6.*alpha110+yc7.*alpha115+yc12.*alpha114+yc16.*alpha118;
%alpha51=alpha15;
alpha25=yc6.*alpha210+yc7.*alpha215+yc12.*alpha214+yc16.*alpha218;
%alpha52=alpha25;
alpha35=yc6.*alpha310+yc7.*alpha315+yc12.*alpha314+yc16.*alpha318;
%alpha53=alpha35;
alpha16=yc4.*alpha81+yc5.*alpha101;
%alpha61=alpha16;
alpha26=yc4.*alpha82+yc5.*alpha102;
%alpha62=alpha26;
alpha36=yc4.*alpha83+yc5.*alpha103;
%alpha63=alpha36;
alpha17=yc4.*alpha91+yc5.*alpha151;
%alpha71=alpha17;
alpha27=yc4.*alpha92+yc5.*alpha152;
%alpha72=alpha27;
alpha37=yc4.*alpha93+yc5.*alpha153;
%alpha73=alpha37;
alpha112=yc4.*alpha131+yc5.*alpha141;
%alpha121=alpha112;
alpha212=yc4.*alpha132+yc5.*alpha142;
%alpha122=alpha212;
alpha312=yc4.*alpha133+yc5.*alpha143;
%alpha123=alpha312;
alpha116=yc4.*alpha171+yc5.*alpha181;
%alpha161=alpha116;
alpha216=yc4.*alpha172+yc5.*alpha182;
%alpha162=alpha216;
alpha316=yc4.*alpha173+yc5.*alpha183;
%alpha163=alpha316;
alpha46=yc5.*alpha810;
%alpha64=alpha46;
alpha47=yc5.*alpha915;
%alpha74=alpha47;
alpha412=yc5.*alpha1314;
%alpha124=alpha412;
alpha416=yc5.*alpha1718;
%alpha164=alpha416;
alpha56=yc4.*alpha108;
%alpha65=alpha56;
alpha57=yc4.*alpha159;
%alpha75=alpha57;
alpha512=yc4.*alpha1413;
%alpha125=alpha512;
alpha516=yc4.*alpha1817;
%alpha165=alpha516;
G14=yc6.*G18+yc7.*G19+yc12.*G113+yc16.*G117;
%G41=G14;
G24=yc6.*G28+yc7.*G29+yc12.*G213+yc16.*G217;
%G42=G24;
G34=yc6.*G38+yc7.*G39+yc12.*G313+yc16.*G317;
%G43=G34;
G15=yc6.*G110+yc7.*G115+yc12.*G114+yc16.*G118;
%G51=G15;
G25=yc6.*G210+yc7.*G215+yc12.*G214+yc16.*G218;
%G52=G25;
G35=yc6.*G310+yc7.*G315+yc12.*G314+yc16.*G318;
%G53=G35;
G16=yc4.*G81+yc5.*G101;
%G61=G16;
G26=yc4.*G82+yc5.*G102;
%G62=G26;
G36=yc4.*G83+yc5.*G103;
%G63=G36;
G17=yc4.*G91+yc5.*G151;
%G71=G17;
G27=yc4.*G92+yc5.*G152;
%G72=G27;
G37=yc4.*G93+yc5.*G153;
%G73=G37;
G112=yc4.*G131+yc5.*G141;
%G121=G112;
G212=yc4.*G132+yc5.*G142;
%G122=G212;
G312=yc4.*G133+yc5.*G143;
%G123=G312;
G116=yc4.*G171+yc5.*G181;
%G161=G116;
G216=yc4.*G172+yc5.*G182;
%G162=G216;
G316=yc4.*G173+yc5.*G183;
%G163=G316;
G46=yc5.*G810;
%G64=G46;
G47=yc5.*G915;
%G74=G47;
G412=yc5.*G1314;
%G124=G412;
G416=yc5.*G1718;
%G164=G416;
G56=yc4.*G108;
%G65=G56;
G57=yc4.*G159;
%G75=G57;
G512=yc4.*G1413;
%G125=G512;
G516=yc4.*G1817;
%G165=G516;
tau14=-log(G14)./alpha14;
%tau41=tau14;
tau24=-log(G24)./alpha24;
%tau42=tau24;
tau34=-log(G34)./alpha34;
%tau43=tau34;
tau15=-log(G15)./alpha15;
%tau51=tau15;
tau25=-log(G25)./alpha25;
%tau52=tau25;
tau35=-log(G35)./alpha35;
%tau53=tau35;
tau16=-log(G16)./alpha16;
%tau61=tau16;
tau26=-log(G26)./alpha26;
%tau62=tau26;
tau36=-log(G36)./alpha36;
%tau63=tau36;
tau17=-log(G17)./alpha17;
%tau71=tau17;
tau27=-log(G27)./alpha27;
%tau72=tau27;
tau37=-log(G37)./alpha37;
%tau73=tau37;
tau112=-log(G112)./alpha112;
%tau121=tau112;
tau212=-log(G212)./alpha212;
%tau122=tau212;
tau312=-log(G312)./alpha312;
%tau123=tau312;
tau116=-log(G116)./alpha116;
%tau161=tau116;
tau216=-log(G216)./alpha216;
%tau162=tau216;
tau316=-log(G316)./alpha316;
%tau163=tau316;
tau46=-log(G46)./alpha46;
%tau64=tau46;
tau47=-log(G47)./alpha47;
%tau74=tau47;
tau412=-log(G412)./alpha412;
%tau124=tau412;
tau416=-log(G416)./alpha416;
%tau164=tau416;
tau56=-log(G56)./alpha56;
%tau65=tau56;
tau57=-log(G57)./alpha57;
%tau75=tau57;
tau512=-log(G512)./alpha512;
%tau125=tau512;
tau516=-log(G516)./alpha516;
%tau165=tau516;
ln_y1_1=G12.*X2.*tau12+ G31.*X3.*tau13+ G14.*X4.*tau14+G15.*X5.*tau15+G16.*X6.*tau16+G17.*X7.*tau17+G112.*X12.*tau112+G116.*X16.*tau116;
ln_y1_2=G12.*X2+ G13.*X3+ G14.*X4+G15.*X5+G16.*X6+G17.*X7+G112.*X12+G116.*X16;
ln_y2_1=G21.*X1.*tau12+ G32.*X3.*tau32+ G24.*X4.*tau24+G25.*X5.*tau25+G26.*X6.*tau26+G27.*X7.*tau27+G212.*X12.*tau212+G216.*X16.*tau216;
ln_y2_2=G12.*X1+ G23.*X3+G24.*X4+G25.*X5+G26.*X6+G27.*X7+G212.*X12+G216.*X16;
ln_y3_1=G13.*X1.*tau13+ G23.*X3.*tau23+ G34.*X4.*tau34+G35.*X5.*tau35+G36.*X6.*tau36+G37.*X7.*tau37+G312.*X12.*tau312+G316.*X16.*tau316;
ln_y3_2=G13.*X1+ G23.*X3+ G34.*X4+G35.*X5+G36.*X6+G37.*X7+G312.*X12+G316.*X16;
ln_y4_1=G14.*X1.*tau14+G24.*X2.*tau24+G34.*X3.*tau34+G46.*X6.*tau46+G47.*X7.*tau47+G412.*X12.*tau412+G416.*X16.*tau416;
ln_y4_2=G14.*X1+G24.*X2+G34.*X3+G46.*X6+G47.*X7+G412.*X12+G416.*X16;
ln_y5_1=G15.*X1.*tau15+G25.*X2.*tau25+G35.*X3.*tau35+G56.*X6.*tau56+G57.*X7.*tau57+G512.*X12.*tau512+G516.*X16.*tau516;
ln_y5_2=G15.*X1+G25.*X2+G35.*X3+G56.*X6+G57.*X7+G512.*X12+G516.*X16;
ln_y6_1=G16.*X1.*tau16+G26.*X2.*tau26+G36.*X3.*tau36+G46.*X4.*tau46+G56.*X5.*tau56;
ln_y6_2=G16.*X1+G26.*X2+G36.*X3+G46.*X4+G56.*X5;
ln_y7_1=G17.*X1.*tau17+G27.*X2.*tau27+G37.*X3.*tau37+G47.*X4.*tau47+G57.*X5.*tau57;
ln_y7_2=G17.*X1+G27.*X2+G37.*X3+G47.*X4+G57.*X5;
ln_y12_1=G112.*X1.*tau112+G212.*X2.*tau212+G312.*X3.*tau312+G412.*X4.*tau412+G512.*X5.*tau512;
ln_y12_2=G112.*X1+G212.*X2+G312.*X3+G412.*X4+G512.*X5;
ln_y16_1=G116.*X1.*tau116+G216.*X2.*tau216+G316.*X3.*tau316+G416.*X4.*tau416+G516.*X5.*tau516;
ln_y16_2=G116.*X1+G216.*X2+G316.*X3+G416.*X4+G516.*X5;
ln_y1_3=(((X2.*G12)./ln_y2_2).*(tau12-(ln_y2_1)./(ln_y2_2)))+(((X3.*G13)./ln_y3_2).*(tau13-(ln_y3_1)./(ln_y3_2)));
ln_y1_4=(((X6.*G16)./ln_y6_2).*(tau16- (ln_y6_1./ln_y6_2))) + (((X7.*G17)./ln_y7_2).*(tau17- (ln_y7_1./ln_y7_2)))+(((X12.*G12)./ln_y12_2).*(tau112- (ln_y12_1./ln_y12_2)))+(((X16.*G16)./ln_y16_2).*(tau116- (ln_y16_1./ln_y16_2)));
ln_y1_5=(((X4.*G14)./ln_y4_2).*(tau14- (ln_y4_1./ln_y4_2))) + (((X5.*G15)./ln_y5_2).*(tau15- (ln_y5_1./ln_y5_2)));
yk=exp((ln_y1_1./ln_y1_2) + ln_y1_3 + ln_y1_4+ ln_y1_5) % activity coefficient for H2O
end
........................................
Another function where above function to be called.....
% calling the function act_coeff to estimate the binary interaction parameters
for i=1:112
filename = 'EagelsDATA.xlsx'; %reading VLE data from excel file
Data = xlsread(filename);
x(:,1) = Data([10:15 17:19],16);
x(:,2) = Data([10:15 17:19],1);
x(:,3)= Data([10:15 17:19],2);
taua=(randi([-5,5],1,112));
yk=[0.0606 (values calculated from above function and will be used for fitting)
0.4327
0.6545
0.9417
1.2570
1.6881
1.9108
1.7777
1.3821]
% taua =[ -2 3 4 -3 -4 1 4 -2 4 -4 -1 4 5 -3 3 2 -5 3 -4
% 1 4 1 5 -1 -1 -3 2 -3 4 3 4 2 5 4 -2 4 3 -1
% 1 0 -5 -5 -5 -3 4 2 1 4 0 2 -3 -4 5 0 -3 2 5
% 1 0 5 1 -3 5 4 1 5 2 3 2 0 -5 -4 -2 1 -2 5
%-5 5 -2 -2 4 1 -1 3 -1 1 5 -1 0 -1 4 5 5 1 4
% 1 0 4 -4 4 0 -1 -2 -5 -3 -4 -5
% -5 0 -2 0 -5] (random values for which yk was calculted from the command
taua= randi([-5,5],1,112))
try % try-catch used to continue the loop without stopping on encountering an error
[taua1]= nlinfit(x,yk,#activity_coefficientE,taua)
catch exception
continue
end
end
I am not able to attach excel sheet here so data from excel sheet is as:
x =[0.1577 0.1492 0.1462 0.1366 0.1299 0.1180 0.0863 0.0761 0.0550; column 1
0.8278 0.7945 0.7678 0.7450 0.6979 0.6309 0.4611 0.4114 0.2952 ; column 2
0.0145 0.0563 0.0860 0.1184 0.1722 0.2511 0.4526 0.5125 0.6498 ]; column 3
I found 3 major problems with what you did.
Problem #1 - errors
The reason you get the error is because your function "activity_coefficientE" can sometimes return NaN or inf values. My suggestion is to look for these values and set the value of "yk" to a large value so that the optimizer in "nlinfit" will stay away from coefficients that produce infinite or NaN values. This is the code at the bottom of the function so that you avoid crashes:
if any(~isfinite(yk))
yk = 10 * ones(size(yk));
end
Problem #2 - random initial guesses
The trouble with using random numbers for your initial conditions is that every time you run it you get a different answer, so sometimes it works and sometimes it doesn't. If you set the random number generator seed, you can get the same random numbers each time you run the script. If you change you seed, you can get a different set of random numbers. I shortened your main script to this, where I try 100 different random seeds (and store the results of each attempt) to see what answers result:
for i=1:100
rng(i)
taua = randi([-5,5],1,112);
taua1(i, :) = nlinfit(x,yk,#activity_coefficientE,taua);
end
Each row of "taua1" is a set of 111 coefficients.
Problem #3 - Trying to fit 9 points with 112 coefficients
Every time nlinfit is called, you get this warning:
Warning: Rank deficient
because you have more coefficients (112) that you are asking nlinfit to find than data points you are fitting (9). It's like trying to find the 2nd order equation that best fits 2 points, there are an infinite number of solutions. When curve fitting you should have more data points than coefficients to make sure you're not fitting noise. You need more data points in "yk" and "x" and/or fewer coefficients to fit. I've done a lot of curve fitting and I've never seen an equation with 112 coefficients, so I am thinking that you are not solving the problem correctly. Perhaps the 112 coefficients aren't really independent or there are 112 data points and 9 coefficients that you want to find.
For completeness, here is my edited version of the activity_coefficientE.m function that I created to work on this solution. In general, I never see Matlab code with this many variables with similar names. Much of this code could be greatly simplified by using vector operations. Most of my changes involve formatting, adding semicolons, and the checks for non-finite values at the end.
function yk=activity_coefficientE(taua,x)
T=523;
alpha12=0.3; alpha13=0.3; alpha21=0.3; alpha23=0.3; alpha31=0.3; alpha32=0.3;
alpha18=0.2; alpha81=0.2; alpha28=0.2; alpha82=0.2; alpha38=0.2; alpha83=0.3;
alpha19=0.2; alpha91=0.2; alpha29=0.2; alpha92=0.2; alpha39=0.2; alpha93=0.2;
alpha110=0.2;alpha101=0.2;alpha210=0.2;alpha102=0.2;alpha310=0.2;alpha103=0.2;
alpha113=0.2;alpha131=0.2;alpha213=0.2;alpha132=0.2;alpha313=0.2;alpha133=0.2;
alpha114=0.2;alpha141=0.2;alpha214=0.2;alpha142=0.2;alpha314=0.2;alpha143=0.2;
alpha115=0.2;alpha151=0.2;alpha215=0.2;alpha152=0.2;alpha315=0.2;alpha153=0.2;
alpha117=0.2;alpha171=0.2;alpha217=0.2;alpha172=0.2;alpha317=0.2;alpha173=0.2;
alpha118=0.2;alpha181=0.2;alpha218=0.2;alpha182=0.2;alpha318=0.2;alpha183=0.2;
alpha810=0.2;alpha915=0.2;alpha1314=0.2;alpha108=0.2;alpha159=0.2;alpha1413=0.2;
alpha1718=0.2;alpha1817=0.2;
tau12=0; tau13=0; tau21=0; tau23=0; tau31=0; tau32=0;
tau18=taua(1)+taua(57)/T;
tau81=taua(2)+taua(58)/T;
tau28=taua(3)+taua(59)/T;
tau82=taua(4)+taua(60)/T;
tau38=taua(5)+taua(61)/T;
tau83=taua(6)+taua(62)/T;
tau19=taua(7)+taua(63)/T;
tau91=taua(8)+taua(64)/T;
tau29=taua(9)+taua(65)/T;
tau92=taua(10)+taua(66)/T;
tau39=taua(11)+taua(67)/T;
tau93=taua(12)+taua(68)/T;
tau110=taua(13)+taua(69)/T;
tau101=taua(14)+taua(70)/T;
tau210=taua(15)+taua(71)/T;
tau102=taua(16)+taua(72)/T;
tau310=taua(17)+taua(73)/T;
tau103=taua(18)+taua(74)/T;
tau113=taua(19)+taua(75)/T;
tau131=taua(20)+taua(76)/T;
tau213=taua(21)+taua(77)/T;
tau132=taua(22)+taua(78)/T;
tau313=taua(23)+taua(79)/T;
tau133=taua(24)+taua(80)/T;
tau114=taua(25)+taua(81)/T;
tau141=taua(26)+taua(82)/T;
tau214=taua(27)+taua(83)/T;
tau142=taua(28)+taua(84)/T;
tau314=taua(29)+taua(85)/T;
tau143=taua(30)+taua(86)/T;
tau115=taua(31)+taua(87)/T;
tau151=taua(32)+taua(88)/T;
tau215=taua(33)+taua(89)/T;
tau152=taua(34)+taua(90)/T;
tau315=taua(35)+taua(91)/T;
tau153=taua(36)+taua(92)/T;
tau117=taua(37)+taua(93)/T;
tau171=taua(38)+taua(94)/T;
tau217=taua(39)+taua(95)/T;
tau172=taua(40)+taua(96)/T;
tau317=taua(41)+taua(97)/T;
tau173=taua(42)+taua(98)/T;
tau118=taua(43)+taua(99)/T;
tau181=taua(44)+taua(100)/T;
tau218=taua(45)+taua(101)/T;
tau182=taua(46)+taua(102)/T;
tau318=taua(47)+taua(103)/T;
tau183=taua(48)+taua(104)/T;
tau810=taua(49)+taua(105)/T;
tau108=taua(50)+taua(106)/T;
tau915=taua(51)+taua(107)/T;
tau159=taua(52)+taua(108)/T;
tau1314=taua(53)+taua(109)/T;
tau1413=taua(54)+taua(110)/T;
tau1718=taua(55)+taua(111)/T;
tau1817=taua(56)+taua(112)/T;
G12=exp(-(tau12*alpha12));
G21=exp(-(tau21*alpha21));
G13=exp(-(tau13*alpha13));
G31=exp(-(tau31*alpha31));
G23=exp(-(tau23*alpha23));
G32=exp(-(tau32*alpha32));
G18=exp(-(tau18*alpha18));
G81=exp(-(tau81*alpha81));
G28=exp(-(tau28*alpha28));
G82=exp(-(tau82*alpha82));
G38=exp(-(tau38*alpha83));
G83=exp(-(tau83*alpha83));
G19=exp(-(tau19*alpha19));
G91=exp(-(tau91*alpha91));
G29=exp(-(tau29*alpha29));
G92=exp(-(tau92*alpha92));
G39=exp(-(tau39*alpha39));
G93=exp(-(tau93*alpha93));
G110=exp(-(tau110*alpha110));
G101=exp(-(tau101*alpha101));
G210=exp(-(tau210*alpha210));
G102=exp(-(tau102*alpha102));
G310=exp(-(tau310*alpha310));
G103=exp(-(tau103*alpha103));
G113=exp(-(tau113*alpha113));
G131=exp(-(tau131*alpha131));
G213=exp(-(tau213*alpha213));
G132=exp(-(tau132*alpha132));
G313=exp(-(tau313*alpha313));
G133=exp(-(tau133*alpha133));
G114=exp(-(tau114*alpha114));
G141=exp(-(tau141*alpha141));
G214=exp(-(tau214*alpha214));
G142=exp(-(tau142*alpha142));
G314=exp(-(tau314*alpha314));
G143=exp(-(tau143*alpha143));
G115=exp(-(tau115*alpha115));
G151=exp(-(tau151*alpha151));
G215=exp(-(tau215*alpha215));
G152=exp(-(tau152*alpha152));
G315=exp(-(tau315*alpha315));
G153=exp(-(tau153*alpha153));
G117=exp(-(tau117*alpha117));
G171=exp(-(tau171*alpha171));
G217=exp(-(tau217*alpha217));
G172=exp(-(tau172*alpha172));
G317=exp(-(tau317*alpha317));
G173=exp(-(tau173*alpha173));
G118=exp(-(tau118*alpha118));
G181=exp(-(tau181*alpha181));
G218=exp(-(tau218*alpha218));
G182=exp(-(tau182*alpha182));
G318=exp(-(tau318*alpha318));
G183=exp(-(tau183*alpha183));
G810=exp(-(tau810*alpha810));
G108=exp(-(tau108*alpha108));
G915=exp(-(tau915*alpha915));
G159=exp(-(tau159*alpha159));
G1314=exp(-(tau1314*alpha1314));
G1413=exp(-(tau1413*alpha1413));
G1718=exp(-(tau1718*alpha1718));
G1817=exp(-(tau1817*alpha1817));
%calculating mole fractions of ionic species
x1=x(:,1);
x2=x(:,2);
x3=x(:,3);
A=[0.0674243 0.0773881 0.0843400 0.0865343 0.0899223 0.0882858 0.0715087 0.0643867 0.0483658];
B=[0.0141081 0.0479814 0.0643151 0.0737477 0.0820756 0.0838701 0.0701576 0.0634457 0.0479639];
C=[0.0565665 0.0450072 0.0387724 0.0313828 0.02506094 0.0186280 0.0092734 0.0073438 0.0041595 ];
D=[0.0336447 0.0267694 0.0230611 0.0186659 0.0149058 0.0110795 0.0055157 0.0043679 0.0024739 ];
E=[0.0008148 0.0008756 0.00087131 0.0008794 0.0008711 0.0008441 0.0007384 0.0006997 0.0005980 ];
N=length(A);
x1n=zeros(N,1);x2n=zeros(N,1);x3n=zeros(N,1);
X1=zeros(N,1);X2=zeros(N,1);X3=zeros(N,1);X4=zeros(N,1);X5=zeros(N,1);X6=zeros(N,1);X7=zeros(N,1);
X12=zeros(N,1);X16=zeros(N,1);
for i=1:N
x1n(i)=(x1(i)-A(i)-D(i)-2*E(i)-C(i)+3*B(i));
x2n(i)=(x2(i)-A(i)-C(i)-D(i));
x3n(i)=(x3(i)-B(i));
X1(i)=(x1n(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)));
X2(i)=(x2n(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)));
X3(i)=(x3n(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)));
X4(i)=(A(i)+D(i)+E(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)));
X5(i)=(C(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)));
X6(i)=(A(i)-B(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)));
X7(i)=(B(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)));
X12(i)=(E(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)));
X16(i)=(C(i)+D(i)/(x1n(i)+x2n(i)+x3n(i)+2*A(i)+4*B(i)+2*C(i)+2*D(i)+2*E(i)));
end
yc4=X4./(X4+X5);
yc5=X5./(X4+X5);
yc6=X6./(X6+X7+X12+X16);
yc7=X7./(X6+X7+X12+X16);
yc12=X12./(X6+X7+X12+X16);
yc16=X16./(X6+X7+X12+X16);
alpha14=yc6.*alpha18+yc7.*alpha19+yc12.*alpha113+yc16.*alpha117;
alpha24=yc6.*alpha28+yc7.*alpha29+yc12.*alpha213+yc16.*alpha217;
alpha34=yc6.*alpha38+yc7.*alpha39+yc12.*alpha313+yc16.*alpha317;
alpha15=yc6.*alpha110+yc7.*alpha115+yc12.*alpha114+yc16.*alpha118;
alpha25=yc6.*alpha210+yc7.*alpha215+yc12.*alpha214+yc16.*alpha218;
alpha35=yc6.*alpha310+yc7.*alpha315+yc12.*alpha314+yc16.*alpha318;
alpha16=yc4.*alpha81+yc5.*alpha101;
alpha26=yc4.*alpha82+yc5.*alpha102;
alpha36=yc4.*alpha83+yc5.*alpha103;
alpha17=yc4.*alpha91+yc5.*alpha151;
alpha27=yc4.*alpha92+yc5.*alpha152;
alpha37=yc4.*alpha93+yc5.*alpha153;
alpha112=yc4.*alpha131+yc5.*alpha141;
alpha212=yc4.*alpha132+yc5.*alpha142;
alpha312=yc4.*alpha133+yc5.*alpha143;
alpha116=yc4.*alpha171+yc5.*alpha181;
alpha216=yc4.*alpha172+yc5.*alpha182;
alpha316=yc4.*alpha173+yc5.*alpha183;
alpha46=yc5.*alpha810;
alpha47=yc5.*alpha915;
alpha412=yc5.*alpha1314;
alpha416=yc5.*alpha1718;
alpha56=yc4.*alpha108;
alpha57=yc4.*alpha159;
alpha512=yc4.*alpha1413;
alpha516=yc4.*alpha1817;
G14=yc6.*G18+yc7.*G19+yc12.*G113+yc16.*G117;
G24=yc6.*G28+yc7.*G29+yc12.*G213+yc16.*G217;
G34=yc6.*G38+yc7.*G39+yc12.*G313+yc16.*G317;
G15=yc6.*G110+yc7.*G115+yc12.*G114+yc16.*G118;
G25=yc6.*G210+yc7.*G215+yc12.*G214+yc16.*G218;
G35=yc6.*G310+yc7.*G315+yc12.*G314+yc16.*G318;
G16=yc4.*G81+yc5.*G101;
G26=yc4.*G82+yc5.*G102;
G36=yc4.*G83+yc5.*G103;
G17=yc4.*G91+yc5.*G151;
G27=yc4.*G92+yc5.*G152;
G37=yc4.*G93+yc5.*G153;
G112=yc4.*G131+yc5.*G141;
G212=yc4.*G132+yc5.*G142;
G312=yc4.*G133+yc5.*G143;
G116=yc4.*G171+yc5.*G181;
G216=yc4.*G172+yc5.*G182;
G316=yc4.*G173+yc5.*G183;
G46=yc5.*G810;
G47=yc5.*G915;
G412=yc5.*G1314;
G416=yc5.*G1718;
G56=yc4.*G108;
G57=yc4.*G159;
G512=yc4.*G1413;
G516=yc4.*G1817;
tau14=-log(G14)./alpha14;
tau24=-log(G24)./alpha24;
tau34=-log(G34)./alpha34;
tau15=-log(G15)./alpha15;
tau25=-log(G25)./alpha25;
tau35=-log(G35)./alpha35;
tau16=-log(G16)./alpha16;
tau26=-log(G26)./alpha26;
tau36=-log(G36)./alpha36;
tau17=-log(G17)./alpha17;
tau27=-log(G27)./alpha27;
tau37=-log(G37)./alpha37;
tau112=-log(G112)./alpha112;
tau212=-log(G212)./alpha212;
tau312=-log(G312)./alpha312;
tau116=-log(G116)./alpha116;
tau216=-log(G216)./alpha216;
tau316=-log(G316)./alpha316;
tau46=-log(G46)./alpha46;
tau47=-log(G47)./alpha47;
tau412=-log(G412)./alpha412;
tau416=-log(G416)./alpha416;
tau56=-log(G56)./alpha56;
tau57=-log(G57)./alpha57;
tau512=-log(G512)./alpha512;
tau516=-log(G516)./alpha516;
ln_y1_1=G12.*X2.*tau12+ G31.*X3.*tau13+ G14.*X4.*tau14+G15.*X5.*tau15+G16.*X6.*tau16+G17.*X7.*tau17+G112.*X12.*tau112+G116.*X16.*tau116;
ln_y1_2=G12.*X2+ G13.*X3+ G14.*X4+G15.*X5+G16.*X6+G17.*X7+G112.*X12+G116.*X16;
ln_y2_1=G21.*X1.*tau12+ G32.*X3.*tau32+ G24.*X4.*tau24+G25.*X5.*tau25+G26.*X6.*tau26+G27.*X7.*tau27+G212.*X12.*tau212+G216.*X16.*tau216;
ln_y2_2=G12.*X1+ G23.*X3+G24.*X4+G25.*X5+G26.*X6+G27.*X7+G212.*X12+G216.*X16;
ln_y3_1=G13.*X1.*tau13+ G23.*X3.*tau23+ G34.*X4.*tau34+G35.*X5.*tau35+G36.*X6.*tau36+G37.*X7.*tau37+G312.*X12.*tau312+G316.*X16.*tau316;
ln_y3_2=G13.*X1+ G23.*X3+ G34.*X4+G35.*X5+G36.*X6+G37.*X7+G312.*X12+G316.*X16;
ln_y4_1=G14.*X1.*tau14+G24.*X2.*tau24+G34.*X3.*tau34+G46.*X6.*tau46+G47.*X7.*tau47+G412.*X12.*tau412+G416.*X16.*tau416;
ln_y4_2=G14.*X1+G24.*X2+G34.*X3+G46.*X6+G47.*X7+G412.*X12+G416.*X16;
ln_y5_1=G15.*X1.*tau15+G25.*X2.*tau25+G35.*X3.*tau35+G56.*X6.*tau56+G57.*X7.*tau57+G512.*X12.*tau512+G516.*X16.*tau516;
ln_y5_2=G15.*X1+G25.*X2+G35.*X3+G56.*X6+G57.*X7+G512.*X12+G516.*X16;
ln_y6_1=G16.*X1.*tau16+G26.*X2.*tau26+G36.*X3.*tau36+G46.*X4.*tau46+G56.*X5.*tau56;
ln_y6_2=G16.*X1+G26.*X2+G36.*X3+G46.*X4+G56.*X5;
ln_y7_1=G17.*X1.*tau17+G27.*X2.*tau27+G37.*X3.*tau37+G47.*X4.*tau47+G57.*X5.*tau57;
ln_y7_2=G17.*X1+G27.*X2+G37.*X3+G47.*X4+G57.*X5;
ln_y12_1=G112.*X1.*tau112+G212.*X2.*tau212+G312.*X3.*tau312+G412.*X4.*tau412+G512.*X5.*tau512;
ln_y12_2=G112.*X1+G212.*X2+G312.*X3+G412.*X4+G512.*X5;
ln_y16_1=G116.*X1.*tau116+G216.*X2.*tau216+G316.*X3.*tau316+G416.*X4.*tau416+G516.*X5.*tau516;
ln_y16_2=G116.*X1+G216.*X2+G316.*X3+G416.*X4+G516.*X5;
ln_y1_3=(((X2.*G12)./ln_y2_2).*(tau12-(ln_y2_1)./(ln_y2_2)))+(((X3.*G13)./ln_y3_2).*(tau13-(ln_y3_1)./(ln_y3_2)));
ln_y1_4=(((X6.*G16)./ln_y6_2).*(tau16- (ln_y6_1./ln_y6_2))) + (((X7.*G17)./ln_y7_2).*(tau17- (ln_y7_1./ln_y7_2)))+(((X12.*G12)./ln_y12_2).*(tau112- (ln_y12_1./ln_y12_2)))+(((X16.*G16)./ln_y16_2).*(tau116- (ln_y16_1./ln_y16_2)));
ln_y1_5=(((X4.*G14)./ln_y4_2).*(tau14- (ln_y4_1./ln_y4_2))) + (((X5.*G15)./ln_y5_2).*(tau15- (ln_y5_1./ln_y5_2)));
yk=exp((ln_y1_1./ln_y1_2) + ln_y1_3 + ln_y1_4+ ln_y1_5)'; % activity coefficient for H2O
if any(~isfinite(yk))
yk = 10 * ones(size(yk));
end
i'm learning pytorch, and taking the anpr project,which is based tensorflow
(https://github.com/matthewearl/deep-anpr,
http://matthewearl.github.io/2016/05/06/cnn-anpr/)
as a exercise, transplant it to pytorch platform.
there is a problem,i'm using nn.CrossEntropyLoss() as loss function:
criterion=nn.CrossEntropyLoss()
the output.data of model is:
- 1.00000e-02 *
- 2.5552 2.7582 2.5368 ... 5.6184 1.2288 -0.0076
- 0.7033 1.3167 -1.0966 ... 4.7249 1.3217 1.8367
- 0.7592 1.4777 1.8095 ... 0.8733 1.2417 1.1521
- 0.1040 -0.7054 -3.4862 ... 4.7703 2.9595 1.4263
- [torch.FloatTensor of size 4x253]
and targets.data is:
- 1 0 0 ... 0 0 0
- 1 0 0 ... 0 0 0
- 1 0 0 ... 0 0 0
- 1 0 0 ... 0 0 0
- [torch.DoubleTensor of size 4x253]
when i call:
loss=criterion(output,targets)
error occured,information is:
TypeError: FloatClassNLLCriterion_updateOutput received an invalid combination of arguments - got (int, torch.FloatTensor, **torch.DoubleTensor**, torch.FloatTensor, bool, NoneType, torch.FloatTensor), but expected (int state, torch.FloatTensor input, **torch.LongTensor** target, torch.FloatTensor output, bool sizeAverage, [torch.FloatTensor weights or None], torch.FloatTensor total_weight)
'expected torch.LongTensor'......'got torch.DoubleTensor',but if i convert the targets into LongTensor:
torch.LongTensor(numpy.array(targets.data.numpy(),numpy.long))
call loss=criterion(output,targets), the error is:
RuntimeError: multi-target not supported at /data/users/soumith/miniconda2/conda-bld/pytorch-0.1.10_1488752595704/work/torch/lib/THNN/generic/ClassNLLCriterion.c:20
my last exercise is mnist, a example from pytorch,i made a bit modification,batch_size is 4,the loss function:
loss = F.nll_loss(outputs, labels)
outputs.data:
- -2.3220 -2.1229 -2.3395 -2.3391 -2.5270 -2.3269 -2.1055 -2.2321 -2.4943 -2.2996
-2.3653 -2.2034 -2.4437 -2.2708 -2.5114 -2.3286 -2.1921 -2.1771 -2.3343 -2.2533
-2.2809 -2.2119 -2.3872 -2.2190 -2.4610 -2.2946 -2.2053 -2.3192 -2.3674 -2.3100
-2.3715 -2.1455 -2.4199 -2.4177 -2.4565 -2.2812 -2.2467 -2.1144 -2.3321 -2.3009
[torch.FloatTensor of size 4x10]
labels.data:
- 8
- 6
- 0
- 1
- [torch.LongTensor of size 4]
the labels, for a input image,must be a single element, in upper example, there is 253 numbers, and in 'mnist',there is only one number, the shape of outputs is difference from labels.
i review the tensorflow manual, tf.nn.softmax_cross_entropy_with_logits,
'Logits and labels must have the sameshape [batch_size, num_classes] and the same dtype (either float32 or float64).'
does pytorch support the same function in tensorflow?
many thks
You can convert the targets that you have to a categorical representation.
In the example that you provide, you would have 1 0 0 0.. 0 if the class is 0, 0 1 0 0 ... if the class is 1, 0 0 1 0 0 0... if the class is 2 etc.
One quick way that I can think of is first convert the target Tensor to a numpy array, then convert it from one hot to a categorical array, and convert it back to a pytorch Tensor. Something like this:
targetnp=targets.numpy()
idxs=np.where(targetnp>0)[1]
new_targets=torch.LongTensor(idxs)
loss=criterion(output,new_targets)
CrossEntropyLoss is equivalent to tf.nn.softmax_cross_entropy_with_logits. The input to CrossEntropyLoss is a categorical vector of shape [batch_size]. Use .view() to change the tensor shapes.
labels = labels.view(-1)
output = output.view(labels.size(0), -1)
loss = criterion(output, loss)
calling .view(x, y, -1) causes the tensor to use the remaining datapoints to fill the -1 dimension and will cause an error if there is not enough to make a full dimension
labels.size(0) gives the size of the 0th dimension of the label tensor
Additional
to convert between tensor types you can call the type on the tensor, for example 'labels = labels.long()`
Second Additional
If you unpack the data from a variable like so output.data then you will lose the gradients for that output and be unable to backprop when the time comes
I was wondering if the procedure applied trying to download the sample rate was the appropriate as follows the instruction: y = downsample(x,n)
downsamp_rate = 40;
downsampled_data = downsample(X,downsamp_rate);
.. because my doubt relays in why the first column from both matrices is exactly the same (the original matrix and the sample donwloaded)maintaining the same data....
then the other data have already transformed to a lower sample rate.
Thank you so much!
Best!
edited: Sample data. I pasted the data but I can upload de .mat files.
Original data.
column 1 column 2 column 3
-0,593600000000000 -0,592699999999996 -0,591899999999995
2,42180000000000 2,41010000000000 2,40360000000000
1,78550000000000 1,79020000000000 1,79530000000000
-1,30590000000000 -1,31520000000000 -1,31530000000000
-0,707800000000003 -0,712699999999999 -0,727700000000003
-0,986500000000001 -0,996000000000002 -1,00460000000000
-0,989699999999999 -0,989699999999999 -0,989699999999999
1,23500000000000 1,22970000000000 1,21880000000000
0,122899999999998 0,127899999999997 0,128899999999998
0,938300000000003 0,937500000000002 0,936200000000004
0,248600000000004 0,248500000000002 0,248700000000002
-0,381499999999996 -0,393199999999999 -0,393699999999997
0,294099999999997 0,279299999999999 0,271299999999997
-0,223200000000001 -0,223699999999999 -0,227299999999997
0,0879999999999992 0,117300000000004 0,122500000000003
-0,167899999999999 -0,170999999999999 -0,174800000000003
-0,687499999999996 -0,697199999999998 -0,701600000000002
-0,681700000000002 -0,682200000000000 -0,683000000000000
1,19659999999999 1,19670000000000 1,19490000000000
-0,565500000000008 -0,565199999999999 -0,557400000000008
Downsampled data
column 1 column 2 column 3
-0,593600000000000 0,821900000000003 0,936300000000001
2,42180000000000 1,14610000000000 -0,255400000000000
1,78550000000000 2,86550000000000 3,66890000000000
-1,30590000000000 7,01950000000000 12,9564000000000
-0,707800000000003 3,05920000000000 0,852999999999998
-0,986500000000001 -0,372200000000000 -0,951000000000002
-0,989699999999999 -0,988000000000000 -1,21730000000000
1,23500000000000 5,79700000000000 3,40880000000000
0,122899999999998 5,32230000000000 5,19260000000000
0,938300000000003 4,88130000000000 7,55900000000000
0,248600000000004 4,79290000000000 2,96620000000000
-0,381499999999996 -0,400000000000000 0,641500000000000
0,294099999999997 -0,131400000000004 -1,20040000000000
-0,223200000000001 1,49610000000000 1,59030000000000
0,0879999999999992 0,418700000000000 -0,0114999999999976
-0,167899999999999 0,0149999999999983 -0,857500000000000
-0,687499999999996 -0,593100000000002 0,119700000000000
-0,681700000000002 -0,170000000000003 0,126799999999999
1,19659999999999 1,17670000000000 1,15780000000000
-0,565500000000008 8,89019999999999 6,58569999999999
A possible for your output is a periodic input signal with a period length of downsamp_rate-1. To give a short demonstration:
>> X=repmat(1:39,1,10);
>> downsampled_data = downsample(X,downsamp_rate);
>> downsampled_data
downsampled_data =
Columns 1 through 9
1 2 3 4 5 6 7 8 9
Column 10
10
Thus, take a look at your rows 40,41,42. I assume the first value is identical to your row 1,2,3
I am trying to use sequentialfs on a logistic regression in order to determine the variables to include. I have tried to modify the answer from here Sequential feature selection Matlab in order to make it work, but the handle part is tricky to me!
I am using
[b, dev, stats] = glmfit(X_train,y_train,'binomial','link','logit')
in order to fit a model, and then I use
y_hat = glmval(b,X_test,'logit');
to evaluate the model output. I have tried to make a handle
f = #(X_test, y_test)...
(sum(y_test ~= round(glmval(b,X_test,'logit'))))
Which just says that I have "Too many input arguments", when I use
sequentialfs(f,X,y)
Can anyone help identify the relevant variables? A data set is below, where all variables will be deemed relevant, but the real problem is of a much larger scale which leads to overparametrization of the model.
X=
0,0305780001742986 0,0293310740486058 0,0289653631914407 0,0313646568650811 0,0308948854477814 0,0298740323895053
0,0221062699789144 0,0213746063391538 0,0196872068542263 0,0209915418572080 0,0206064713419377 0,0198587113064423
0,0275588428138312 0,0273957214651399 0,0291622596392042 0,0313729847567230 0,0314439993783026 0,0307137185244424
0,0262451954064372 0,0230629198767100 0,0251192876802874 0,0243459679829053 0,0235752627390268 0,0245208219450122
0,0232074343987232 0,0272778269415268 0,0288725913067116 0,0274565324127577 0,0283032894902223 0,0287391056368953
0,0237589488887855 0,0251929947662669 0,0209989755767701 0,0146662148369109 0,0193830305676060 0,0198900627308170
0,0275142606053146 0,0311683593689258 0,0287109246912083 0,0307544961383919 0,0293964246310913 0,0291758079280418
0,0284337063240141 0,0249820611584738 0,0261664330153102 0,0270312804219022 0,0269178494606530 0,0273467522864029
0,0279150521116060 0,0314021886143824 0,0291020356476994 0,0278247389567505 0,0294200854382611 0,0306460336509255
0,0270622115094110 0,0317795784913586 0,0278283619288299 0,0307757941373800 0,0292513615541838 0,0283900407512898
0,0270275432108930 0,0330384417745352 0,0323886885104962 0,0330255939800101 0,0329789138848656 0,0333091935226094
0,0263417729025468 0,0243442097895390 0,0253328659546050 0,0270828343149025 0,0262845355278762 0,0257915212526289
0,0247503544929709 0,0282150822748136 0,0282408722769508 0,0306907484707723 0,0284025718962319 0,0280291257508206
0,0282116130443164 0,0259317921438547 0,0316179116969559 0,0300579055064814 0,0315134680888256 0,0299693403497154
0,0221040769734790 0,0232354360252008 0,0231588261739581 0,0240414200785524 0,0209509517094598 0,0223174875964419
0,0264965936690525 0,0312918915850473 0,0297480867085914 0,0349060220702562 0,0307640365732823 0,0299291946182921
0,0283027112468824 0,0288419304885060 0,0275801208398665 0,0239401671924088 0,0263296648119700 0,0260497226349653
0,0298063469363023 0,0253535298515575 0,0245113899712628 0,0158158669753461 0,0228044538675689 0,0221885556611738
0,0276409442724517 0,0283430130710139 0,0303893043659674 0,0314511518633802 0,0315673022208602 0,0302375851656905
0,0270849987059202 0,0312381323489334 0,0301662309393833 0,0290482017036615 0,0299207348490636 0,0295746114571195
0,0225683074444599 0,0297455473987182 0,0241145950178924 0,0233857691372279 0,0259911200866772 0,0248345888658664
0,0267870191916224 0,0254332496269710 0,0270915983261551 0,0263567311441536 0,0267470932454238 0,0279742674970829
0,0271933786309775 0,0274722435013798 0,0249484244285920 0,0301299698898670 0,0255527349811283 0,0263901147510067
0,0262114755708772 0,0168593285634855 0,0205147916736994 0,0227484518393022 0,0187327306255277 0,0197581601499656
0,0314796048983783 0,0281771847088388 0,0318159664952504 0,0325586660052902 0,0315277330661507 0,0324593871738733
0,0265694239044569 0,0239306067986609 0,0263341531447523 0,0277011505766640 0,0274385429019891 0,0258861916796922
0,0248731179403750 0,0253801474063385 0,0258606627949811 0,0234446644496543 0,0262626821946271 0,0265908368467206
0,0268335079060221 0,0327877888534457 0,0292050848084788 0,0286811931594028 0,0288058286572012 0,0297311873407772
0,0289655102183149 0,0297912585631799 0,0289955796469846 0,0301374575223736 0,0286017508461882 0,0294957275181626
0,0291599221376467 0,0284752300175276 0,0294302899166185 0,0291600417637315 0,0304544724881292 0,0299842921064845
0,0306574107981617 0,0282562949634004 0,0290757741762068 0,0266759231631069 0,0276079132984815 0,0283490628634946
0,0251419690541645 0,0190208495552799 0,0219740565645893 0,0229700198654972 0,0231913971059666 0,0218104715761714
0,0279646047804379 0,0213712975401601 0,0266486742241442 0,0282351537040943 0,0256010069497723 0,0254151479565295
0,0291819765619095 0,0274841050997730 0,0277287252877124 0,0253910521460239 0,0270103968729900 0,0280719282161061
0,0267034888169756 0,0230641558482702 0,0254891453796317 0,0246876229024146 0,0232209026694445 0,0255548967304529
0,0282525579813869 0,0275995431014968 0,0263616638576222 0,0282844384170093 0,0262147272435204 0,0257109782177451
0,0293098464976821 0,0323503780295293 0,0266630772869637 0,0228075737924046 0,0263296732877124 0,0254313353506367
y=
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
The function you use for sequentialfs should take in four variables. Yours only accepts two:
criterion = fun(XTRAIN,ytrain,XTEST,ytest)
This is because, even though you give only two variables, X and y, sequentialfs splits these into a training and testing subset (straight from the docs):
Starting from an empty feature set, sequentialfs creates candidate
feature subsets by sequentially adding each of the features not yet
selected. For each candidate feature subset, sequentialfs performs
10-fold cross-validation by repeatedly calling fun with different
training subsets of X and y, XTRAIN and ytrain, and test subsets of X
and y, XTEST and ytest
So the function passed for sequentialfs must take both the training and test subsets, e.g.:
function criterion = my_function(X_train,y_train,X_test,y_test)
[b, dev, stats] = glmfit(X_train,y_train,'binomial','link','logit')
y_hat = glmval(b,X_test,'logit');
criterion = sum(y_test ~= round(y_hat));
end
If this function is on your path you can pass it as #my_function, you don't have to make an anonymous function to get a handle.
I have a set of data shown bellow:
flow Rate (L/min)
Speed(rpm) 1 1.25 1.5 1.75 2 2.25 2.5 2.77 ... 6
Pressure (Pa)
2000 15251.2 15232 15200 15168 15027.2 14912 14752 0 ... 0
2050 16000 15840 15808 15744 15680 15520 15488 15232 ... 0
2100 16384 16256 16217.6 16192 16128 16064 16032 15872 ... 0
2150 17088 17024 16992 16960 16928 16832 16704 16512 ... 0
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
4250 61120 60800 60768 60736 60672 60736 60608 60416 ... 56960
At a specific speed (from 2000-4250rpm) and flow rate (from 1-6 L/min) as shown there are different pressures.
1) i want to know how can i insert a new row in between two of these speeds for example if i have a speed of 2030rpm i want to be able to find in between which two values the 2030rpm is and insert a row on matlab
demonstration hown below:
2000 15251.2 15232 15200 15168 15027.2 14912 14752 0 ... 0
2030 0 0 0 0 0 0 0 0
2050 16000 15840 15808 15744 15680 15520 15488 15232 ... 0
2) my second problem is how can i interpolate between the two values below (where the zero is and get a value.
15232
0
16000
I really appreciate if any one can answer any of my questions preferably the first one so ic an actually get to the second step lol
Thank you so much
newmat = zeros(size(oldmat,1)+1,size(oldmat,2))
newmat(1:x) = oldmat(1:x)
newmat(x+2:end) = oldmat(x+1:end)
where oldmat, newmat are the old and new versions of your matrix and x+1 the index of the row of 0s inserted into newmat.
Then, supposing that you want linear interpolation, something like:
newmat(x+1,:) = newmat(x,:)+0.6*(newmat(x+2,:)-newmat(x,:))
I expect I've made some small errors, and this is quite specific to your example, if you have trouble fixing and generalising, update your question or comment.
Assuming the data is stored in a matrix called p, for automatically positioning the new row in correct sequence:
Append the new row at the end of p, then:
p = sortrows(p)
Following up on the comments, we have:
newrow = [2130, zeros(1,size(test,2)-1)]
p(size(p,1)+1,:) = newrow
p = sortrows(p)
(if 2130 is the first value of the new row.)
This may help you:
% Matrix dimensions
nCols = 10;
nRows = 8;
% Synthetic data
matrix = [ linspace(2000,4250,nRows)' , 2000*rand(nRows,nCols-1)];
matrix([2,4],2:end) = zeros(2,nCols-1); % where some rows are zeros (2 and 4 on this example)
matrix
matrix =
1.0e+03 *
2.0000 1.7810 1.3674 1.4983 0.7329 1.5439 1.5639 0.2246 0.8653 1.5379
2.3214 0 0 0 0 0 0 0 0 0
2.6429 1.4687 1.4454 1.4801 1.3701 0.7765 0.5881 0.5831 0.2195 0.5459
2.9643 0 0 0 0 0 0 0 0 0
3.2857 0.1458 0.2350 1.4699 1.5787 0.4579 1.0617 1.9288 0.3749 1.3466
3.6071 0.1771 1.2814 1.9412 0.7353 1.2839 0.1830 0.8650 0.5324 0.8591
3.9286 1.5967 0.6576 1.7339 0.4121 0.9690 0.8106 1.3895 1.5957 0.9035
4.2500 1.8860 1.3076 0.1725 0.1733 0.3037 0.2097 1.5162 0.9752 1.2197
If you just want to fill the rows whose elements from the second to the last columns are zeros with the average value of the previous and the next rows.
for i=2:nRows-1
if ( sum(matrix(i,2:end))==0 )
matrix(i,2:end) = mean( matrix([i-1,i+1],2:end) );
end
end
matrix
matrix =
1.0e+03 *
2.0000 1.7810 1.3674 1.4983 0.7329 1.5439 1.5639 0.2246 0.8653 1.5379
2.3214 1.6248 1.4064 1.4892 1.0515 1.1602 1.0760 0.4039 0.5424 1.0419
2.6429 1.4687 1.4454 1.4801 1.3701 0.7765 0.5881 0.5831 0.2195 0.5459
2.9643 0.8072 0.8402 1.4750 1.4744 0.6172 0.8249 1.2560 0.2972 0.9462
3.2857 0.1458 0.2350 1.4699 1.5787 0.4579 1.0617 1.9288 0.3749 1.3466
3.6071 0.1771 1.2814 1.9412 0.7353 1.2839 0.1830 0.8650 0.5324 0.8591
3.9286 1.5967 0.6576 1.7339 0.4121 0.9690 0.8106 1.3895 1.5957 0.9035
4.2500 1.8860 1.3076 0.1725 0.1733 0.3037 0.2097 1.5162 0.9752 1.2197
This code assumes that:
You want to fill rows were only the first column element is non-zero.
You want to replace the zeros with the average between previous and next rows values.
You only interpolate inner rows.
I hope it helps.