Related
I am doing two updates for h and x given in this [paper] http://paris.cs.illinois.edu/pubs/nasser-icassp2015.pdf (You don't have to read the paper, just look for the equations 4,10 and 14 for updating h and x given on page 2 and 3).
This is the code snippet that I have tried so far. Can you tell me if its correct? Also, is there any way to optimize these for loops?
In some cases (t-tau) term was negative and MATLAB was giving an error. So, I put a condition that only implement if (t-tau)>0. Doing this is correct or is there any other way to take care of the negative indices?
%updates
Lh=10;
lambda = 0.1*(sum(reverberatedspeechspec(:))/(size(Y,1)*size(Y,2)));
S=reverberatedspeechspec;
W=basis_mel_act; W=gather(W); W = double(W); %W is the dictionary
%---initialization for H(RIR)----
H=rand(size(Y,1),Lh);
nmfIter = 50;
%---initialization for X(Activations)-----
W_trans=W';
X=W_trans*S; X=double(X);
Y = zeros(size(S,1));
for idx = 1 : size(Y,1)
Y(idx,:) = filter(S(idx,:),1,H(idx,:));
end
Stilde = zeros(size(S));
Ytilde = zeros(size(S));
for iter=1:nmfIter
% update for H
Stilde = W*X;
Ytilde = zeros(size(S));
for j=1:size(Stilde,1)
Ytilde (j,:) = filter(Stilde(j,:),1,H(j,:));
end
ratio = Y./Ytilde;
numerator = zeros(size(H));
denominator = numerator;
for k = 1 :size(Y,1)
for tau = 1:Lh
for t= 1:size(Y,2)
if gt (t-tau , 0)
numerator (k,tau) = numerator(k,tau) + ratio(k,t) * Stilde(k,t-tau);
denominator(k,tau) = denominator (k,tau) + Stilde (k,t-tau);
end
end
end
end
H = H .* numerator ./denominator ;
%updating Ytilde after getting a new value for H
for j=1:size(Stilde,1)
Ytilde (j,:) = filter(Stilde(j,:),1,H(j,:));
end
%update for X
ratio = Y./Ytilde;
ratio = [ratio zeros(size(Y,1),Lh)]; %zero padding Y and Ytilde for (t+tau) term in update of X.
Product = H.' * W; % Product of H_transpose and W in th update which is equivalent to the term ∑H(k,tau)W(k,r)
numerator = zeros(size(H));
denominator = numerator;
for r = 1:size(W,2)
for t = 1:size(Y,2)
for k = 1:size(Y,1)
for tau = 1:Lh
numerator(r,t) = numerator(r,t) + ratio(k,t+tau) * Product(tau,r);
denominator(r,t) = denominator(r,t) + Product(tau,r) + lambda;
end
end
end
end
X = X .* numerator ./denominator;
end
I'm writing a script for an aerodynamics class and I'm getting the following error:
Undefined function or variable 'dCt_dx'.
Error in Project2_Iteration (line 81)
Ct = trapz(x,dCt_dx)
I'm not sure what the cause is. It's something to do with my if statement. My script is below:
clear all
clc
global dr a n Vinf Vr w rho k x c cl dr B R beta t
%Environmental Parameters
n = 2400; %rpm
Vinf = 154; %KTAS
rho = 0.07647 * (.7429/.9450); %from mattingly for 8kft
a = 1084; %speed of sound, ft/s, 8000 ft
n = n/60; %convert to rps
w = 2*pi*n;
Vinf = (Vinf*6076.12)/3600; %convert from KTAS to ft/s
k = length(c);
dr = R/k; %length of each blade element
for i = 1:k
r(i) = i*dr - (.5*dr); %radius at center of blade element
dA = 2*pi*r*dr; %Planform area of blade element
x(i) = r(i)/R;
if x(i) > .15 && x(i-1) < .15
i_15 = i;
end
if x(i) > .75 && x(i-1) < .75
i_75h = i;
i_75l = i-1;
end
Vr(i) = w*r(i) + Vinf;
%Aerodynamic Parameters
M = Vr(i)/a;
if M > 0.9
M = 0.9;
end
m0 = 0.9*(2*pi/(1-M^2)^0.5); %lift-curve slope (2pi/rad)
%1: Calculate phi
phi = atan(Vinf/(2*pi*n*r(i)));
%2: Choose Vo
Vo = .00175*Vinf;
%3: Calculate Theta
theta = atan((Vinf + Vo)/(2*pi*n*r(i)))-phi;
%4:
if option == 1
%calculate cl(i) from c(i)
sigma = (B*c(i))/(pi*R);
if sigma > 0
cl(i) = (8*x(i)*theta*cos(phi)*tan(phi+theta))/sigma;
else
cl(i) = 0;
end
else %option == 2
%calculate c(i) from cl(i)
if cl(i) ~= 0
sigma = (8*x(i)*theta*cos(phi)*tan(phi+theta))/cl(i);
else
sigma = 0;
end
c(i) = (sigma*pi*R)/B;
if c(i) < 0
c(i) = 0;
end
end
%5: Calculate cd
cd(i) = 0.0090 + 0.0055*(cl(i)-0.1)^2;
%6: calculate alpha
alpha = cl(i)/m0;
%7: calculate beta
beta(i) = phi + alpha + theta;
%8: calculate dCt/dx and dCq/dx
phi0 = phi+theta;
lambda_t = (1/(cos(phi)^2))*(cl(i)*cos(phi0) - cd(i)*sin(phi0));
lambda_q = (1/(cos(phi)^2))*(cl(i)*sin(phi0) + cd(i)*cos(phi0));
if x(i) >= 0.15
dCt_dx(i) = ((pi^3)*(x(i)^2)*sigma*lambda_t)/8; %Roskam eq. 7.47, pg. 280
dCq_dx(i) = ((pi^3)*(x(i)^3)*sigma*lambda_q)/16; %Roskam eq. 7.48, pg 280
else
dCt_dx(i) = 0;
dCq_dx(i) = 0;
end
%calculate Mdd
t(i) = (0.04/(x(i)^1.2))*c(i);
Mdd(i) = 0.94 - (t(i)/c(i)) - cl(i)/10;
end
%9: calculate Ct, Cq, Cd
Ct = trapz(x,dCt_dx)
Cq = trapz(x,dCq_dx)
D = 2*R;
Q=(rho*(n^2)*(D^5)*Cq)
T=(rho*(n^2)*(D^4)*Ct)
When I step through your script, I see that the the entire for i = 1:k loop is skipped because k=0. You set k = length(c), but c was never initialized to a value, so it has length zero.
Because of this, dCt_dx is never given a value--and more importantly the majority of your script is never run.
If you're going to be using MATLAB in the future, I really suggest learning how to do this. It makes it a lot easier to find bugs. Try looking at this video.
I have a 3D volume and a 2D image and an approximate mapping (affine transformation with no skwewing, known scaling, rotation and translation approximately known and need fitting) between the two. Because there is an error in this mapping and I would like to further register the 2D image to the 3D volume. I have not written code for registration purposes before, but because I can't find any programs or code to solve this I would like to try and do this. I believe the standard for registration is to optimize mutual information. I think this would also be suitable here, because the intensities are not equal between the two images. So I think I should make a function for the transformation, a function for the mutual information and a function for optimization.
I did find some Matlab code on a mathworks thread from two years ago, based on an article. The OP reports that she managed to get the code to work, but I'm not getting how she did that exactly. Also in the IP package for matlab there is an implementation, but I dont have that package and there does not seem to be an equivalent for octave. SPM is a program that uses matlab and has registration implemented, but does not cope with 2d to 3d registration. On the file exchange there is a brute force method that registers two 2D images using mutual information.
What she does is pass a multi planar reconstruction function and an similarity/error function into a minimization algorithm. But the details I don't quite understand. Maybe it would be better to start fresh:
load mri; volume = squeeze(D);
phi = 3; theta = 2; psi = 5; %some small angles
tx = 1; ty = 1; tz = 1; % some small translation
dx = 0.25, dy = 0.25, dz = 2; %different scales
t = [tx; ty; tz];
r = [phi, theta, psi]; r = r*(pi/180);
dims = size(volume);
p0 = [round(dims(1)/2);round(dims(2)/2);round(dims(3)/2)]; %image center
S = eye(4); S(1,1) = dx; S(2,2) = dy; S(3,3) = dz;
Rx=[1 0 0 0;
0 cos(r(1)) sin(r(1)) 0;
0 -sin(r(1)) cos(r(1)) 0;
0 0 0 1];
Ry=[cos(r(2)) 0 -sin(r(2)) 0;
0 1 0 0;
sin(r(2)) 0 cos(r(2)) 0;
0 0 0 1];
Rz=[cos(r(3)) sin(r(3)) 0 0;
-sin(r(3)) cos(r(3)) 0 0;
0 0 1 0;
0 0 0 1];
R = S*Rz*Ry*Rx;
%make affine matrix to rotate about center of image
T1 = ( eye(3)-R(1:3,1:3) ) * p0(1:3);
T = T1 + t; %add translation
A = R;
A(1:3,4) = T;
Rold2new = A;
Rnew2old = inv(Rold2new);
%the transformation
[xx yy zz] = meshgrid(1:dims(1),1:dims(2),1:1);
coordinates_axes_new = [xx(:)';yy(:)';zz(:)'; ones(size(zz(:)))'];
coordinates_axes_old = Rnew2old*coordinates_axes_new;
Xcoordinates = reshape(coordinates_axes_old(1,:), dims(1), dims(2), dims(3));
Ycoordinates = reshape(coordinates_axes_old(2,:), dims(1), dims(2), dims(3));
Zcoordinates = reshape(coordinates_axes_old(3,:), dims(1), dims(2), dims(3));
%interpolation/reslicing
method = 'cubic';
slice= interp3(volume, Xcoordinates, Ycoordinates, Zcoordinates, method);
%so now I have my slice for which I would like to find the correct position
% first guess for A
A0 = eye(4); A0(1:3,4) = T1; A0(1,1) = dx; A0(2,2) = dy; A0(3,3) = dz;
% this is pretty close to A
% now how would I fit the slice to the volume by changing A0 and examining some similarity measure?
% probably maximize mutual information?
% http://www.mathworks.com/matlabcentral/fileexchange/14888-mutual-information-computation/content//mi/mutualinfo.m
Ok I was hoping for someone else's approach, that would probably have been better than mine as I have never done any optimization or registration before. So I waited for Knedlsepps bounty to almost finish. But I do have some code thats working now. It will find a local optimum so the initial guess must be good. I wrote some functions myself, took some functions from the file exchange as is and I extensively edited some other functions from the file exchange. Now that I put all the code together to work as an example here, the rotations are off, will try and correct that. Im not sure where the difference in code is between the example and my original code, must have made a typo in replacing some variables and data loading scheme.
What I do is I take the starting affine transformation matrix, decompose it to an orthogonal matrix and an upper triangular matrix. I then assume the orthogonal matrix is my rotation matrix so I calculate the euler angles from that. I directly take the translation from the affine matrix and as stated in the problem I assume I know the scaling matrix and there is no shearing. So then I have all degrees of freedom for the affine transformation, which my optimisation function changes and constructs a new affine matrix from, applies it to the volume and calculates the mutual information. The matlab optimisation function patternsearch then minimises 1-MI/MI_max.
What I noticed when using it on my real data which are multimodal brain images is that it works much better on brain extracted images, so with the skull and tissue outside of the skull removed.
%data
load mri; volume = double(squeeze(D));
%transformation parameters
phi = 3; theta = 1; psi = 5; %some small angles
tx = 1; ty = 1; tz = 3; % some small translation
dx = 0.25; dy = 0.25; dz = 2; %different scales
t = [tx; ty; tz];
r = [phi, theta, psi]; r = r*(pi/180);
%image center and size
dims = size(volume);
p0 = [round(dims(1)/2);round(dims(2)/2);round(dims(3)/2)];
%slice coordinate ranges
range_x = 1:dims(1)/dx;
range_y = 1:dims(2)/dy;
range_z = 1;
%rotation
R = dofaffine([0;0;0], r, [1,1,1]);
T1 = ( eye(3)-R(1:3,1:3) ) * p0(1:3); %rotate about p0
%scaling
S = eye(4); S(1,1) = dx; S(2,2) = dy; S(3,3) = dz;
%translation
T = [[eye(3), T1 + t]; [0 0 0 1]];
%affine
A = T*R*S;
% first guess for A
r00 = [1,1,1]*pi/180;
R00 = dofaffine([0;0;0], r00, [1 1 1]);
t00 = T1 + t + ( eye(3) - R00(1:3,1:3) ) * p0;
A0 = dofaffine( t00, r00, [dx, dy, dz] );
[ t0, r0, s0 ] = dofaffine( A0 );
x0 = [ t0.', r0, s0 ];
%the transformation
slice = affine3d(volume, A, range_x, range_y, range_z, 'cubic');
guess = affine3d(volume, A0, range_x, range_y, range_z, 'cubic');
%initialisation
Dt = [1; 1; 1];
Dr = [2 2 2].*pi/180;
Ds = [0 0 0];
Dx = [Dt', Dr, Ds];
%limits
LB = x0-Dx;
UB = x0+Dx;
%other inputs
ref_levels = length(unique(slice));
Qref = imquantize(slice, ref_levels);
MI_max = MI_GG(Qref, Qref);
%patternsearch options
options = psoptimset('InitialMeshSize',0.03,'MaxIter',20,'TolCon',1e-5,'TolMesh',5e-5,'TolX',1e-6,'PlotFcns',{#psplotbestf,#psplotbestx});
%optimise
[x2, MI_norm_neg, exitflag_len] = patternsearch(#(x) AffRegOptFunc(x, slice, volume, MI_max, x0), x0,[],[],[],[],LB(:),UB(:),options);
%check
p0 = [round(size(volume)/2).'];
R0 = dofaffine([0;0;0], x2(4:6)-x0(4:6), [1 1 1]);
t1 = ( eye(3) - R0(1:3,1:3) ) * p0;
A2 = dofaffine( x2(1:3).'+t1, x2(4:6), x2(7:9) ) ;
fitted = affine3d(volume, A2, range_x, range_y, range_z, 'cubic');
overlay1 = imfuse(slice, guess);
overlay2 = imfuse(slice, fitted);
figure(101);
ax(1) = subplot(1,2,1); imshow(overlay1, []); title('pre-reg')
ax(2) = subplot(1,2,2); imshow(overlay2, []); title('post-reg');
linkaxes(ax);
function normed_score = AffRegOptFunc( x, ref_im, reg_im, MI_max, x0 )
t = x(1:3).';
r = x(4:6);
s = x(7:9);
rangx = 1:size(ref_im,1);
rangy = 1:size(ref_im,2);
rangz = 1:size(ref_im,3);
ref_levels = length(unique(ref_im));
reg_levels = length(unique(reg_im));
t0 = x0(1:3).';
r0 = x0(4:6);
s0 = x0(7:9);
p0 = [round(size(reg_im)/2).'];
R = dofaffine([0;0;0], r-r0, [1 1 1]);
t1 = ( eye(3) - R(1:3,1:3) ) * p0;
t = t + t1;
Ap = dofaffine( t, r, s );
reg_im_t = affine3d(reg_im, A, rangx, rangy, rangz, 'cubic');
Qref = imquantize(ref_im, ref_levels);
Qreg = imquantize(reg_im_t, reg_levels);
MI = MI_GG(Qref, Qreg);
normed_score = 1-MI/MI_max;
end
function [ varargout ] = dofaffine( varargin )
% [ t, r, s ] = dofaffine( A )
% [ A ] = dofaffine( t, r, s )
if nargin == 1
%affine to degrees of freedom (no shear)
A = varargin{1};
[T, R, S] = decompaffine(A);
r = GetEulerAngles(R(1:3,1:3));
s = [S(1,1), S(2,2), S(3,3)];
t = T(1:3,4);
varargout{1} = t;
varargout{2} = r;
varargout{3} = s;
elseif nargin == 3
%degrees of freedom to affine (no shear)
t = varargin{1};
r = varargin{2};
s = varargin{3};
R = GetEulerAngles(r); R(4,4) = 1;
S(1,1) = s(1); S(2,2) = s(2); S(3,3) = s(3); S(4,4) = 1;
T = eye(4); T(1,4) = t(1); T(2,4) = t(2); T(3,4) = t(3);
A = T*R*S;
varargout{1} = A;
else
error('incorrect number of input arguments');
end
end
function [ T, R, S ] = decompaffine( A )
%I assume A = T * R * S
T = eye(4);
R = eye(4);
S = eye(4);
%decompose in orthogonal matrix q and upper triangular matrix r
%I assume q is a rotation matrix and r is a scale and shear matrix
%matlab 2014 can force real solution
[q r] = qr(A(1:3,1:3));
R(1:3,1:3) = q;
S(1:3,1:3) = r;
% A*S^-1*R^-1 = T*R*S*S^-1*R^-1 = T*R*I*R^-1 = T*R*R^-1 = T*I = T
T = A*inv(S)*inv(R);
t = T(1:3,4);
T = [eye(4) + [[0 0 0;0 0 0;0 0 0;0 0 0],[t;0]]];
end
function [varargout]= GetEulerAngles(R)
assert(length(R)==3)
dims = size(R);
if min(dims)==1
rx = R(1); ry = R(2); rz = R(3);
R = [[ cos(ry)*cos(rz), -cos(ry)*sin(rz), sin(ry)];...
[ cos(rx)*sin(rz) + cos(rz)*sin(rx)*sin(ry), cos(rx)*cos(rz) - sin(rx)*sin(ry)*sin(rz), -cos(ry)*sin(rx)];...
[ sin(rx)*sin(rz) - cos(rx)*cos(rz)*sin(ry), cos(rz)*sin(rx) + cos(rx)*sin(ry)*sin(rz), cos(rx)*cos(ry)]];
varargout{1} = R;
else
ry=asin(R(1,3));
rz=acos(R(1,1)/cos(ry));
rx=acos(R(3,3)/cos(ry));
if nargout > 1 && nargout < 4
varargout{1} = rx;
varargout{2} = ry;
varargout{3} = rz;
elseif nargout == 1
varargout{1} = [rx ry rz];
else
error('wrong number of output arguments');
end
end
end
I asked a question a few days before but I guess it was a little too complicated and I don't expect to get any answer.
My problem is that I need to use ANN for classification. I've read that much better cost function (or loss function as some books specify) is the cross-entropy, that is J(w) = -1/m * sum_i( yi*ln(hw(xi)) + (1-yi)*ln(1 - hw(xi)) ); i indicates the no. data from training matrix X. I tried to apply it in MATLAB but I find it really difficult. There are couple things I don't know:
should I sum each outputs given all training data (i = 1, ... N, where N is number of inputs for training)
is the gradient calculated correctly
is the numerical gradient (gradAapprox) calculated correctly.
I have following MATLAB codes. I realise I may ask for trivial thing but anyway I hope someone can give me some clues how to find the problem. I suspect the problem is to calculate gradients.
Many thanks.
Main script:
close all
clear all
L = #(x) (1 + exp(-x)).^(-1);
NN = #(x,theta) theta{2}*[ones(1,size(x,1));L(theta{1}*[ones(size(x,1),1) x]')];
% theta = [10 -30 -30];
x = [0 0; 0 1; 1 0; 1 1];
y = [0.9 0.1 0.1 0.1]';
theta0 = 2*rand(9,1)-1;
options = optimset('gradObj','on','Display','iter');
thetaVec = fminunc(#costFunction,theta0,options,x,y);
theta = cell(2,1);
theta{1} = reshape(thetaVec(1:6),[2 3]);
theta{2} = reshape(thetaVec(7:9),[1 3]);
NN(x,theta)'
Cost function:
function [jVal,gradVal,gradApprox] = costFunction(thetaVec,x,y)
persistent index;
% 1 x x
% 1 x x
% 1 x x
% x = 1 x x
% 1 x x
% 1 x x
% 1 x x
m = size(x,1);
if isempty(index) || index > size(x,1)
index = 1;
end
L = #(x) (1 + exp(-x)).^(-1);
NN = #(x,theta) theta{2}*[ones(1,size(x,1));L(theta{1}*[ones(size(x,1),1) x]')];
theta = cell(2,1);
theta{1} = reshape(thetaVec(1:6),[2 3]);
theta{2} = reshape(thetaVec(7:9),[1 3]);
Dew = cell(2,1);
DewApprox = cell(2,1);
% Forward propagation
a0 = x(index,:)';
z1 = theta{1}*[1;a0];
a1 = L(z1);
z2 = theta{2}*[1;a1];
a2 = L(z2);
% Back propagation
d2 = 1/m*(a2 - y(index))*L(z2)*(1-L(z2));
Dew{2} = [1;a1]*d2;
d1 = [1;a1].*(1 - [1;a1]).*theta{2}'*d2;
Dew{1} = [1;a0]*d1(2:end)';
% NNRes = NN(x,theta)';
% jVal = -1/m*sum(NNRes-y)*NNRes*(1-NNRes);
jVal = -1/m*(a2 - y(index))*a2*(1-a2);
gradVal = [Dew{1}(:);Dew{2}(:)];
gradApprox = CalcGradApprox(0.0001);
index = index + 1;
function output = CalcGradApprox(epsilon)
output = zeros(size(gradVal));
for n=1:length(thetaVec)
thetaVecMin = thetaVec;
thetaVecMax = thetaVec;
thetaVecMin(n) = thetaVec(n) - epsilon;
thetaVecMax(n) = thetaVec(n) + epsilon;
thetaMin = cell(2,1);
thetaMax = cell(2,1);
thetaMin{1} = reshape(thetaVecMin(1:6),[2 3]);
thetaMin{2} = reshape(thetaVecMin(7:9),[1 3]);
thetaMax{1} = reshape(thetaVecMax(1:6),[2 3]);
thetaMax{2} = reshape(thetaVecMax(7:9),[1 3]);
a2min = NN(x(index,:),thetaMin)';
a2max = NN(x(index,:),thetaMax)';
jValMin = -1/m*(a2min-y(index))*a2min*(1-a2min);
jValMax = -1/m*(a2max-y(index))*a2max*(1-a2max);
output(n) = (jValMax - jValMin)/2/epsilon;
end
end
end
EDIT:
Below I present the correct version of my costFunction for those who may be interested.
function [jVal,gradVal,gradApprox] = costFunction(thetaVec,x,y)
m = size(x,1);
L = #(x) (1 + exp(-x)).^(-1);
NN = #(x,theta) L(theta{2}*[ones(1,size(x,1));L(theta{1}*[ones(size(x,1),1) x]')]);
theta = cell(2,1);
theta{1} = reshape(thetaVec(1:6),[2 3]);
theta{2} = reshape(thetaVec(7:9),[1 3]);
Delta = cell(2,1);
Delta{1} = zeros(size(theta{1}));
Delta{2} = zeros(size(theta{2}));
D = cell(2,1);
D{1} = zeros(size(theta{1}));
D{2} = zeros(size(theta{2}));
jVal = 0;
for in = 1:size(x,1)
% Forward propagation
a1 = [1;x(in,:)']; % added bias to a0
z2 = theta{1}*a1;
a2 = [1;L(z2)]; % added bias to a1
z3 = theta{2}*a2;
a3 = L(z3);
% Back propagation
d3 = a3 - y(in);
d2 = theta{2}'*d3.*a2.*(1 - a2);
Delta{2} = Delta{2} + d3*a2';
Delta{1} = Delta{1} + d2(2:end)*a1';
jVal = jVal + sum( y(in)*log(a3) + (1-y(in))*log(1-a3) );
end
D{1} = 1/m*Delta{1};
D{2} = 1/m*Delta{2};
jVal = -1/m*jVal;
gradVal = [D{1}(:);D{2}(:)];
gradApprox = CalcGradApprox(x(in,:),0.0001);
% Nested function to calculate gradApprox
function output = CalcGradApprox(x,epsilon)
output = zeros(size(thetaVec));
for n=1:length(thetaVec)
thetaVecMin = thetaVec;
thetaVecMax = thetaVec;
thetaVecMin(n) = thetaVec(n) - epsilon;
thetaVecMax(n) = thetaVec(n) + epsilon;
thetaMin = cell(2,1);
thetaMax = cell(2,1);
thetaMin{1} = reshape(thetaVecMin(1:6),[2 3]);
thetaMin{2} = reshape(thetaVecMin(7:9),[1 3]);
thetaMax{1} = reshape(thetaVecMax(1:6),[2 3]);
thetaMax{2} = reshape(thetaVecMax(7:9),[1 3]);
a3min = NN(x,thetaMin)';
a3max = NN(x,thetaMax)';
jValMin = 0;
jValMax = 0;
for inn=1:size(x,1)
jValMin = jValMin + sum( y(inn)*log(a3min) + (1-y(inn))*log(1-a3min) );
jValMax = jValMax + sum( y(inn)*log(a3max) + (1-y(inn))*log(1-a3max) );
end
jValMin = 1/m*jValMin;
jValMax = 1/m*jValMax;
output(n) = (jValMax - jValMin)/2/epsilon;
end
end
end
I've only had a quick eyeball over your code. Here are some pointers.
Q1
should I sum each outputs given all training data (i = 1, ... N, where
N is number of inputs for training)
If you are talking in relation to the cost function, it is normal to sum and normalise by the number of training examples in order to provide comparison between.
I can't tell from the code whether you have a vectorised implementation which will change the answer. Note that the sum function will only sum up a single dimension at a time - meaning if you have a (M by N) array, sum will result in a 1 by N array.
The cost function should have a scalar output.
Q2
is the gradient calculated correctly
The gradient is not calculated correctly - specifically the deltas look wrong. Try following Andrew Ng's notes [PDF] they are very good.
Q3
is the numerical gradient (gradAapprox) calculated correctly.
This line looks a bit suspect. Does this make more sense?
output(n) = (jValMax - jValMin)/(2*epsilon);
EDIT: I actually can't make heads or tails of your gradient approximation. You should only use forward propagation and small tweaks in the parameters to compute the gradient. Good luck!
I've written some code to implement an algorithm that takes as input a vector q of real numbers, and returns as an output a complex matrix R. The Matlab code below produces a plot showing the input vector q and the output matrix R.
Given only the complex matrix output R, I would like to obtain the input vector q. Can I do this using least-squares optimization? Since there is a recursive running sum in the code (rs_r and rs_i), the calculation for a column of the output matrix is dependent on the calculation of the previous column.
Perhaps a non-linear optimization can be set up to recompose the input vector q from the output matrix R?
Looking at this in another way, I've used an algorithm to compute a matrix R. I want to run the algorithm "in reverse," to get the input vector q from the output matrix R.
If there is no way to recompose the starting values from the output, thereby treating the problem as a "black box," then perhaps the mathematics of the model itself can be used in the optimization? The program evaluates the following equation:
The Utilde(tau, omega) is the output matrix R. The tau (time) variable comprises the columns of the response matrix R, whereas the omega (frequency) variable comprises the rows of the response matrix R. The integration is performed as a recursive running sum from tau = 0 up to the current tau timestep.
Here are the plots created by the program posted below:
Here is the full program code:
N = 1001;
q = zeros(N, 1); % here is the input
q(1:200) = 55;
q(201:300) = 120;
q(301:400) = 70;
q(401:600) = 40;
q(601:800) = 100;
q(801:1001) = 70;
dt = 0.0042;
fs = 1 / dt;
wSize = 101;
Glim = 20;
ginv = 0;
R = get_response(N, q, dt, wSize, Glim, ginv); % R is output matrix
rows = wSize;
cols = N;
figure; plot(q); title('q value input as vector');
ylim([0 200]); xlim([0 1001])
figure; imagesc(abs(R)); title('Matrix output of algorithm')
colorbar
Here is the function that performs the calculation:
function response = get_response(N, Q, dt, wSize, Glim, ginv)
fs = 1 / dt;
Npad = wSize - 1;
N1 = wSize + Npad;
N2 = floor(N1 / 2 + 1);
f = (fs/2)*linspace(0,1,N2);
omega = 2 * pi .* f';
omegah = 2 * pi * f(end);
sigma2 = exp(-(0.23*Glim + 1.63));
sign = 1;
if(ginv == 1)
sign = -1;
end
ratio = omega ./ omegah;
rs_r = zeros(N2, 1);
rs_i = zeros(N2, 1);
termr = zeros(N2, 1);
termi = zeros(N2, 1);
termr_sub1 = zeros(N2, 1);
termi_sub1 = zeros(N2, 1);
response = zeros(N2, N);
% cycle over cols of matrix
for ti = 1:N
term0 = omega ./ (2 .* Q(ti));
gamma = 1 / (pi * Q(ti));
% calculate for the real part
if(ti == 1)
Lambda = ones(N2, 1);
termr_sub1(1) = 0;
termr_sub1(2:end) = term0(2:end) .* (ratio(2:end).^-gamma);
else
termr(1) = 0;
termr(2:end) = term0(2:end) .* (ratio(2:end).^-gamma);
rs_r = rs_r - dt.*(termr + termr_sub1);
termr_sub1 = termr;
Beta = exp( -1 .* -0.5 .* rs_r );
Lambda = (Beta + sigma2) ./ (Beta.^2 + sigma2); % vector
end
% calculate for the complex part
if(ginv == 1)
termi(1) = 0;
termi(2:end) = (ratio(2:end).^(sign .* gamma) - 1) .* omega(2:end);
else
termi = (ratio.^(sign .* gamma) - 1) .* omega;
end
rs_i = rs_i - dt.*(termi + termi_sub1);
termi_sub1 = termi;
integrand = exp( 1i .* -0.5 .* rs_i );
if(ginv == 1)
response(:,ti) = Lambda .* integrand;
else
response(:,ti) = (1 ./ Lambda) .* integrand;
end
end % ti loop
No, you cannot do so unless you know the "model" itself for this process. If you intend to treat the process as a complete black box, then it is impossible in general, although in any specific instance, anything can happen.
Even if you DO know the underlying process, then it may still not work, as any least squares estimator is dependent on the starting values, so if you do not have a good guess there, it may converge to the wrong set of parameters.
It turns out that by using the mathematics of the model, the input can be estimated. This is not true in general, but for my problem it seems to work. The cumulative integral is eliminated by a partial derivative.
N = 1001;
q = zeros(N, 1);
q(1:200) = 55;
q(201:300) = 120;
q(301:400) = 70;
q(401:600) = 40;
q(601:800) = 100;
q(801:1001) = 70;
dt = 0.0042;
fs = 1 / dt;
wSize = 101;
Glim = 20;
ginv = 0;
R = get_response(N, q, dt, wSize, Glim, ginv);
rows = wSize;
cols = N;
cut_val = 200;
imagLogR = imag(log(R));
Mderiv = zeros(rows, cols-2);
for k = 1:rows
val = deriv_3pt(imagLogR(k,:), dt);
val(val > cut_val) = 0;
Mderiv(k,:) = val(1:end-1);
end
disp('Running iteration');
q0 = 10;
q1 = 500;
NN = cols - 2;
qout = zeros(NN, 1);
for k = 1:NN
data = Mderiv(:,k);
qout(k) = fminbnd(#(q) curve_fit_to_get_q(q, dt, rows, data),q0,q1);
end
figure; plot(q); title('q value input as vector');
ylim([0 200]); xlim([0 1001])
figure;
plot(qout); title('Reconstructed q')
ylim([0 200]); xlim([0 1001])
Here are the supporting functions:
function output = deriv_3pt(x, dt)
% Function to compute dx/dt using the 3pt symmetrical rule
% dt is the timestep
N = length(x);
N0 = N - 1;
output = zeros(N0, 1);
denom = 2 * dt;
for k = 2:N0
output(k - 1) = (x(k+1) - x(k-1)) / denom;
end
function sse = curve_fit_to_get_q(q, dt, rows, data)
fs = 1 / dt;
N2 = rows;
f = (fs/2)*linspace(0,1,N2); % vector for frequency along cols
omega = 2 * pi .* f';
omegah = 2 * pi * f(end);
ratio = omega ./ omegah;
gamma = 1 / (pi * q);
termi = ((ratio.^(gamma)) - 1) .* omega;
Error_Vector = termi - data;
sse = sum(Error_Vector.^2);