Search Results

Search found 1051 results on 43 pages for 'matlab'.

Page 31/43 | < Previous Page | 27 28 29 30 31 32 33 34 35 36 37 38  | Next Page >

  • Generating lognormally distributed random number from mean, coeff of variation

    - by Richie Cotton
    Most functions for generating lognormally distributed random numbers take the mean and standard deviation of the associated normal distribution as parameters. My problem is that I only know the mean and the coefficient of variation of the lognormal distribution. It is reasonably straight forward to derive the parameters I need for the standard functions from what I have: If mu and sigma are the mean and standard deviation of the associated normal distribution, we know that coeffOfVar^2 = variance / mean^2 = (exp(sigma^2) - 1) * exp(2*mu + sigma^2) / exp(mu + sigma^2/2)^2 = exp(sigma^2) - 1 We can rearrange this to sigma = sqrt(log(coeffOfVar^2 + 1)) We also know that mean = exp(mu + sigma^2/2) This rearranges to mu = log(mean) - sigma^2/2 Here's my R implementation rlnorm0 <- function(mean, coeffOfVar, n = 1e6) { sigma <- sqrt(log(coeffOfVar^2 + 1)) mu <- log(mean) - sigma^2 / 2 rlnorm(n, mu, sigma) } It works okay for small coefficients of variation r1 <- rlnorm0(2, 0.5) mean(r1) # 2.000095 sd(r1) / mean(r1) # 0.4998437 But not for larger values r2 <- rlnorm0(2, 50) mean(r2) # 2.048509 sd(r2) / mean(r2) # 68.55871 To check that it wasn't an R-specific issue, I reimplemented it in MATLAB. (Uses stats toolbox.) function y = lognrnd0(mean, coeffOfVar, sizeOut) if nargin < 3 || isempty(sizeOut) sizeOut = [1e6 1]; end sigma = sqrt(log(coeffOfVar.^2 + 1)); mu = log(mean) - sigma.^2 ./ 2; y = lognrnd(mu, sigma, sizeOut); end r1 = lognrnd0(2, 0.5); mean(r1) % 2.0013 std(r1) ./ mean(r1) % 0.5008 r2 = lognrnd0(2, 50); mean(r2) % 1.9611 std(r2) ./ mean(r2) % 22.61 Same problem. The question is, why is this happening? Is it just that the standard deviation is not robust when the variation is that wide? Or have a screwed up somewhere?

    Read the article

  • function taking in an input image and different kernel size

    - by drifterOcean19
    I have this filtering function that takes an input image, performs convolution using a given kernel, and returns the resulting image. However, I can't seem to work it out how to make it takes different kernel sizes.For example instead of pre-defined 3x3 kernel as below in the code, it could instead take 5x5 or 7x7. and then the user could input the type of kernel/filter they want(Depending on the intended effect). I can't seem to put my head around it. i'm quite new to matlab. function [newImg] = kernelFunc(imgB) img=imread(imgB); figure,imshow(img); img2=zeros(size(img)+2); newImg=zeros(size(img)); for rgb=1:3 for x=1:size(img,1) for y=1:size(img,2) img2(x+1,y+1,rgb)=img(x,y,rgb); end end end for rgb=1:3 for i= 1:size(img2,1)-2 for j=1:size(img2,2)-2 window=zeros(9,1); inc=1; for x=1:3 for y=1:3 window(inc)=img2(i+x-1,j+y-1,rgb); inc=inc+1; end end kernel=[1;2;1;2;4;2;1;2;1]/16; med=window.*kernel; disp(med); med=sum(med); med=floor(med); newImg(i,j,rgb)=med; end end end newImg=uint8(newImg); figure,imshow(newImg); end

    Read the article

  • Is it possible that we override global variables?

    - by Ram Moj
    I have this function: function example(y) global TICTOC; tic TICTOC=5; toc end and I expect TICTOC=5 change the result of toc, since TICTOC is a global variable in tic and toc functions; but this is not the case; Does anyone know the reason? I like to know the answer, because I 'm worried to declare a global variable, which it's name has been declared global in some other functions, I'm not aware of. I saw this function in matlab 2008b help function tic % TIC Start a stopwatch timer. % TIC; any stuff; TOC % prints the time required. % See also: TOC, CLOCK. global TICTOC TICTOC = clock; function t = toc % TOC Read the stopwatch timer. % TOC prints the elapsed time since TIC was used. % t = TOC; saves elapsed time in t, does not print. % See also: TIC, ETIME. global TICTOC if nargout < 1 elapsed_time = etime(clock, TICTOC) else t = etime(clock, TICTOC); end thanks.

    Read the article

  • How can I correctly calculate the direction for a moving object?

    - by Jakub Hampl
    I'm solving the following problem: I have an object and I know its position now and its position 300ms ago. I assume the object is moving. I have a point to which I want the object to get. What I need is to get the angle from my current object to the destination point in such a format that I know whether to turn left or right. The idea is to assume the current angle from the last known position and the current position. I'm trying to solve this in MATLAB. I've tried using several variations with atan2 but either I get the wrong angle in some situations (like when my object is going in circles) or I get the wrong angle in all situations. Examples of code that screws up: a = new - old; b = dest - new; alpha = atan2(a(2) - b(2), a(1) - b(1); where new is the current position (eg. x = 40; y = 60; new = [x y];), old is the 300ms old position and dest is the destination point. Edit Here's a picture to demonstrate the problem with a few examples: In the above image there are a few points plotted and annotated. The black line indicates our estimated current facing of the object. If the destination point is dest1 I would expect an angle of about 88°. If the destination point is dest2 I would expect an angle of about 110°. If the destination point is dest3 I would expect an angle of about -80°.

    Read the article

  • Out-of-memory algorithms for addressing large arrays

    - by reve_etrange
    I am trying to deal with a very large dataset. I have k = ~4200 matrices (varying sizes) which must be compared combinatorially, skipping non-unique and self comparisons. Each of k(k-1)/2 comparisons produces a matrix, which must be indexed against its parents (i.e. can find out where it came from). The convenient way to do this is to (triangularly) fill a k-by-k cell array with the result of each comparison. These are ~100 X ~100 matrices, on average. Using single precision floats, it works out to 400 GB overall. I need to 1) generate the cell array or pieces of it without trying to place the whole thing in memory and 2) access its elements (and their elements) in like fashion. My attempts have been inefficient due to reliance on MATLAB's eval() as well as save and clear occurring in loops. for i=1:k [~,m] = size(data{i}); cur_var = ['H' int2str(i)]; %# if i == 1; save('FileName'); end; %# If using a single MAT file and need to create it. eval([cur_var ' = cell(1,k-i);']); for j=i+1:k [~,n] = size(data{j}); eval([cur_var '{i,j} = zeros(m,n,''single'');']); eval([cur_var '{i,j} = compare(data{i},data{j});']); end save(cur_var,cur_var); %# Add '-append' when using a single MAT file. clear(cur_var); end The other thing I have done is to perform the split when mod((i+j-1)/2,max(factor(k(k-1)/2))) == 0. This divides the result into the largest number of same-size pieces, which seems logical. The indexing is a little more complicated, but not too bad because a linear index could be used. Does anyone know/see a better way?

    Read the article

  • Computing, storing, and retrieving values to and from an N-Dimensional matrix

    - by Adam S
    This question is probably quite different from what you are used to reading here - I hope it can provide a fun challenge. Essentially I have an algorithm that uses 5(or more) variables to compute a single value, called outcome. Now I have to implement this algorithm on an embedded device which has no memory limitations, but has very harsh processing constraints. Because of this, I would like to run a calculation engine which computes outcome for, say, 20 different values of each variable and stores this information in a file. You may think of this as a 5(or more)-dimensional matrix or 5(or more)-dimensional array, each dimension being 20 entries long. In any modern language, filling this array is as simple as having 5(or more) nested for loops. The tricky part is that I need to dump these values into a file that can then be placed onto the embedded device so that the device can use it as a lookup table. The questions now, are: What format(s) might be acceptable for storing the data? What programs (MATLAB, C#, etc) might be best suited to compute the data? C# must be used to import the data on the device - is this possible given your answer to #1?

    Read the article

  • find consecutive nonzero values

    - by thymeandspace
    I am trying to write a simple MATLAB program that will find the first chain (more than 70) of consecutive nonzero values and return the starting value of that consecutive chain. I am working with movement data from a joystick and there are a few thousand rows of data with a mix of zeros and nonzero values before the actual trial begins (coming from subjects slightly moving the joystick before the trial actually started). I need to get rid of these rows before I can start analyzing the movement from the trials. I am sure this is a relatively simple thing to do so I was hoping someone could offer insight. Thank you in advance -Lilly EDIT: Here's what I tried: s = zeros(size(x1)); for i=2:length(x1) if(x1(i-1) ~= 0) s(i) = 1 + s(i-1); end end display(S); for a vector x1 which has a max chain of 72 but I dont know how to find the max chain and return its first value, so I know where to trim. I also really don't think this is the best strategy, since the max chain in my data will be tens of thousands of values. Thanks for helping me edit, Steve. :)

    Read the article

  • How could I eliminate the meter names with an 'x' after them?

    - by Rose Comete
    Hi I imported some Excel data into MatLab - it is a list of about 200 meter names with about 28 rows each, but the problem is that for each there is a duplicate for the other direction - with the same meter name with an 'x' after it. Does anyone have any ideas as to how I can eliminate these ones with an 'x' after wards? Attached is the part of my code that imports the data, but unfortunately I have not got enough points on this site yet to be allowed to upload data/photos x clear all fid=fopen('sue1.csv'); % Open the file sue1.csv and read it all and put it into an array data = textscan(fid,'%s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s %s','Delimiter',',','CollectOutput',1); fclose(fid) j = 1; k = 1; % j - turbine number, k - date number for i = 1:length(data{1,1}) % Run through all the data if strcmp(data{1,1}(i),'') == 0 meterold{j}(k,:) = data{1,1}(i,:); % if strcmp(data{1,1}(i),'MeterName') == 0 % nummeter{j}(k,:) = str2num(data{1,1}(i,3:end)); % end k = k + 1; else % These commands are followed in the strings match (empty line) k = 1; % Reset the day counter as we're back to the beginning j = j + 1; % Add one to the meter counter as we're now looking at % a new turbine end end

    Read the article

  • Parallelize or vectorize all-against-all operation on a large number of matrices?

    - by reve_etrange
    I have approximately 5,000 matrices with the same number of rows and varying numbers of columns (20 x ~200). Each of these matrices must be compared against every other in a dynamic programming algorithm. In this question, I asked how to perform the comparison quickly and was given an excellent answer involving a 2D convolution. Serially, iteratively applying that method, like so list = who('data_matrix_prefix*') H = cell(numel(list),numel(list)); for i=1:numel(list) for j=1:numel(list) if i ~= j eval([ 'H{i,j} = compare(' char(list(i)) ',' char(list(j)) ');']); end end end is fast for small subsets of the data (e.g. for 9 matrices, 9*9 - 9 = 72 calls are made in ~1 s). However, operating on all the data requires almost 25 million calls. I have also tried using deal() to make a cell array composed entirely of the next element in data, so I could use cellfun() in a single loop: # who(), load() and struct2cell() calls place k data matrices in a 1D cell array called data. nextData = cell(k,1); for i=1:k [nextData{:}] = deal(data{i}); H{:,i} = cellfun(@compare,data,nextData,'UniformOutput',false); end Unfortunately, this is not really any faster, because all the time is in compare(). Both of these code examples seem ill-suited for parallelization. I'm having trouble figuring out how to make my variables sliced. compare() is totally vectorized; it uses matrix multiplication and conv2() exclusively (I am under the impression that all of these operations, including the cellfun(), should be multithreaded in MATLAB?). Does anyone see a (explicitly) parallelized solution or better vectorization of the problem?

    Read the article

  • Filter design for audio signal.

    - by beanyblue
    What I am trying to do is simple. I have a few .wav files. I want to remove noise and filter out specific frequencies. I don't have matlab and I intend to write my own code for all the filters. Right now, I have a way to read the .wav file and dump out the structure into a text file. My questions are the following: Can I directly apply the digital filters on this sampled data?{ ie, can I directly do a convolution between my input samples and h(n) for the filter function that i choose?). How do I choose the number of coefficients for the Window function? I have octave, so if someone can point me to anything that gives me some idea on how to process the .wav file using octave, that would be great too. I want to be able to filter out the frequency and then listen to the sound again. Is this possible with octave? I'm just a beginner with these kinds of things, so please bear with me if my questions are too naive. Any help will be great.

    Read the article

  • Have a trouble with the function roots

    - by user3707462
    Hey guys I have multiple problems with using function 'roots'. I Have to find zeros of 's^1000 + 1'. I made Y = zeros(1,1000) then manually changed the 1000th matrice to '1'. but then 'root' function does not work with it ! Another problem is that I am having trouble with matrix multiplication. The question is finding zeros(roots) of (s^6 + 6*s^5 + 15*s^4 + 20*s^3 + 15*s^2 + 6*s +1)*(s^6 + 6s^5 + 15*s^4 +15*s^2 +6*s +1) so i did: a = [1 6 15 20 15 6 1] b = [1 6 15 0 15 6 1] y = a.*b; roots(y) but this gives me -27.9355 + 0.0000i -8.2158 + 0.0000i 0.1544 + 0.9880i 0.1544 - 0.9880i -0.1217 + 0.0000i -0.0358 + 0.0000i where I calculate the original equation with wolfram then I have made matrix as : p = [1 12 66 200 375 492 524 492 375 200 66 12 1] roots(p) and this gives me : -3.1629 + 2.5046i -3.1629 - 2.5046i 0.3572 + 0.9340i 0.3572 - 0.9340i -1.0051 + 0.0000i -1.0025 + 0.0044i -1.0025 - 0.0044i -0.9975 + 0.0044i -0.9975 - 0.0044i -0.9949 + 0.0000i -0.1943 + 0.1539i -0.1943 - 0.1539i and I think the second solution is right (that is what wolfram alpha gave me) How would you answer these two questions through matlab guys?

    Read the article

  • Rotation Matrix calculates by column not by row

    - by pinnacler
    I have a class called forest and a property called fixedPositions that stores 100 points (x,y) and they are stored 250x2 (rows x columns) in MatLab. When I select 'fixedPositions', I can click scatter and it will plot the points. Now, I want to rotate the plotted points and I have a rotation matrix that will allow me to do that. The below code should work: theta = obj.heading * pi/180; apparent = [cos(theta) -sin(theta) ; sin(theta) cos(theta)] * obj.fixedPositions; But it wont. I get this error. ??? Error using == mtimes Inner matrix dimensions must agree. Error in == landmarkslandmarks.get.apparentPositions at 22 apparent = [cos(theta) -sin(theta) ; sin(theta) cos(theta)] * obj.fixedPositions; When I alter forest.fixedPositions to store the variables 2x250 instead of 250x2, the above code will work, but it wont plot. I'm going to be plotting fixedPositions constantly in a simulation, so I'd prefer to leave it as it, and make the rotation work instead. Any ideas? Also, fixed positions, is the position of the xy points as if you were looking straight ahead. i.e. heading = 0. heading is set to 45, meaning I want to rotate points clockwise 45 degrees. Here is my code: classdef landmarks properties fixedPositions %# positions in a fixed coordinate system. [x, y] heading = 45; %# direction in which the robot is facing end properties (Dependent) apparentPositions end methods function obj = landmarks(numberOfTrees) %# randomly generates numberOfTrees amount of x,y coordinates and set %the array or matrix (not sure which) to fixedPositions obj.fixedPositions = 100 * rand([numberOfTrees,2]) .* sign(rand([numberOfTrees,2]) - 0.5); end function obj = set.apparentPositions(obj,~) theta = obj.heading * pi/180; [cos(theta) -sin(theta) ; sin(theta) cos(theta)] * obj.fixedPositions; end function apparent = get.apparentPositions(obj) %# rotate obj.positions using obj.facing to generate the output theta = obj.heading * pi/180; apparent = [cos(theta) -sin(theta) ; sin(theta) cos(theta)] * obj.fixedPositions; end end end P.S. If you change one line to this: obj.fixedPositions = 100 * rand([2,numberOfTrees]) .* sign(rand([2,numberOfTrees]) - 0.5); Everything will work fine... it just wont plot.

    Read the article

  • Why is this class re-initialized every time?

    - by pinnacler
    I have 4 files and the code 'works' as expected. I try to clean everything up, place code into functions, etc... and everything looks fine... and it doesn't work. Can somebody please explain why MatLab is so quirky... or am I just stupid? Normally, I type terminator = simulation(100,20,0,0,0,1); terminator.animate(); and it should produce a map of trees with the terminator walking around in a forest. Everything rotates to his perspective. When I break it into functions... everything ceases to work. I really only changed a few lines of code, shown in comments. Code that works: classdef simulation properties landmarks robot end methods function obj = simulation(mapSize, trees, x,y,heading,velocity) obj.landmarks = landmarks(mapSize, trees); obj.robot = robot(x,y,heading,velocity); end function animate(obj) %Setup Plots fig=figure; xlabel('meters'), ylabel('meters') set(fig, 'name', 'Phil''s AWESOME 80''s Robot Simulator') xymax = obj.landmarks.mapSize*3; xymin = -(obj.landmarks.mapSize*3); l=scatter([0],[0],'bo'); axis([xymin xymax xymin xymax]); obj.landmarks.apparentPositions %Simulation Loop THIS WAS ORGANIZED for n = 1:720, %Calculate and Set Heading/Location obj.robot.headingChange = navigate(n); %Update Position obj.robot.heading = obj.robot.heading + obj.robot.headingChange; obj.landmarks.heading = obj.robot.heading; y = cosd(obj.robot.heading); x = sind(obj.robot.heading); obj.robot.x = obj.robot.x + (x*obj.robot.velocity); obj.robot.y = obj.robot.y + (y*obj.robot.velocity); obj.landmarks.x = obj.robot.x; obj.landmarks.y = obj.robot.y; %Animate set(l,'XData',obj.landmarks.apparentPositions(:,1),'YData',obj.landmarks.apparentPositions(:,2)); rectangle('Position',[-2,-2,4,4]); drawnow end end end end ----------- classdef landmarks properties fixedPositions %# positions in a fixed coordinate system. [ x, y ] mapSize = 10; %Map Size. Value is side of square x=0; y=0; heading=0; headingChange=0; end properties (Dependent) apparentPositions end methods function obj = landmarks(mapSize, numberOfTrees) obj.mapSize = mapSize; obj.fixedPositions = obj.mapSize * rand([numberOfTrees, 2]) .* sign(rand([numberOfTrees, 2]) - 0.5); end function apparent = get.apparentPositions(obj) %-STILL ROTATES AROUND ORIGINAL ORIGIN currentPosition = [obj.x ; obj.y]; apparent = bsxfun(@minus,(obj.fixedPositions)',currentPosition)'; apparent = ([cosd(obj.heading) -sind(obj.heading) ; sind(obj.heading) cosd(obj.heading)] * (apparent)')'; end end end ---------- classdef robot properties x y heading velocity headingChange end methods function obj = robot(x,y,heading,velocity) obj.x = x; obj.y = y; obj.heading = heading; obj.velocity = velocity; end end end ---------- function headingChange = navigate(n) %steeringChange = 5 * rand(1) * sign(rand(1) - 0.5); Most chaotic shit %Draw an S if n <270 headingChange=1; elseif n<540 headingChange=-1; elseif n<720 headingChange=1; else headingChange=1; end end Code that does not work... classdef simulation properties landmarks robot end methods function obj = simulation(mapSize, trees, x,y,heading,velocity) obj.landmarks = landmarks(mapSize, trees); obj.robot = robot(x,y,heading,velocity); end function animate(obj) %Setup Plots fig=figure; xlabel('meters'), ylabel('meters') set(fig, 'name', 'Phil''s AWESOME 80''s Robot Simulator') xymax = obj.landmarks.mapSize*3; xymin = -(obj.landmarks.mapSize*3); l=scatter([0],[0],'bo'); axis([xymin xymax xymin xymax]); obj.landmarks.apparentPositions %Simulation Loop for n = 1:720, %Calculate and Set Heading/Location %Update Position headingChange = navigate(n); obj.robot.updatePosition(headingChange); obj.landmarks.updatePerspective(obj.robot.heading, obj.robot.x, obj.robot.y); %Animate set(l,'XData',obj.landmarks.apparentPositions(:,1),'YData',obj.landmarks.apparentPositions(:,2)); rectangle('Position',[-2,-2,4,4]); drawnow end end end end ----------------- classdef landmarks properties fixedPositions; %# positions in a fixed coordinate system. [ x, y ] mapSize; %Map Size. Value is side of square x; y; heading; headingChange; end properties (Dependent) apparentPositions end methods function obj = createLandmarks(mapSize, numberOfTrees) obj.mapSize = mapSize; obj.fixedPositions = obj.mapSize * rand([numberOfTrees, 2]) .* sign(rand([numberOfTrees, 2]) - 0.5); end function apparent = get.apparentPositions(obj) %-STILL ROTATES AROUND ORIGINAL ORIGIN currentPosition = [obj.x ; obj.y]; apparent = bsxfun(@minus,(obj.fixedPositions)',currentPosition)'; apparent = ([cosd(obj.heading) -sind(obj.heading) ; sind(obj.heading) cosd(obj.heading)] * (apparent)')'; end function updatePerspective(obj,tempHeading,tempX,tempY) obj.heading = tempHeading; obj.x = tempX; obj.y = tempY; end end end ----------------- classdef robot properties x y heading velocity end methods function obj = robot(x,y,heading,velocity) obj.x = x; obj.y = y; obj.heading = heading; obj.velocity = velocity; end function updatePosition(obj,headingChange) obj.heading = obj.heading + headingChange; tempy = cosd(obj.heading); tempx = sind(obj.heading); obj.x = obj.x + (tempx*obj.velocity); obj.y = obj.y + (tempy*obj.velocity); end end end The navigate function is the same... I would appreciate any help as to why things aren't working. All I did was take the code from the first section from under comment: %Simulation Loop THIS WAS ORGANIZED and break it into 2 functions. One in robot and one in landmarks. Is a new instance created every time because it's constantly printing the same heading for this line int he robot class obj.heading = obj.heading + headingChange;

    Read the article

  • Problems with real-valued input deep belief networks (of RBMs)

    - by Junier
    I am trying to recreate the results reported in Reducing the dimensionality of data with neural networks of autoencoding the olivetti face dataset with an adapted version of the MNIST digits matlab code, but am having some difficulty. It seems that no matter how much tweaking I do on the number of epochs, rates, or momentum the stacked RBMs are entering the fine-tuning stage with a large amount of error and consequently fail to improve much at the fine-tuning stage. I am also experiencing a similar problem on another real-valued dataset. For the first layer I am using a RBM with a smaller learning rate (as described in the paper) and with negdata = poshidstates*vishid' + repmat(visbiases,numcases,1); I'm fairly confident I am following the instructions found in the supporting material but I cannot achieve the correct errors. Is there something I am missing? See the code I'm using for real-valued visible unit RBMs below, and for the whole deep training. The rest of the code can be found here. rbmvislinear.m: epsilonw = 0.001; % Learning rate for weights epsilonvb = 0.001; % Learning rate for biases of visible units epsilonhb = 0.001; % Learning rate for biases of hidden units weightcost = 0.0002; initialmomentum = 0.5; finalmomentum = 0.9; [numcases numdims numbatches]=size(batchdata); if restart ==1, restart=0; epoch=1; % Initializing symmetric weights and biases. vishid = 0.1*randn(numdims, numhid); hidbiases = zeros(1,numhid); visbiases = zeros(1,numdims); poshidprobs = zeros(numcases,numhid); neghidprobs = zeros(numcases,numhid); posprods = zeros(numdims,numhid); negprods = zeros(numdims,numhid); vishidinc = zeros(numdims,numhid); hidbiasinc = zeros(1,numhid); visbiasinc = zeros(1,numdims); sigmainc = zeros(1,numhid); batchposhidprobs=zeros(numcases,numhid,numbatches); end for epoch = epoch:maxepoch, fprintf(1,'epoch %d\r',epoch); errsum=0; for batch = 1:numbatches, if (mod(batch,100)==0) fprintf(1,' %d ',batch); end %%%%%%%%% START POSITIVE PHASE %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% data = batchdata(:,:,batch); poshidprobs = 1./(1 + exp(-data*vishid - repmat(hidbiases,numcases,1))); batchposhidprobs(:,:,batch)=poshidprobs; posprods = data' * poshidprobs; poshidact = sum(poshidprobs); posvisact = sum(data); %%%%%%%%% END OF POSITIVE PHASE %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% poshidstates = poshidprobs > rand(numcases,numhid); %%%%%%%%% START NEGATIVE PHASE %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% negdata = poshidstates*vishid' + repmat(visbiases,numcases,1);% + randn(numcases,numdims) if not using mean neghidprobs = 1./(1 + exp(-negdata*vishid - repmat(hidbiases,numcases,1))); negprods = negdata'*neghidprobs; neghidact = sum(neghidprobs); negvisact = sum(negdata); %%%%%%%%% END OF NEGATIVE PHASE %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% err= sum(sum( (data-negdata).^2 )); errsum = err + errsum; if epoch>5, momentum=finalmomentum; else momentum=initialmomentum; end; %%%%%%%%% UPDATE WEIGHTS AND BIASES %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% vishidinc = momentum*vishidinc + ... epsilonw*( (posprods-negprods)/numcases - weightcost*vishid); visbiasinc = momentum*visbiasinc + (epsilonvb/numcases)*(posvisact-negvisact); hidbiasinc = momentum*hidbiasinc + (epsilonhb/numcases)*(poshidact-neghidact); vishid = vishid + vishidinc; visbiases = visbiases + visbiasinc; hidbiases = hidbiases + hidbiasinc; %%%%%%%%%%%%%%%% END OF UPDATES %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% end fprintf(1, '\nepoch %4i error %f \n', epoch, errsum); end dofacedeepauto.m: clear all close all maxepoch=200; %In the Science paper we use maxepoch=50, but it works just fine. numhid=2000; numpen=1000; numpen2=500; numopen=30; fprintf(1,'Pretraining a deep autoencoder. \n'); fprintf(1,'The Science paper used 50 epochs. This uses %3i \n', maxepoch); load fdata %makeFaceData; [numcases numdims numbatches]=size(batchdata); fprintf(1,'Pretraining Layer 1 with RBM: %d-%d \n',numdims,numhid); restart=1; rbmvislinear; hidrecbiases=hidbiases; save mnistvh vishid hidrecbiases visbiases; maxepoch=50; fprintf(1,'\nPretraining Layer 2 with RBM: %d-%d \n',numhid,numpen); batchdata=batchposhidprobs; numhid=numpen; restart=1; rbm; hidpen=vishid; penrecbiases=hidbiases; hidgenbiases=visbiases; save mnisthp hidpen penrecbiases hidgenbiases; fprintf(1,'\nPretraining Layer 3 with RBM: %d-%d \n',numpen,numpen2); batchdata=batchposhidprobs; numhid=numpen2; restart=1; rbm; hidpen2=vishid; penrecbiases2=hidbiases; hidgenbiases2=visbiases; save mnisthp2 hidpen2 penrecbiases2 hidgenbiases2; fprintf(1,'\nPretraining Layer 4 with RBM: %d-%d \n',numpen2,numopen); batchdata=batchposhidprobs; numhid=numopen; restart=1; rbmhidlinear; hidtop=vishid; toprecbiases=hidbiases; topgenbiases=visbiases; save mnistpo hidtop toprecbiases topgenbiases; backpropface; Thanks for your time

    Read the article

  • Problems with real-valued deep belief networks (of RBMs)

    - by Junier
    I am trying to recreate the results reported in Reducing the dimensionality of data with neural networks of autoencoding the olivetti face dataset with an adapted version of the MNIST digits matlab code, but am having some difficulty. It seems that no matter how much tweaking I do on the number of epochs, rates, or momentum the stacked RBMs are entering the fine-tuning stage with a large amount of error and consequently fail to improve much at the fine-tuning stage. I am also experiencing a similar problem on another real-valued dataset. For the first layer I am using a RBM with a smaller learning rate (as described in the paper) and with negdata = poshidstates*vishid' + repmat(visbiases,numcases,1); I'm fairly confident I am following the instructions found in the supporting material but I cannot achieve the correct errors. Is there something I am missing? See the code I'm using for real-valued visible unit RBMs below, and for the whole deep training. The rest of the code can be found here. rbmvislinear.m: epsilonw = 0.001; % Learning rate for weights epsilonvb = 0.001; % Learning rate for biases of visible units epsilonhb = 0.001; % Learning rate for biases of hidden units weightcost = 0.0002; initialmomentum = 0.5; finalmomentum = 0.9; [numcases numdims numbatches]=size(batchdata); if restart ==1, restart=0; epoch=1; % Initializing symmetric weights and biases. vishid = 0.1*randn(numdims, numhid); hidbiases = zeros(1,numhid); visbiases = zeros(1,numdims); poshidprobs = zeros(numcases,numhid); neghidprobs = zeros(numcases,numhid); posprods = zeros(numdims,numhid); negprods = zeros(numdims,numhid); vishidinc = zeros(numdims,numhid); hidbiasinc = zeros(1,numhid); visbiasinc = zeros(1,numdims); sigmainc = zeros(1,numhid); batchposhidprobs=zeros(numcases,numhid,numbatches); end for epoch = epoch:maxepoch, fprintf(1,'epoch %d\r',epoch); errsum=0; for batch = 1:numbatches, if (mod(batch,100)==0) fprintf(1,' %d ',batch); end %%%%%%%%% START POSITIVE PHASE %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% data = batchdata(:,:,batch); poshidprobs = 1./(1 + exp(-data*vishid - repmat(hidbiases,numcases,1))); batchposhidprobs(:,:,batch)=poshidprobs; posprods = data' * poshidprobs; poshidact = sum(poshidprobs); posvisact = sum(data); %%%%%%%%% END OF POSITIVE PHASE %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% poshidstates = poshidprobs > rand(numcases,numhid); %%%%%%%%% START NEGATIVE PHASE %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% negdata = poshidstates*vishid' + repmat(visbiases,numcases,1);% + randn(numcases,numdims) if not using mean neghidprobs = 1./(1 + exp(-negdata*vishid - repmat(hidbiases,numcases,1))); negprods = negdata'*neghidprobs; neghidact = sum(neghidprobs); negvisact = sum(negdata); %%%%%%%%% END OF NEGATIVE PHASE %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% err= sum(sum( (data-negdata).^2 )); errsum = err + errsum; if epoch>5, momentum=finalmomentum; else momentum=initialmomentum; end; %%%%%%%%% UPDATE WEIGHTS AND BIASES %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% vishidinc = momentum*vishidinc + ... epsilonw*( (posprods-negprods)/numcases - weightcost*vishid); visbiasinc = momentum*visbiasinc + (epsilonvb/numcases)*(posvisact-negvisact); hidbiasinc = momentum*hidbiasinc + (epsilonhb/numcases)*(poshidact-neghidact); vishid = vishid + vishidinc; visbiases = visbiases + visbiasinc; hidbiases = hidbiases + hidbiasinc; %%%%%%%%%%%%%%%% END OF UPDATES %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% end fprintf(1, '\nepoch %4i error %f \n', epoch, errsum); end dofacedeepauto.m: clear all close all maxepoch=200; %In the Science paper we use maxepoch=50, but it works just fine. numhid=2000; numpen=1000; numpen2=500; numopen=30; fprintf(1,'Pretraining a deep autoencoder. \n'); fprintf(1,'The Science paper used 50 epochs. This uses %3i \n', maxepoch); load fdata %makeFaceData; [numcases numdims numbatches]=size(batchdata); fprintf(1,'Pretraining Layer 1 with RBM: %d-%d \n',numdims,numhid); restart=1; rbmvislinear; hidrecbiases=hidbiases; save mnistvh vishid hidrecbiases visbiases; maxepoch=50; fprintf(1,'\nPretraining Layer 2 with RBM: %d-%d \n',numhid,numpen); batchdata=batchposhidprobs; numhid=numpen; restart=1; rbm; hidpen=vishid; penrecbiases=hidbiases; hidgenbiases=visbiases; save mnisthp hidpen penrecbiases hidgenbiases; fprintf(1,'\nPretraining Layer 3 with RBM: %d-%d \n',numpen,numpen2); batchdata=batchposhidprobs; numhid=numpen2; restart=1; rbm; hidpen2=vishid; penrecbiases2=hidbiases; hidgenbiases2=visbiases; save mnisthp2 hidpen2 penrecbiases2 hidgenbiases2; fprintf(1,'\nPretraining Layer 4 with RBM: %d-%d \n',numpen2,numopen); batchdata=batchposhidprobs; numhid=numopen; restart=1; rbmhidlinear; hidtop=vishid; toprecbiases=hidbiases; topgenbiases=visbiases; save mnistpo hidtop toprecbiases topgenbiases; backpropface; Thanks for your time

    Read the article

  • How would i down-sample a .wav file then reconstruct it using nyquist? - in matlab [closed]

    - by martin
    This is all done in MatLab 2010 My objective is to show the results of: undersampling, nyquist rate/ oversampling First i need to downsample the .wav file to get an incomplete/ or impartial data stream that i can then reconstuct. Heres the flow chart of what im going to be doing So the flow is analog signal - sampling analog filter - ADC - resample down - resample up - DAC - reconstruction analog filter what needs to be achieved: F= Frequency F(Hz=1/s) E.x. 100Hz = 1000 (Cyc/sec) F(s)= 1/(2f) Example problem: 1000 hz = Highest frequency 1/2(1000hz) = 1/2000 = 5x10(-3) sec/cyc or a sampling rate of 5ms This is my first signal processing project using matlab. what i have so far. % Fs = frequency sampled (44100hz or the sampling frequency of a cd) [test,fs]=wavread('test.wav'); % loads the .wav file left=test(:,1); % Plot of the .wav signal time vs. strength time=(1/44100)*length(left); t=linspace(0,time,length(left)); plot(t,left) xlabel('time (sec)'); ylabel('relative signal strength') **%this is were i would need to sample it at the different frequecys (both above and below and at) nyquist frequency.*I think.*** soundsc(left,fs) % shows the resaultant audio file , which is the same as original ( only at or above nyquist frequency however) Can anyone tell me how to make it better, and how to do the various sampling at different frequencies?

    Read the article

  • MATLAB: What is an appropriate Data Structure for a Matrix with Random Variable Entries?

    - by user12707
    I'm working in an area that is related to simulation and trying to design a data structure that can include random variables within matrices. I am currently coding in MATLAB. To motivate this let me say I have the following matrix: [a b; c d] I want to find a data structure that will allow for a, b, c, d to be either real numbers or random variables. As an example, let's say that a = 1, b = -1, c = 2 but let d be a normally distributed random variable with mean 20 and SD 40. The data structure that I have in mind will give no value to d. However, I also want to be able to design a function that can take in the structure, simulate an uniform(0,1), obtain a value for d using an inverse CDF and then spit out an actual matrix. I have several ideas to do this (all related to the MATLAB icdf function) but would like to know how more experienced programmers would do it. In this application, it's important that the structure is as "lean" as possible since I will be working with very very large matrices and memory will be an issue.

    Read the article

  • How can I flush the output of disp in Octave?

    - by Nathan Fellman
    I have a program in Octave that has a loop - running a function with various parameters, not something that I can turn into matrices. At the beginning of each iteration I print the current parameters using disp. The first times I ran it I had a brazillion warnings, and then I also got these prints. Now that I cleaned them up, I no longer see them. My guess is that they're stuck in a buffer, and I'll see them when the program ends or the buffer fills. Is there any way to force a flush of the print buffer so that I can see my prints?

    Read the article

  • Simulink: type consistency errors

    - by stanigator
    Using this Simulink model file as a reference, I'm trying to figure out the two following errors: I have no idea what has gone wrong with the data type consistency/conversion problems. Do you know what the error messages mean exactly in the context of a model? It would be great to get an interpretation of the problem to solve it. Thanks in advance.

    Read the article

  • the easiest way to convert matrix to one row vector

    - by niko
    Hi, Does anyone know what is the best way to create one row matrix (vector) from M x N matrix by putting all rows, from 1 to M, of the original matrix into first row of new matrix the following way: A = [row1; row2, ..., rowM] B = [row1, row2, ..., rowM] Example: A = [1 1 0 0; 0 1 0 1] B = [1 1 0 0 0 1 0 1] I would be very thankful if anyone suggested any simple method or perhaps points out a function if it already exists that could generate matrix B from original matrix A.

    Read the article

  • uncomplete ezplot drawing

    - by Alireza
    why when i use ezplot in for example [1 1.5] interval, a discontinuity will appear in some pieces of lines but when i use a closer interval like [1.3 1.5], the discontinuity will be annihilated?

    Read the article

  • Saving/Associating slider values with a pop-up menu

    - by James
    Hi, Following on from a question I posted yesterday about GUIs, I have another problem I've been working with. This question related to calculating the bending moment on a beam under different loading conditions. On the GUI I have developed so far, I have a number of sliders (which now work properly) and a pop-up menu which defines the load case. I would like to be able to select the load case from the pop-up menu and position the loads as appropriate, in order to define each load case in turn. The output that I need is an array defining the load case number (the rows) and a number of loading parameters (the itensity and position of the loads, which are controlled by the sliders). The problem I am having is that I can produce this array (of the size I need) and define the loading for one load case (by selecting the pop-up menu) using the sliders, but when I change the popup menu again, the array only keeps the loading for the load case selected by the pop-up menu. Can anyone suggest an approach I can take with (specifically to store the variables from each load case) or an example that illustrates a similar solution to the problem? The probem may be a bit vague, so please let me know if anything needs clearing up. Many Thanks, James

    Read the article

< Previous Page | 27 28 29 30 31 32 33 34 35 36 37 38  | Next Page >