I want to plot a confusion matrix in MATLAB. Here's my code;
data = rand(3, 3)
imagesc(data)
colormap(gray)
colorbar
When I run this, a confusion matrix with a color bar is shown. But usually, I have seen confusion matrix in MATLAB will give counts as well as probabilities. How can I get them? How can I change the class labels which will be shown as 1,2,3, etc.?
I want a matrix like this:
If you do not have the neural network toolbox, you can use plotConfMat. It gets you the following result.
I have also included an independent example below without the need for the function:
% sample data
confmat = magic(3);
labels = {'Dog', 'Cat', 'Horse'};
numlabels = size(confmat, 1); % number of labels
% calculate the percentage accuracies
confpercent = 100*confmat./repmat(sum(confmat, 1),numlabels,1);
% plotting the colors
imagesc(confpercent);
title('Confusion Matrix');
ylabel('Output Class'); xlabel('Target Class');
% set the colormap
colormap(flipud(gray));
% Create strings from the matrix values and remove spaces
textStrings = num2str([confpercent(:), confmat(:)], '%.1f%%\n%d\n');
textStrings = strtrim(cellstr(textStrings));
% Create x and y coordinates for the strings and plot them
[x,y] = meshgrid(1:numlabels);
hStrings = text(x(:),y(:),textStrings(:), ...
'HorizontalAlignment','center');
% Get the middle value of the color range
midValue = mean(get(gca,'CLim'));
% Choose white or black for the text color of the strings so
% they can be easily seen over the background color
textColors = repmat(confpercent(:) > midValue,1,3);
set(hStrings,{'Color'},num2cell(textColors,2));
% Setting the axis labels
set(gca,'XTick',1:numlabels,...
'XTickLabel',labels,...
'YTick',1:numlabels,...
'YTickLabel',labels,...
'TickLength',[0 0]);
If you have the neural network toolbox you can use the function plotconfusion. You can create a copy and edit it to customise it further, for example to print custom labels.
Related
I have four sets of data, the distribution of which I would like to represent in MATLAB in one figure. Current code is:
[n1,x1]=hist([dataset1{:}]);
[n2,x2]=hist([dataset2{:}]);
[n3,x3]=hist([dataset3{:}]);
[n4,x4]=hist([dataset4{:}]);
bar(x1,n1,'hist');
hold on; h1=bar(x1,n1,'hist'); set(h1,'facecolor','g')
hold on; h2=bar(x2,n2,'hist'); set(h2,'facecolor','g')
hold on; h3=bar(x3,n3,'hist'); set(h3,'facecolor','g')
hold on; h4=bar(x4,n4,'hist'); set(h4,'facecolor','g')
hold off
My issue is that I have different sampling sizes for each group, dataset1 has an n of 69, dataset2 has an n of 23, dataset3 and dataset4 have n's of 10. So how do I normalize the distributions when representing these three groups together?
Is there some way to..for example..divide the instances in each bin by the sampling for that group?
You can normalize your histograms by dividing by the total number of elements:
[n1,x1] = histcounts(randn(69,1));
[n2,x2] = histcounts(randn(23,1));
[n3,x3] = histcounts(randn(10,1));
[n4,x4] = histcounts(randn(10,1));
hold on
bar(x4(1:end-1),n4./sum(n4),'histc');
bar(x3(1:end-1),n3./sum(n3),'histc');
bar(x2(1:end-1),n2./sum(n2),'histc');
bar(x1(1:end-1),n1./sum(n1),'histc');
hold off
ax = gca;
set(ax.Children,{'FaceColor'},mat2cell(lines(4),ones(4,1),3))
set(ax.Children,{'FaceAlpha'},repmat({0.7},4,1))
However, as you can see above, you can do some more things to make your code more simple and short:
You only need to hold on once.
Instead of collecting all the bar handles, use the axes handle.
Plot the bar in ascending order of the number of elements in the dataset, so all histograms will be clearly visible.
With the axes handle set all properties at one command.
and as a side note - it's better to use histcounts.
Here is the result:
EDIT:
If you want to also plot the pdf line from histfit, then you can save it first, and then plot it normalized:
dataset = {randn(69,1),randn(23,1),randn(10,1),randn(10,1)};
fits = zeros(100,2,numel(dataset));
hold on
for k = numel(dataset):-1:1
total = numel(dataset{k}); % for normalizing
f = histfit(dataset{k}); % draw the histogram and fit
% collect the curve data and normalize it:
fits(:,:,k) = [f(2).XData; f(2).YData./total].';
x = f(1).XData; % collect the bar positions
n = f(1).YData; % collect the bar counts
f.delete % delete the histogram and the fit
bar(x,n./total,'histc'); % plot the bar
end
ax = gca; % get the axis handle
% set all color and transparency for the bars:
set(ax.Children,{'FaceColor'},mat2cell(lines(4),ones(4,1),3))
set(ax.Children,{'FaceAlpha'},repmat({0.7},4,1))
% plot all the curves:
plot(squeeze(fits(:,1,:)),squeeze(fits(:,2,:)),'LineWidth',3)
hold off
Again, there are some other improvements you can introduce to your code:
Put everything in a loop to make thigs more easily changed later.
Collect all the curves data to one variable so you can plot them all together very easily.
The new result is:
I'm trying to draw a set of rectangles, each with a fill color representing some value between 0 and 1. Ideally, I would like to use any standard colormap.
Note that the rectangles are not placed in a nice grid, so using imagesc, surf, or similar seems unpractical. Also, the scatter function does not seem to allow me to assign a custom marker shape. Hence, I'm stuck to plotting a bunch of Rectangles in a for-loop and assigning a FillColor by hand.
What's the most efficient way to compute RGB triplets from the scalar values? I've been unable to find a function along the lines of [r,g,b] = val2rgb(value,colormap). Right now, I've built a function which computes 'jet' values, after inspecting rgbplot(jet). This seems a bit silly. I could, of course, obtain values from an arbitrary colormap by interpolation, but this would be slow for large datasets.
So, what would an efficient [r,g,b] = val2rgb(value,colormap) look like?
You have another way to handle it: Draw your rectangles using patch or fill specifying the color scale value, C, as the third parameter. Then you can add and adjust the colorbar:
x = [1,3,3,1,1];
y = [1,1,2,2,1];
figure
for ii = 1:10
patch(x + 4 * rand(1), y + 2 * rand(1), rand(1), 'EdgeColor', 'none')
end
colorbar
With this output:
I think erfan's patch solution is much more elegant and flexible than my rectangle approach.
Anyway, for those who seek to convert scalars to RGB triplets, I'll add my final thoughts on the issue. My approach to the problem was wrong: colors should be drawn from the closest match in the colormap without interpolation. The solution becomes trivial; I've added some code for those who stumble upon this issue in the future.
% generate some data
x = randn(1,1000);
% pick a range of values that should map to full color scale
c_range = [-1 1];
% pick a colormap
colormap('jet');
% get colormap data
cmap = colormap;
% get the number of rows in the colormap
cmap_size = size(cmap,1);
% translate x values to colormap indices
x_index = ceil( (x - c_range(1)) .* cmap_size ./ (c_range(2) - c_range(1)) );
% limit indices to array bounds
x_index = max(x_index,1);
x_index = min(x_index,cmap_size);
% read rgb values from colormap
x_rgb = cmap(x_index,:);
% plot rgb breakdown of x values; this should fall onto rgbplot(colormap)
hold on;
plot(x,x_rgb(:,1),'ro');
plot(x,x_rgb(:,2),'go');
plot(x,x_rgb(:,3),'bo');
axis([c_range 0 1]);
xlabel('Value');
ylabel('RGB component');
With the following result:
I have a d1×d2 array A of integer values and want to color code display them on the xy-plane in a rectangle scaled to d1×d2 using colors that match the magnitude of the value of array at that location. I also want to display the color chart which indicates which color is which magnitude as in the following figure:-
Is there a simple code that can do this?
Or is this kind of charting need special packages?
Will this work ('A' is matrix with non-negative entries)?
function plot2Ddistprf(A, Length, Breadth)
Amax=max(A(:));
A=A/Amax;
G1 = linspace(0,Length,size(A,1));
G2 = linspace(0,Breadth,size(A,2));
[X,Y] = meshgrid(G1,G2);
% plot data
figure; % create a new figure
contourf(X,Y,A); % plot data as surface
caxis([1,100]); % set limits of colormap
colorbar; % display the colorbar
No external library or special package is needed to create such a plot. You can use contourf to plot the data. Then, set the colormap to gray. With caxis you can control the range of colors. colorbar shows the bar on the right side.
The result looks like this:
Here is the code:
% generate sample data
d1 = linspace(-3,3,200);
d2 = linspace(-3,3,200);
[X,Y] = meshgrid(d1,d2);
A = -abs(peaks(X,Y))+100;
% plot data
figure; % create a new figure
contourf(X,Y,A); % plot data as surface
colormap(gray); % use gray colormap
caxis([91,100]); % set limits of colormap
colorbar; % display the colorbar
title('The Title');
xlabel('y');
ylabel('x');
Try the contourf function and then add the colorbar
contourf(A)
colorbar
I would like to plot a 3D surface using the Matlab surf function. The whole surface should be in gray scale, then I need to highlight a specific cut of the surface using a different color.
I thought this code would've worked but it doesn't.
Mat = randi(100); % Matrix to be plotted in gray scale
ind_highlight = 10; % Row of the matrix to be highlighted
Mat2 = Mat;
Mat2([1:ind_highlight-1, ind_highlight+1:end] ,:) = NaN;
figure
surf(X,Y,Mat)
colormap gray
hold on
% Highlight the ind_highlight row
surf(X,Y,Mat2)
colormap hsv
Any help would be highly appreciated!
It seems there is no way to use different colormap to obtain the desired effect since the colormap "belongs" to the figure.
I've found a possible solution which does not use colormap.
It is based on the specifying the color matrix in the call to surf, one for the whole matrix, one for the section to be highlighted, then superimposing the second one to the first one.
Unfortunately, I've not been able to set the first ad gray.
I've used the peaks matrix instead of your "randi" in order to have a more smooth surface to work with and inserted the script in a for loop to highlight different section of the matrix
% Mat = randi(100,100,100); % Matrix to be plotted in gray scale
% Alternative definition of the Matrix to be displayed
n_pt=50;
Mat=peaks(n_pt);
% Generate meshgrid
x=1:n_pt;
y=1:n_pt;
[X,Y]=meshgrid(x,y);
ind_highlight_2 = 5; % Number of rows of the matrix to be highlighted
% Generate two set of color matrix
% The first on for the whole surf
% The second one for the section to be highlighted
a=randi(2,n_pt,n_pt);
b=randi(10,n_pt,n_pt);
for i=1:n_pt-ind_highlight_2
ind_highlight_1 = i; % Starting row of the matrix to be highlighted
Mat2 = Mat;
% Modified set of data (in the original just one row was left
% Mat2([1:ind_highlight-1, ind_highlight+1:end] ,:) = NaN
Mat2(ind_highlight_1:ind_highlight_1+ind_highlight_2,:) = NaN;
COL=a;
COL(ind_highlight_1:ind_highlight_1+ind_highlight_2,:)=b(ind_highlight_1:ind_highlight_1+ind_highlight_2,:);
% Plot the surf specifying the color
s_h=surf(X,Y,Mat,COL);
shading interp
% view([0 90])
% f_name=['jpg_name_' num2str(i)]
% print('-djpeg75',f_name)
pause(.1);
end
Hope this helps.
Hi i am trying to make a simple distribution histogram using some code from stack overflow
however i am unable to get it to work. i know that there are is a simple method for this using statistic toolbox but form a learning point of view i prefer a more explanatory code - can any one help me ?
%%
clear all
load('Mini Project 1.mat')
% Set data to var2
data = var2;
% Set the number of bins
nbins = 0:.8:8;
% Create a histogram plot of data sorted into (nbins) equally spaced bins
n = hist(data,nbins);
% Plot a bar chart with y values at each x value.
% Notice that the length(x) and length(y) have to be same.
bar(nbins,n);
MEAN = mean(data);
STD = sqrt(mean((data - MEAN).^2)); % can also use the simple std(data)
f = ( 1/(STD*sqrt(2*pi)) ) * exp(-0.5*((nbins-MEAN)/STD).^2 );
f = f*sum(nbins)/sum(f);
hold on;
% Plots a 2-D line plot(x,y) with the normal distribution,
% c = color cyan , Width of line = 2
plot (data,f, 'c', 'LineWidth', 2);
xlabel('');
ylabel('Firmness of apples after one month of storage')
title('Histogram compared to normal distribution');
hold of
You are confusing
hist
with
histc
Read up on both.
Also, you are not defining the number of bins, you are defining the bins themselves .
I don't have Matlab at hand now, but try the following:
If you want to compare a normal distribution to the bar plot bar(nbins,n), you should first normalize it:
bar(nbins,n/sum(n))
See if this solves your problem.
If not, try also removing the line f = f*sum(nbins)/sum(f);.