Plot Heatmap from distinct coordinates and data array of floats

Hi,

I have an array of floats of size 200x2 where each row is the (x,y) coordinate of a specific point on the heatmap (x=column 1 and y=column 2). I also have an array of floats of size 1x200, which is the data I want to plot, I'll denote it as z. At each row of the coordinates array, the point (x,y) is associated with the value in z at the same index. I want to create a heatmap using this data.

I tried creating a meashgrid from the two columns of the coordinate array, but then I have this problem that z doesn't have the proper dimantions. I believe I need to create an array filed with zeros, and fill it with z values at the adequate coordinates, but than I have the problem that the coordinates are floats.

This is the code I used (which is of course problematic):

coords = load('grid_points.mat').array;                 
accuracies = load('test_accuracies.mat').array;

xcoord = coords(:,1);
ycoord = coords(:,2);

[X,Y] = meshgrid(xcoord,ycoord);

x = X(:);                                       % Your varaible x should something like this as a column vector
y = Y(:);                                       % Similar for variable y
z = accuracies(:);                              % Similar for variable z

tbl = table(x,y,z);                             % Combine them in a table
h = heatmap(tbl,'x','y','ColorVariable','z');   % Generate heat map

patial view of the coordinates array:

patial view of the data array:

I'd appreciate it if anyone could help me understand how can I overcome the fact that the coordinates array has floats in it (I don't want to convert it into integers, I want to keep it as floats if possible).

Thank you very much for your time and attention

Answer 1

[KSSV]

It depends on your data whether it is structured or unstructured. Let x, y, z be your column data vectors.

%%structured 
xi = unique(x) ; yi = unique(y) ;
[X,Y] = meshgrid(xi,yi) ;
Z = reshape(z,size(X)) ;
figure
surf(X,Y,Z)
%%unstructured 
dt = delaunayTriangulation(x,y) ;
tri = dt.ConnectivityList ;
figure
trisurf(tri,x,y,z)

Also have a look on griddata and scatteredInterpolant

[Elinor Ginzburg]

Thank you for your comment. I don't have a lot of experience with matlab, so forgive me for my questions - what does it mean if my data is structured or unstuctured?

in case it is structured - I'm not able to reshape z, since size(X) is 200x200 after performing meshgrid but z has only 200 elements (double numbers)

in case it is unstructured - my x and y variables are of class 'single' but I'm not able to cast them to double matrix as the function delaunayTriangulation requires.

Perhaps you have an idea what can I do?

[KSSV]

You can convert single to double using the function double. You may attach your data, so that I can look into and help you.

[Elinor Ginzburg]

Thank you very much

Yeah, I used double() but it's not a matrix apperantly so the function throws me an error. I attached my data.

Thanks again!

[KSSV]

load test_accuracies.mat
z = array' ;
load grid_points.mat ;
x = double(array(:,1)) ;
y = double(array(:,2)) ;

pgon = polyshape(x,y) ;
xp = pgon.Vertices(:,1) ;
yp = pgon.Vertices(:,2) ;
F = scatteredInterpolant(x,y,z) ;
zp = F(xp,yp) ;

x1 = x(1:100) ; y1 = y(1:100) ;
x2 = x(101:200) ; y2 = y(101:200) ;


xi = linspace(min(x),max(x),500) ;
yi = linspace(min(y),max(y),500) ;
[X,Y] = meshgrid(xi,yi) ;

idx = inpolygon(X,Y,xp,yp) ;
id = ~isnan(zp) ;
Z = griddata(xp(id),yp(id),zp(id),X,Y) ;

Z(~idx) = NaN ;

pcolor(X,Y,Z)
shading interp

[Elinor Ginzburg]

I'm sorry for the late reply. Thank you so much for your help!