Heat Map Representation of error data

I have an error matrix consisting of positive and negative difference between the theoritical and practical values. I want to select the cell containing lowest positive number using heat map and ignore the negative numbers. How to do that ?

P.S. : This is a part of bigger program where difference in RMSE values of training and validation datasets represents the above error matrix.

Answer 1

[Mathieu NOE]

hello

I could suggest these two solutions

either based on the regular heatmap function and creating a specific colormap

the limited range that includes your min positive value is coded in red (up to you to change to another color)

or second solution is based on a Fex submission (as far as I can remember, tx to his creator !)

err = readmatrix("Error Matrix.xlsx");
err_pos = err;
err_pos(err_pos<0) = NaN;
[min_val,id]= min(err_pos,[],'all','linear'); % min positive value  and linear indice
[r,c] = ind2sub(size(err),id); % convert linear indice to row / col indices

%% solution #1 
figure(1)
max_val = max(abs(err),[],'all');

% this colormap will span values from -max_val to +max_val
n = 2*ceil(max_val/min_val)+1; % odd number
cmap = gray(n);
% create a "red" zone between 0 and a value slightly above the min error 
rz = (n-1)/2; % row index correponding to zero
cmap((rz:rz+2),1) = 1; % 
cmap((rz:rz+2),2:3) = 0; % 

% create the heatmap
h = heatmap(err,'GridVisible',false);
% apply the colormap
colormap(cmap)
% % apply the color limits (full range of the data)
caxis([-max_val max_val])

%% solution #2 
figure(2)
matvisual(err, 'annotation')
hold on
plot(c,r,'ro','markersize',60);
hold off

% +------------------------------------------------------+
% |   Matrix Visualization with MATLAB Implementation    | 
% |                                                      |
% | Author: Ph.D. Eng. Hristo Zhivomirov        12/02/18 | 
% +------------------------------------------------------+
% 
% function: matvisual(A, varargin)
%
% Input:
% A - m-by-n-by-p matrix to be visualized
% varargin - type 'annotation' in the place of varargin if one wants to
%            annotate the matrix plot (x-label, y-label, etc.)
% example : 
% %% generate a test matrix
% % A = eye(10);
% % A = magic(30);
% % A = ones(10);
% % A = zeros(10);
% % A = NaN(10);
% % A = Inf(10);
% % A = rand(20, 10, 3);
% A = round(10*rand(50, 50));
% % A = round(10*rand(10, 5, 5));
% % A = abs(fft(1:10));
% % A = complex(rand(5,5), rand(5,5));
% % A = char(1300*ones(10));
% % A = zeros(0, 20);
% 
% %% visualize the matrix
% matvisual(A, 'annotation')


function matvisual(A, varargin)

% input validation
validateattributes(A, {'single', 'double', 'logical'}, ...
                      {'3d', 'nonempty', 'real'}, ...
                      '', 'A', 1)

% determine the matrix size
[M, N, P] = size(A);

% loop through the matrix pages
for p = 1:P
    % get prepared for page-by-page visualization
    if P > 1, subplot(1, P, p), end 
    
    % visualize the matrix
    himg = imagesc(A(:, :, p));
    grid on
      
    % values labeling
    for m = 1:M
        for n = 1:N
            text(n, m, num2str(A(m, n, p), 3), ...
                'FontName', 'Times New Roman', ...
                'FontSize', round(6 + 50./sqrt(M.*N)), ...
                'HorizontalAlignment', 'center', ...
                'Rotation', 45)
        end
    end
    
    % annotation
    if strcmp(varargin, 'annotation')
        % x-label, y-label, x-ticks, y-ticks, title
        set(gca, 'FontName', 'Times New Roman', 'FontSize', 12)
        xlabel('Column number')
        ylabel('Row number')
        if P > 1, title(['Matrix page ' num2str(p)]), end
        if M <= 50, set(gca, 'YTick', 1:M), end
        if N <= 50, set(gca, 'XTick', 1:N), end
    end
    
    % set the datatip UpdateFcn
    cursorMode = datacursormode(gcf);
    set(cursorMode, 'UpdateFcn', {@datatiptxt, himg})
end

end

function text_to_display = datatiptxt(~, hDatatip, himg)

% determine the current datatip position
pos = get(hDatatip, 'Position');

% form the datatip label
text_to_display = {['Row: ' num2str(pos(2))], ...
                   ['Column: ' num2str(pos(1))], ...
                   ['Value: ' num2str(himg.CData(pos(2), pos(1)))]};

end

[Mathieu NOE]

hello again

do you mind accepting my answer (if it has fullfiled your expectations ) ? tx