how to set concentric rectangle dataset？

I want that the number of each color data point can be changed

[Adam Danz]

If you show the code that produced the plot, it will be easy to count the number of points within each color.

[Image Analyst]

Since you made this plot, presumably with rand(), you have control over the number of points. Why can't you change it??? It's one of the parameters you gave to rand().

[z cy]

I used this method to get these data. Could you tell how to use rand() to make this plot. Thanks.

r = unifrnd(-0.9,0.9,[1,n1]);%r= r(randperm(length(r)));
t = unifrnd(-0.5,0.5,[1,n1]);%t= t(randperm(length(t)));
x = r + noise1*randn(1,n1);
y = t + noise1*randn(1,n1);
z = sqrt(noise2)*randn(fea_n,n1);
data1 = [x; y;z];

r = unifrnd(-1.9,1.9,[1,n1*2.5]);
t = unifrnd(-1.5,1.5,[1,n1*2.5]);
b = [r;t];
b(:,(abs(b(1,:)) <=1.15)+(abs(b(2,:))<0.65) ==2) =[];
n2 = size(b,2);
x = b(1,:) + noise1*randn(1,n2);
y = b(2,:) + noise1*randn(1,n2);
z = sqrt(noise2)*randn(fea_n,n2);
data2 = [x;y;z];

[Adam Danz]

But that doesn't show how the plot was produced.

[z cy]

oh,sorry. I just use [x;y] of data1 ,data2 and data3(not porvide code,but the way are same with how to get data2). Then plot it.

[Adam Danz]

If the x and y values of data1 are the coordinates of one color and the x and y values of data2 are the coordinates of another color and so one, then why can't you just count the number of (x,y) values?

[z cy]

In my code,just deleted which not satisfazed the range of data. such as

b(:,(abs(b(1,:)) <=1.15)+(abs(b(2,:))<0.65) ==2) =[];

Count the number no useless. I still cannot to adjust the number of each color point.

[Adam Danz]

We cannot run the code you shared due to many missing variables. If you can provide a fully functional, easy to use version of your code I'd be glad to figure out how your data are produced.

[z cy]

n1 = 500 ;
noise1 = 0.02;
r = unifrnd(-0.9,0.9,[1,n1]);
t = unifrnd(-0.5,0.5,[1,n1]);
x = r + noise1*randn(1,n1);
y = t + noise1*randn(1,n1);
data1 = [x;y];

r = unifrnd(-1.9,1.9,[1,n1*2.5]);
t = unifrnd(-1.5,1.5,[1,n1*2.5]);
b = [r;t];
b(:,(abs(b(1,:)) <=1.15)+(abs(b(2,:))<0.65) ==2) =[];
n2 = size(b,2);
x = b(1,:) + noise1*randn(1,n2);
y = b(2,:) + noise1*randn(1,n2);
data2 = [x;y];

r = unifrnd(-2.9,2.9,[1,n1*4]);
t = unifrnd(-2.5,2.5,[1,n1*4]);
b = [r;t];
b(:,(abs(b(1,:)) <=2.15)+(abs(b(2,:))<1.6) ==2) =[];
n3 = size(b,2);
x = b(1,:) + noise1*randn(1,n3);
y = b(2,:) + noise1*randn(1,n3);
data3 = [x;y];

X_train = [data1,data2,data3];
Y_train = [ones(1,n1),2*ones(1,n2),3*ones(1,n3)];

figure
plot(data1(1,:), data1(2,:), 'ro',data2(1,:), data2(2,:), 'g+',data3(1,:), data3(2,:), 'b^','linewidth',1.5);

[Adam Danz]

Instead of deleting some of the points, you could replace those coordinates with a new set of random coordinates. That could be set up within a while-loop where you keep replacing values that meet a set criteria until all of the points meet the requirements.

Answer 1

[Image Analyst]

Try this:

% Define parameters.
numPoints = 1000000; % Something way bigger than you think you'll need.
n1 = 200;
n2 = 300;
n3 = 400;
% Generate numPoints (x,y) locations.
x = 3 * rand(numPoints, 2) - 1.5;
y = 3 * rand(numPoints, 2) - 1.5;
% plot(x, y, '.');
%-----------------------------------------------------------
% Get the inner zone points and plot them.
zone1 = (abs(x) < 0.5) & (abs(y) < 0.5); % Inner zone
zone1i = find(zone1, n1); % Get the actual indexes for the first n1 points.
x1 = x(zone1i);
y1 = y(zone1i);
hold on;
plot(x1, y1, 'ro');

%-----------------------------------------------------------
% Get the outer zone points and plot them.
zone3 = ~((abs(x) < 1) & (abs(y) < 1)); % Outer zone
zone3i = find(zone3, n3); % Get the actual indexes for the first n3 points.
x3 = x(zone3i);
y3 = y(zone3i);
hold on;
plot(x3, y3, 'b^');

%-----------------------------------------------------------
% Get the ring zone points and plot them.
zone2 = ~(zone3 | zone1); % Ring zone is everything except those in zone 1 or zone 3.
zone2i = find(zone2, n2); % Get the actual indexes for the first n2 points.
x2 = x(zone2i);
y2 = y(zone2i);
hold on;
plot(x2, y2, 'g+');
grid on;
axis equal;

Answer 2

[Image Analyst]

In your code, see how n1 and n2 are passed into randn():

r = unifrnd(-0.9,0.9,[1,n1]);%r= r(randperm(length(r)));
t = unifrnd(-0.5,0.5,[1,n1]);%t= t(randperm(length(t)));
x = r + noise1*randn(1,n1);
y = t + noise1*randn(1,n1);
z = sqrt(noise2)*randn(fea_n,n1);
data1 = [x; y;z];

r = unifrnd(-1.9,1.9,[1,n1*2.5]);
t = unifrnd(-1.5,1.5,[1,n1*2.5]);
b = [r;t];
b(:,(abs(b(1,:)) <=1.15)+(abs(b(2,:))<0.65) ==2) =[];
n2 = size(b,2);
x = b(1,:) + noise1*randn(1,n2);
y = b(2,:) + noise1*randn(1,n2);
z = sqrt(noise2)*randn(fea_n,n2);
data2 = [x;y;z];

So you say "I want that the number of each color data point can be changed" so just change the value of n1 to change the number of points. You didn't show how you assigned it, but you must have assigned it before this code or else the code would not have worked.

[z cy]

In real, n1 and n2 have no relationship. I just dont know how to set this plot directly. So that I use delete

b(:,(abs(b(1,:)) <=1.15)+(abs(b(2,:))<0.65) ==2) =[];

n2 = size(b,2);

However,n2 is a number which can not directly to control. What I need is to directly to set n2, not use deleted the data and to get n2.

[Image Analyst]

Do you need to set a specific number for each color of marker, like a specified n1 for red, n2 for green, and n3 for blue? If so, you can just generate a new point and keep it or reject it if it falls in the proper "zone". So I'd just call rand() to get like a few million points. Then get 3 logical vectors to determine which rows are in zone1, zone2, or zone3. Then extract those rows and crop them to just the first n1, n2, or n3 rows. Then plot. See fully worked out code in my other answer on this page.