algorithm - Incremental Linear Programming -

I'm trying to implement 1D linear programming, but I think I remember something in connecting dots. Has been there. I want to calculate the convex hull of a point set, so my obligations are line created by the row pair. I am trying to create a incremental algorithm that combines the numbers / lines one by one.

  min x_m * a + y_m * b s.t. X_i * a + y_i * b + z_i & gt; = 0a ', b' - & gt; Optimize with respect to 1 ... i (x_i * a '+ y_i * b' + z_i & gt; = 0) Do nothing else x_i * a '+ y_i * b' + z_i = 0b '= (- Z_i - X_i * a ') / y_i one' =?  

I think I might have messed up in my signature, and this is misleading me, but how do I calculate 'and B'?

I have tried giving my problem again:

I got an online vector drawer, so I hope that these pictures should be understood.

(This is a strange rotation)

In 2D, the vertical line is defined only as x-coordinate x_m because it There is just an infinite line, I need to find the line drawn, and the way I do this, it is that I start with the line defined by y = ax + b, b = -infinity and a = 0. Then I connect every point in random order one by one. For each point, I check that it is below this line, and if not, then I have found a new line by locking this point, and going through all the other points, the point on the other side of x_m Finding me, give me the biggest one, when I draw a line through the lock point and the new checked point After all, I get a line through two points, with all the points below it piercing x_m.

then max ax_m + b s.t. Y_1 & lt; = Ax_1 + b y_2 & lt; = Ax_2 + b y_i & lt; = Ax_i + b

I'm fine with this job.

In 3D, I want to find the face, which is defined by 3 digits (or a plane z = ax + by + c), along with all other points below, linear programming use. Here, I start with a horizontal plane -infinity, connecting points together, and checking whether they are under my plane if they are not, then I have previously posted L.P. I try to maximize To do this, I should be able to maximize the function of the function, instead of just z = ax +, instead of y = ax + b. (I ignore C as if I ignore b in 2D when I lock one digit and max in 1d).

  min x_m * a + y_m * b s.t. X_1 * a + y_1 * b + z_1 & gt; = 0 s.t. X_2 * a + y_2 * b + z_2 & gt; = 0 s.t. X_i * a + y_i * b + z_i & gt; = 0a ', b' - & gt; Optimize with regard to P_1 ..., p_i if (x_i * a '+ y_i * b' + z_i> = 0) and do nothing else x_i * a '+ y_i * b' + z_i = 0b '= ( - Z_i - x_i * a ') / y_i one' =?  

Does this make sense now?

Actually, I think that I get the form z = ax + by In order to solve the linear programs gradually there is some need.