Minimization solution of three equations in two variables

In summary, the conversation discusses the idea of using a distance formula to find the best fit of multiple lines or planes in the topic of optimization. The speaker also mentions the concept of regression and how it can be simplified using the formula ##X=(A^TA)^{-1}A^TY## if ##A^TA## is nonsingular. This is known as the least sum squares solution. The conversation ends with the realization that this method is surprisingly simple.
  • #1
barryj
856
51
Homework Statement
given these three equations: I know I have more equations than variables. However, isn't there a way to find the closest solution, some sort of regression solution?
2x - y = -3
-2x - y = -4
-2.1x - y = 4
Relevant Equations
2x - y = -3
-2x - y = -4
-2.1x - y = 4
I do not know the solution.
 
Physics news on Phys.org
  • #2
I am thinking about how regression is performed. Let's assume I plot the three equations and they form a triange where they intersect. I can use the "distance from a point to a line" formula to get the distance from an arbitrary point , (X0,Y0) within the triangle to each of the lines. I think i could then square and add the distances and try to minimize the resulting function D(X0,Y0) . It seems that this could be extended to find the best fit of multiple lines or even planes. I guess this is the topic of optimization. I do not know if this is a good way or not.
 
  • #3
barryj said:
isn't there a way to find the closest solution, some sort of regression solution?
Given ##AX=Y##, ##A^TAX=A^TY##.
IF ##A^TA## is nonsingular you have ##X=(A^TA)^{-1}A^TY##.
This can be shown to be the least sum squares solution.
 
  • #4
Amazing! Hard to imagine it is this simple
Thanks.
 
Back
Top