Your (correct) work so far enables you to say that the first three columns span the column space of the matrix. So take the original three columns and apply Gram-Schmidt to them. Do you understand the Gram-Schmidt procedure?
1 3 1 2 2
-2 1 3 3 -2
1 2 1 1 1
1 2 2 1 0
3 1 2 -2 -1
Gauss elimination gives me the pivot positions in colonn number 1,2 and 3.
I think i have to use a normalised Gram-Smith method but i dont understand how to do it. plz help.
/Maria
Your (correct) work so far enables you to say that the first three columns span the column space of the matrix. So take the original three columns and apply Gram-Schmidt to them. Do you understand the Gram-Schmidt procedure?
To produce an ON basis by using the Gram-Schmidt procedure on a linearly independent set follow these steps:
1. Let
2. For the remaining vectors in the ON basis, follow a two-step cumulative procedure: first, take the next vector and subtract off the amount of that vector that is in the direction of all previously computed vectors in the ON basis. In this case, set Here is the inner product, or the dot product.
3. Normalize to get the next ON vector:
4. Set (You need to subtract off the amount in the directions of both the previous two ON vectors.)
5. Then
Continue. Make sense?
I'm not sure i understand but ill try.
The first thing I do is divide my first vector [1,2,1,1,3] with the length of the vector (sqrt(16)). This gives me [1/4,2/4,1/4,1/4,3/4], and this is the first vector in my On base right?
After this step is done (if its right) I am supposed to subtract the inner product from the next vector in my original matrix.
This is what i don't understand how to do.
Also I think I am mixing up the terms, in my textbook it says I should use the following formula to calculate a Orthogonal basis. Can i use this and then normalize the answers that i get?
v1=x1
v2=x2-((x2v1)/(v1^2))*v1
v3=x3-((x3v1)/(v1^2))*v1-((x3v2)/(v2^2))*v1
Thanks in advance
/Maria
The formulas you wrote down are the orthogonal Gram-Schmidt, not the orthonormal Gram-Schmidt. You can normalize all the vectors afterwards if you like, as you suggested. Either way should work. The amount of work will be about the same in either case. Let's say you continue down the path I laid out. (Incidentally, you have sign error in your first vector. It should be a negative 2/4 in the second position.) Your next calculation would be
To get the normalized vector, divide by its length, which in this case is Therefore, we have
See how this works?