We have a subspace (the span of ), and we want to find the vector () closest to in . We also want to find such that . Diagrammatically, We know , so: We also know and (), so:

So finally,

Orthogonal Projection

Let non-zero , and . The orthogonal projection of onto is the vector in the span of that is closest to .

\text{proj}_\vec{u}\vec{y}=\frac{\vec{y}\cdot\vec{u}}{\vec{u} \cdot \vec{u}}\vec{u}

Also, and,

From this we can conclude (look at the diagram),

||\vec{y}||^2=||\text{proj}_\vec{u}\vec{y}||^2+||\vec{z}||^2

Best Approximation

Best Approximation Theorem

Let be a subspace of , and is the orthogonal projection of onto . Then for any , we have


Orthogonal Decomposition

Orthogonal Decomposition Theorem

Let be a subspace of . Then, each has a unique decomposition.

If is the orthogonal basis for ,

is the orthogonal projection of onto