Orthogonal Projection onto a Subspace
You'll learn: How to project any vector onto a higher-dimensional subspace using orthogonal bases, extending the line projection idea to planes and beyond.
The Core Idea
You've already learned to project a vector onto a single line. Now we generalize: what if you want to project a vector onto a plane in ℝ³, or any subspace of dimension greater than one?
The key insight is decomposition: just as we decomposed vectors using orthogonal complements, we can break any vector v into two parts:
- proj_W(v): the part that lies entirely in the subspace W
- v - proj_W(v): the part orthogonal to W
How It Works
Given a subspace W with an orthogonal basis {u₁, u₂, ..., uₖ}, the projection of v onto W is simply the sum of projections onto each basis vector:
proj_W(v) = proj_u₁(v) + proj_u₂(v) + ... + proj_uₖ(v)
Each term is computed just like projection onto a line: proj_uᵢ(v) = (⟨v, uᵢ⟩ / ⟨uᵢ, uᵢ⟩) uᵢ
If you have an orthonormal basis instead, it's even cleaner: proj_W(v) = ⟨v, u₁⟩u₁ + ⟨v, u₂⟩u₂ + ... + ⟨v, uₖ⟩uₖ
Why Orthogonal Bases Matter
The formula works only with orthogonal bases because orthogonal vectors don't interfere with each other—their contributions are independent. This is why Gram-Schmidt is so powerful: it gives us the right basis to make projection straightforward.
Analogy: Think of projecting a 3D object's shadow onto a tilted plane. The shadow is the "closest" version of the object that fits in that plane—that's your orthogonal projection.
Key Takeaway: Orthogonal projection onto a subspace decomposes a vector into its component within the subspace (the projection) and its orthogonal complement, computed by summing line projections onto an orthogonal basis.