Matrix Functions via Diagonalization
You'll learn how to evaluate complicated functions of matrices—like e^A or sin(A)—by exploiting diagonalization to reduce them to simple scalar operations.
The Core Idea
When you have a diagonalizable matrix A, you can write it as A = PDP⁻¹ where D is diagonal. This decomposition unlocks a powerful shortcut: applying any function to A becomes as simple as applying that function element-wise to the diagonal entries of D.
Why does this work? Because if A = PDP⁻¹, then:
- A² = (PDP⁻¹)(PDP⁻¹) = PD²P⁻¹
- A³ = PD³P⁻¹
- And in general: f(A) = Pf(D)P⁻¹
Since D is diagonal, computing f(D) is trivial—just apply the function to each diagonal entry independently.
How It Works in Practice
Suppose you need to compute e^A (the matrix exponential). The definition involves an infinite series: e^A = I + A + A²/2! + A³/3! + ...
Computing this directly would be nightmarish. But with diagonalization:
- Find P and D such that A = PDP⁻¹
- Compute f(D) by applying the function to diagonal entries:
- If D = diag(λ₁, λ₂, λ₃), then e^D = diag(e^λ₁, e^λ₂, e^λ³)
- Reconstruct: e^A = Pe^DP⁻¹
This same technique works for any function expressible as a power series: sin(A), cos(A), √A, and so on.
Think of it like this: Diagonalization transforms your matrix into a "coordinate system" where it acts like simple scaling. In that friendly coordinate system, complicated functions become elementary arithmetic. Then you transform back.
Key Takeaway: Matrix functions like e^A or sin(A) can be efficiently computed by diagonalizing A, applying the function to the diagonal eigenvalues, then transforming back using f(A) = Pf(D)P⁻¹.