GraphMath

Change of basis

The same vector written in different coordinate systems

What changes when the vector stays the same but the basis changes?

A vector is geometric; its coordinate list depends on the basis used to describe it. This chapter shows how a basis matrix converts basis coordinates into standard coordinates, how its inverse converts standard coordinates back into basis coordinates and how the same idea extends to dimension-lowering coordinates and projections.

Key ideas

Change of basis separates a geometric vector from the coordinates used to describe it.

  • If A = [a₁ | a₂ | … | aₙ] is a square full-rank basis matrix, then x = AxA
  • The coordinate vector in basis A is xA = A⁻¹x
  • To convert coordinates from basis B to basis A, first move from B-coordinates to the vector, then move into A-coordinates: xA = A⁻¹BxB
  • For a tall full-column-rank matrix U, the left inverse U⁺ = (UᵀU)⁻¹Uᵀ maps vectors in ℝ³ to u-basis coordinates in ℝ²
  • The round trip UU⁺ is the projection matrix onto col(U)
  • Changing basis is useful when another coordinate system makes the computation simpler, lower-dimensional or more numerically stable

The chapter uses both 2D coordinate grids and the dimension-lowering case to show that basis choice changes coordinates, not the underlying vector.

Why is A⁻¹ the change-of-coordinates matrix?

The basis matrix A maps A-coordinates back to the original vector: x = AxA. Therefore the inverse map sends the original vector to its A-coordinates: xA = A⁻¹x.

Related chapters

Chapter contents

The PDF is a single document. The page links below are best-effort: most browsers support them, but some viewers may ignore the page hint.

Topic Pages
Change of basis: definition 1–2
Same vector in two coordinate systems 3–5
Change of basis with dimension lowering 5–6
Why change basis? 6–7

What changes in the dimension-lowering case?

A tall full-column-rank matrix U does not have a two-sided inverse. Instead, U⁺ = (UᵀU)⁻¹Uᵀ gives the left inverse that extracts coordinates in the u-basis, and the round trip UU⁺ returns the projection onto col(U).

Was this chapter helpful?

Quick feedback helps us improve the site.