6.23.3 Tensor Component Count by Type
Understanding how tensor components are categorized by their types and counted in algebraic structures.
Tensor Component Count by Type is the refinement of the component count relation that examines what role the specific type (p, q), rather than merely the total order p + q, plays in determining a tensor's component count, particularly in the more general setting where the contravariant slots and covariant slots draw from vector spaces of different dimensions. Where component count by order and by dimension treat d^(p+q) as a function of a single underlying dimension, this topic addresses the case where type must be tracked explicitly because the two kinds of slots are no longer interchangeable in their contribution to the total count.
When Type Does Not Affect the Count
The Single-Space Case
When all contravariant and covariant slots refer to the same vector space V (of dimension d) and its dual V* (necessarily also of dimension d, since dim V* = dim V), the component count N = d^(p+q) depends only on the sum p + q, not on how that sum is split. A type (3,0) tensor and a type (0,3) tensor over the same V both have d³ components, and a type (2,1) tensor has the same count d³ as well — in this setting, type genuinely does not affect the component count, only order does.
When Type Does Affect the Count
Tensors Built From Two Different Spaces
The situation changes for a mixed tensor built from two genuinely different spaces, such as T ∈ Hom(V, W) ≅ V* ⊗ W with dim V = m and dim W = n. Here the single contravariant-style slot (valued in W, dimension n) and the single covariant-style slot (valued in V*, dimension m) contribute different factors, giving a component count of
rather than a single power of a common dimension. This is exactly the familiar fact that an m × n matrix has mn entries, generalized to the tensor setting: the type here is effectively (1 slot valued in W, 1 slot valued in V*), and the count is the product of the two relevant dimensions, not a power of one dimension.
General Multi-Space Formula
For a tensor built from p copies of spaces W₁, ..., W_p (contravariant-style slots) and q copies of dual spaces U₁*, ..., U_q* (covariant-style slots), the component count generalizes to the product
which reduces to d^(p+q) exactly when every space involved shares the common dimension d, recovering the simpler single-space formula as a special case.
Diagram Contrasting the Two Regimes
Why This Distinction Matters in Practice
Linear Maps Between Spaces of Different Size
The type-sensitive formula is the one actually used whenever a linear map goes between spaces of different dimension, such as a map from a 4-dimensional feature space to a 2-dimensional output space, represented by a 2 × 4 matrix with 8 components — a count that could not be recovered from any single-dimension power formula, since no common d exists for the two spaces involved.
Type as the Organizing Data for Multi-Space Tensors
When multiple distinct spaces are involved, it is no longer sufficient to say a tensor has "order 2"; the type must specify not just how many contravariant and covariant slots exist but which specific space each slot draws from, since the component count formula depends on that finer data. This is the sense in which component count by type is a genuinely distinct and more general question than component count by order or by a single shared dimension, becoming essential exactly when the simplifying assumption of one common vector space is dropped.