VECTROID /ˈvek.trɔɪd/ n. A matrix with high aspect ratio, exhibiting vector-like character. ──────────────────────────────────────────────────────────── DEFINITION A vectroid is a matrix where one dimension dominates the other: m >> n (tall and thin) m << n (short and wide) Such matrices behave more like vectors than transformations. ETYMOLOGY vector + -oid ("resembling") cf. ellipse → ellipsoid sphere → spheroid vector → vectroid EXAMPLES VECTROID SQUARE MATRIX [a b] [a b c] [c d] [d e f] [e f] [g h i] [g h] 4×2 = 8 elements 3×3 = 9 elements max rank = 2 max rank = 3 embeds / projects transforms CHARACTER PROPERTY VECTROID SQUARE MATRIX ─────────────────────────────────────────────────────────── Shape m >> n or m << n m ≈ n Maximum rank min(m,n) — small min(m,n) — large Action embed or project transform in-place Invertible? No Possibly Decomposes to outer products rotations + scalings THE RANK-1 EXTREME The maximally vectroid matrix is rank-1: the outer product of two vectors. [a] [ax ay az] [b] · [x y z] = [bx by bz] [c] [cx cy cz] Any matrix decomposes into a sum of rank-1 (vectroid) pieces. This is the basis of low-rank factorization and SVD. IN UNIVERSAL LANGUAGE [I] [[A]] [I] The notation encodes aspect ratio: [ ] single bracket → vectroid (stores) [[ ]] double bracket → square(ish) (transforms) The brackets are self-documenting: [I] narrow, tall/thin, input/output [[A]] wide, balanced, transformation The viewing lens Aᵢ in xᵢ = Aᵢ · Θ + εᵢ is typically a vectroid — extracting a low-dimensional observation from latent structure Θ. TENSOR SPECTRUM vectroid ◄────────────────────────────► matrixoid [ ] [ ] [ ] aspect ratio → [ ] [ ] [ ] [ ] stores transforms embeds rotates projects mixes (The term "matrixoid" is provisional; square matrices may simply be padded.) SEE ALSO • Universal Language — IAI framework • Low-rank approximation • Singular Value Decomposition (SVD) • Embedding matrices in neural networks ──────────────────────────────────────────────────────────── Published 2026-01-07 · iai