Rank-1 tensor

Rank-1 tensor — the outer product of one vector per axis, a ∘ b ∘ c. Spanish: tensor de rango 1.

Read it one entry at a time and the outer product is just a product: T[i, j, k] = a[i] · b[j] · c[k]. In NumPy, np.einsum("i,j,k->ijk", a, b, c) builds it directly, with one output axis per input vector.

The notebook’s animation lights each vector’s peak — a at index 0, b at 2, c at 3 — and the cell where they meet is the loudest entry: T[0, 2, 3] = 4 · 5 · 4 = 80. Its address is read straight off the three argmaxes.