Tensor Hinted Mv Conjectures
Zhao Song
TL;DR
The paper generalizes the Hinted Mv conjecture to tensor inputs, presenting Tensor Hinted Mv in Type I and Type II formulations with structured inputs and tensor operators. It leverages a tensor-trick to convert a k-way interaction into a Hadamard product of per-mode computations, expressed as $ ( \oslash_{j=1}^k P_j )^T ( \oslash_{j=1}^k V_j ) = \odot_{j=1}^k ( P_j^T V_j ) $. The authors propose two phase-based algorithmic strategies, provide polynomial-time considerations for the tensor setting, and state conjectured phase-wise lower bounds that scale with $k$; the $k=1$ case reduces to the original conjecture. A concise algebraic justification shows that the tensor construction satisfies $C = ⊙_{ℓ=1}^k C_{ℓ}$ when $C_{ℓ} = A_{ℓ}^T B_{ℓ}$, supporting the tensor-trick framework.
Abstract
Brand, Nanongkai, and Saranurak introduced a conjecture known as the Hinted Mv Conjecture. Although it was originally formulated for the matrix case, we generalize it here to the tensor setting.
