Dual-Channel Tensor Neural Networks: Finite-Sample Theory and Conformal Structure Selection
Researchers have introduced a Dual-Channel Tensor Neural Network (DC-TNN) designed to handle complex tensor-valued data more effectively. This new model decomposes tensor inputs into a low-rank core and a sparse refinement, processing them through separate, linked neural channels. The framework offers a structure-agnostic approach and includes a novel conformal structure selector for choosing tensor decompositions with finite-sample validity. AI
IMPACT Introduces a new method for analyzing complex tensor-valued data, potentially improving performance in fields like neuroimaging and genomics.