Fig. 2.

Simple Example of Rank-1 Tensor Regression with D=2. Suppose a univariate outcome y with canonical link g(·) and predictor tensor $\mathcal{Z}$ with D=2 dimensions. Each dimension in the predictor tensor corresponds to a feature set, and each feature set contains its own coefficient vectors. The coefficient vectors in this example, β₁ and β₂, may be estimated by collecting outcomes and predictor tensors across multiple subjects and applying the Block Relaxation Algorithm for estimation