From: Bi-linear matrix-variate analyses, integrative hypothesis tests, and case-control studies
Type | Description |
---|---|
(1) | on the projected samples of y _{ t }=w ^{T} x _{ t }, we use the one dimensional case of Equation (24) or the Welchâ€™s t-test to test Equation (1) merely along the normal direction of the boundary. |
(2) | measuring the distances of samples from a separating boundary, we consider \( s_{B}=\frac {\sum _{{\mathbf {x}}\in X^{(1)}_{1}\cup X^{(0)}_{0} } | \frac {{\mathbf {w}}^{T}({\mathbf {x}}-{\mathbf {c}}_0)}{\Vert {\mathbf {w}} \Vert }|{\!~\!}^{q}}{ \sum _{{\mathbf {x}}\in X^{(1)}_{0}\cup X^{(0)}_{1} } | \frac {{\mathbf {w}}^{T}({\mathbf {x}}-{\mathbf {c}}_0) }{\Vert {\mathbf {w}} \Vert }|{\!~\!}^q+\gamma _{B}}, \ q \ge 0. \) with q=2 for the square distance, q=1 for the Euclidean one. |