Objective Functions in Twin Support Vector Machines

I'm reading paper Twin Support Vector Machines for Pattern Classification by Jayadeva et al. (2007). In that paper, the authors proposed using two non-parallel hyperplanes for classifying two classes. The objective function for learning one hyperplane is:

$$ \underset{w^{(1)}, b^{(1)}, q}{\text{Min}} \frac{1}{2} (Aw^{(1)} + e_1 b^{(1)})^T (Aw^{(1)} + e_1 b^{(1)}) + c_1 e_2 q^T \\ \text{subject to} -(B w^{(1)} + e_2 b^{(1)}) + q \ge e_2, q \ge 0 $$

, where $A$ and $B$ are data points belong to class +1 and class -1 respectively, $e_1$ and $e_2$ are vectors of ones with appropriate dimensions.

The authors then stated that the first term in the objective function is the distance between all points of class label +1 to the hyperplane. What I don't understand here is that the distance between a point to a hyperplane is:

$$ d = \frac{|\boldsymbol{w}^t x + b|}{|\boldsymbol{w}|} $$

, then shouldn't the first term in the objective become $\frac{(Aw^{(1)} + e_1 b^{(1)})^T (Aw^{(1)} + e_1 b^{(1)})}{w^T w}$? Why the authors just optimize only the nominator?