Table 1.

The training stage of LWNN.

Input: the training dataset.

Procedure:

(1) Start with a set of predefined classes, Class1, Class2,…, ClassN. For each class, Classi,

(2) Compute its centroid $\overrightarrow{o_i}$ and its associated weighting vector $\overrightarrow{w_i}$ by Equation (1) and (4).