To find out the optimal density, we consider the FAP density as the independent variable and calculate the first derivative of Equ. (28).
We again let [[lambda].sub.f] / [[lambda].sub.u] = r, and using Equ. (30) we obtain:
From Equ. (31), we can discover that the coverage probability only depends on the FAP-to-UE ratio with the same relative change of the FAP density.
However, solving Equ. (32) symbolically is difficult.
Calculate the distance from the new cluster to all other clusters by means of Equ.7.
In another word, Equ. 8 needs to be minimized as follows.
However, if just using Equ.9, an "over-fitting" problem may arise.
The loss functions (Equ. 12-14) are the objective function to be optimized, which is NP-hard [11].