Deep Learning for Speeding Up the Min Slot-Continuity Capacity Loss Spectrum Assignment

This paper presents two classifier models based on deep neural networks to speed up the Min Slot-Continuity Capacity Loss (MSCL) spectrum assignment. The first decides between the use of First-Fit or MSCL heuristic, with the aim of avoiding unnecessary MSCL calls whenever the application of First-Fit would provide the same minimum loss of capacity as MSCL. The second adds the capability of pointing out the correct portion of the spectrum MSCL should look for whenever First-Fit is not selected. Simulation results demonstrate reductions of 28% and 74% on the simulation time between the MSCL and the two proposed models without mitigation on the MSCL performance.