Answers are grounded in this paper and cite their sources.
In mask‐based direct electromagnetic (EM) optimization, sampled specification violations are commonly aggregated into a scalar objective through a selected error norm. Although this step is often treated as a secondary implementation detail, it can modify the observed behavior of the optimization process. This paper investigates the influence of error‐norm selection in direct EM optimization of microwave filters through a controlled comparison of the , , and formulations. Two benchmark problems are considered: a three‐variable low‐pass filter (LPF) and a five‐variable band‐pass filter (BPF). For each benchmark, the EM model, solver settings, frequency sampling, optimization algorithm, and initialization protocol are kept fixed, and only the aggregation norm is varied. The comparison is conducted in terms of final objective value, strict mask satisfaction, tolerance‐based near‐feasibility, number of EM simulations, and sensitivity to initialization over fifteen deterministic seeds per benchmark, resulting in 45 direct EM optimization runs per benchmark. Under the adopted setup, the norm choice altered the observed optimization behavior in both benchmarks. For the LPF case, the formulation provided the most favorable overall behavior, achieving the highest exact‐success rate and the lowest mean and median number of EM evaluations. For the BPF case, the formulation achieved the most favorable behavior under the strict feasibility criterion and at tighter tolerance levels. These results indicate that, in mask‐based direct EM optimization, the aggregation norm is better viewed as an explicit methodological choice rather than a purely conventional implementation detail.
No takes yet. Share an insight, caveat, or question.
Davalos‐Guzman et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: