Table 5 Multi-objective optimization using the introduced procedure and the benchmark algorithms: cost breakdown.

From: Multi-objective artificial-intelligence-based parameter tuning of antennas using variable-fidelity machine learning

Algorithm

Optimization cost# and hypervolume@

Antenna I

Antenna II

Antenna III

Antenna IV

This work (variable-fidelity

ML with ANN surrogates)

150.1 [0.65]

150.4 [0.67]

252.5 [0.77]

264.4 [0.67]

Algorithm 1

N = 400

400 [0.55]

400 [0.43]

400 [0.63]

400 [0.59]

N = 1600

1600 [0.58]

1600 [0.40]

1600 [0.65]

1600 [0.58]

Algorithm 2

N = 400

400 [0.50]

400 [0.42]

400 [0.67]

400 [0.54]

N = 1600

1600 [0.55]

1600 [0.44]

1600 [0.68]

1600 [0.55]

Algorithm 3

390 [0.60]

320 [0.61]

330 [0.76]

340 [0.66]

  1. # The cost of benchmark algorithms is equivalent to the total number of EM simulations performed. For the proposed method, the cost is expressed in terms of the equivalent number of high-fidelity EM simulations by taking into account the time evaluation ratio between the high- and lower-fidelity models.
  2. $Hypervolume, provided in square brackets, is computed for normalized objectives.
Back to article page