A Neural Network-Based Solver for Closed and Open Electromagnetic Structures
Conference
Priota, NZ, Volakis, JL, Pulugurtha, MR et al. (2026). A Neural Network-Based Solver for Closed and Open Electromagnetic Structures
. 2015 IEEE INTERNATIONAL SYMPOSIUM ON ANTENNAS AND PROPAGATION & USNC/URSI NATIONAL RADIO SCIENCE MEETING, 965-967. 10.1109/AP-S/USNC-URSI60190.2026.11674878
Priota, NZ, Volakis, JL, Pulugurtha, MR et al. (2026). A Neural Network-Based Solver for Closed and Open Electromagnetic Structures
. 2015 IEEE INTERNATIONAL SYMPOSIUM ON ANTENNAS AND PROPAGATION & USNC/URSI NATIONAL RADIO SCIENCE MEETING, 965-967. 10.1109/AP-S/USNC-URSI60190.2026.11674878
In this paper, we introduce a neural-network-based framework for accurate modeling of both closed and open radiating electromagnetic structures. A systematic investigation is conducted to assess the impact of activation functions on solution accuracy and convergence behavior when solving electromagnetic problems. In particular, we examine how commonly used state-ofthe-art activation functions, such as tanh, Swish, and sigmoid, influence the performance of neural-network-based electromagnetic solvers.
