From Equations to Algorithms and Data: Transforming Microwave Engineering and Education with Machine Learning
Organizations: ICTEAM, Université catholique de Louvain Ottignies-Louvain-la-Neuve, Belgium
Abstract
Conventional microwave engineering education relies heavily on analytical methods, canonical circuit topologies, and intuition-driven design, which have proven effective at microwave frequencies. However, as systems increasingly operate in the millimeter-wave and terahertz regimes, parasitic effects, process-dependent electromagnetic interactions, and ultra-wideband performance requirements challenge both topology/layout-constrained traditional design methodologies and existing teaching paradigms. This paper proposes a pedagogical shift in microwave and RFIC (Radio Frequency Integrated Circuit) engineering and education by introducing machine-learning (ML) and data-driven electromagnetic synthesis as a complementary design framework for microwave circuits such as power dividers and combiners, couplers, and baluns. Rather than emphasizing predefined topologies, the proposed approach enables topology-agnostic, performance-oriented exploration of the design space, allowing students to directly engage with electromagnetic behavior through specification-driven synthesis. By integrating machine-learning-based inverse design and multi-objective optimization into the curriculum, the framework enhances physical intuition, encourages design creativity, and better aligns microwave education with emerging industrial practices in high-frequency and ultra-wideband system design.
Figures & tables
| Method | Requires Dataset | Algorithms | Advantages | Disadvantages |
| Evolutionary Optimization (EA) | No | NSGA-II, CMA-ES, DE, PSO | Directly optimizes performance using EM simulations without reliance on pre-trained models or datasets. | Sensitive to the initial solution (random or template-based). Runtime scales poorly with EM simulation cost. Performance is highly dependent on the score function definition. |
| Gradient-Free Local Search | No | DFO-TR (continuous), Tabu Search (discrete) | Rapid optimization of a single layout with strong local improvement. | Limited exploration capability. Scales poorly with increasing pixel count and is highly sensitive to the initial design. |
| Bayesian Optimization (BO) | Optional | GP-BO, Batch BO, Multi-fidelity BO | Constructs a probabilistic surrogate of the simulator, enabling sample-efficient optimization and reuse of learned information. | Performance degrades in high-dimensional spaces and is sensitive to acquisition strategy selection. |
| Reinforcement Learning | Optional | PPO, TRPO, DQN | Learns a policy that directly optimizes expected performance without explicit S-parameter matching. | Requires large training budgets and careful tuning of reward formulation and hyperparameters. |
| Neural Surrogate Models | Yes | CNN, DNN, other neural networks | Enables fast inference once trained and significantly reduces reliance on EM simulations. | Requires large datasets and often generalizes poorly across structure families or fabrication technologies. |
| Surrogate Model + EA Hybrid | Yes | Surrogate models + EA (NSGA-II, CMA-ES, PSO) | Combines fast surrogate-driven optimization with periodic EM-based validation. | Limited by surrogate accuracy and training data coverage; extrapolation is unreliable. |