We demonstrate efficient ultra-compact power splitters designed via machine learning algorithm viz. binary-additive reinforcement learning algorithm. Two different splitter geometries (Y- and T- junctions) are shown; each with an area footprint of 1.2 x 1.2 µm2. The simulated insertion loss is < 1dB for both.
Paper
Full text
Ultra-compact Design of Power Splitters via Machine Learning
Semantic Scholar · Engineering · 2020
Abstract
We demonstrate efficient ultra-compact power splitters designed via machine learning algorithm viz. binary-additive reinforcement learning algorithm. Two different splitter geometries (Y- and T- junctions) are shown; each with an area footprint of 1.2 x 1.2 µm2. The simulated insertion loss is < 1dB for both.