Combination of Pareto Optimal Front with Deep Neural Networks in Optimizing and Enhancing Performance of RF Designs RF Tasarimlarinin Performansini Optimize Etmek ve Artirmak Için Pareto Optimal ve Derin Sinir Aǧlari ile Birleştirilmesi


Kouhalvandi L., Matekovits L., Özoğuz İ. S.

31st IEEE Conference on Signal Processing and Communications Applications, SIU 2023, İstanbul, Turkey, 5 - 08 July 2023 identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/siu59756.2023.10223987
  • City: İstanbul
  • Country: Turkey
  • Keywords: Antenna, deep neural network (DNN), long short term memory (LSTM), multiple input, multiple output (MIMO), Pareto optimal front (POF), power amplifier (PA)
  • Istanbul Technical University Affiliated: Yes

Abstract

This paper focuses on the implementation of Pareto optimal front (POF) with deep neural networks (DNNs) for optimizing and integrating of radio frequency (RF) designs. POF typically generates the set of optimal tradeoffs while the DNNs employ the automated environment for estimating the targeted specifications. Hence combination of these two tools provides a strong optimization environment for the high-dimensional RF designs. This work illustrates the importance of this combination and explains the general procedure for optimizing RF active devices (such as amplifiers) and RF passive devices (such as antennas). The optimization method that is based on the POF idea is selected as the Thompson sampling efficient multiobjective optimization (TSEMO) algorithm and the DNN is elected to be constructed by the long short term memory (LSTM) layers. Here two separate configurations are considered: the optimized amplifier is operating in the frequency band from 1.7 GHz to 2.2 GHz. Additionally, the antenna design is operated in the frequency band from 3.1 GHz to 10.6 GHz for proving the importance of POF method where DNN is employed to enhance the overall performance of the RF designs.