Advanced Autoencoder Transfer Function Parameter Extraction Technique for Neuro-TF Parametric Modeling of Microwave Components

Jinyi Liu,Feng Feng,Wei Liu,Jianguo Xue, Shaochang Liu, Fang Gao, Xiaolong Li,Qi-Jun Zhang


Cited 0|Views3
No score
Recently, neuro-transfer function (neuro-TF) has become a recognized method for electromagnetic (EM) parametric modeling. The existing neuro-TF methods use the vector fitting technique to perform transfer function (TF) parameter extraction, commonly encountering nonsmoothness and discontinuity issues for the extracted TF parameters with respect to geometrical parameters. This letter proposes an advanced autoencoder TF parameter extraction technique for neuro-TF parametric modeling of microwave components. In the proposed technique, the autoencoder is introduced to extract a set of TF parameters as TF parameters from the $S$ -parameters with the encoder part and generate a decoder function as the TF in the original neuro-TF model. The TF parameters extracted using the proposed technique behave much smoother than the TF parameters extracted using traditional vector fitting. Meanwhile, the proposed technique avoids the discontinuity problem caused by vector fitting in the standard neuro-TF method. Parametric modeling using the smooth TF parameters can thus have higher accuracy than modeling with nonsmooth TF parameters. The proposed technique is demonstrated by two examples of EM parametric modeling of microwave components.
Translated text
Key words
Microwave filters,Parameter extraction,Vectors,Fitting,Neural networks,Parametric statistics,Decoding,Autoencoder,microwave components,neural networks,parameter extraction,parametric modeling
AI Read Science
Must-Reading Tree
Generate MRT to find the research sequence of this paper
Chat Paper
Summary is being generated by the instructions you defined