Improving the tomographic image by enhancing the machine learning algorithm

Abstract

Hyperparameter optimization in machine learning models may help enhance the efficiency of obtaining high-quality tomographic pictures, the purpose of this paper. In the discipline of electrical impedance tomography, machine learning techniques are utilized to translate voltage measurements into reconstruction pictures. Because of this, the so-called "inverse problem" arises, whereby the optimal answer must be sought. Effective machine learning relies heavily on the appropriate choice of model coefficients (hyperparameters). As a consequence, the strategies used to improve this choice have an indirect effect on the final reconstruction. The K-nearest neighbors strategy may be utilized to improve a machine learning model based on linear regression and classification models, as we show in this paper. Electrical tomography, a technology that analyses flood embankments from the interior to measure their structural integrity, makes use of the methods outlined above. The data gathered shows that the suggested solutions work.