Zgadzam się
Nasza strona zapisuje niewielkie pliki tekstowe, nazywane ciasteczkami (ang. cookies) na Twoim urządzeniu w celu lepszego dostosowania treści oraz dla celów statystycznych. Możesz wyłączyć możliwość ich zapisu, zmieniając ustawienia Twojej przeglądarki. Korzystanie z naszej strony bez zmiany ustawień oznacza zgodę na przechowywanie cookies w Twoim urządzeniu.
Strongly correlated predictive variables in linear models make it difficult to determine the exact influence of these predictors on the dependent variable (output). The use of a simple least squares method to estimate unknown parameters may lead to an erroneous prediction. Adding the penalty factor depending on the number of parameters to the least-squares criterion allows to reduce the variance of estimators and to determine the estimators of the load. This article proposes a new solution based on machine learning methods, which enabled obtaining more accurate and stable reconstruction results in the process of solving the inverse problem. Image reconstructions were carried out using the ElasticNET, least-angle regression (LARS) and ElasticNET + artificial neural networks (ANN) hybrid algorithms. The main task of the tomography is to perform an accurate reconstruction of the image. During the measurements it was found that the measured values from some electrode pairs are strongly correlated. The reason is usually the method of carrying out the measurement. During the research, a hybrid method was also presented, in which ElasticNET reduces the vector of predictors, which is then processed by ANN. This approach speeds up the processes of training neural networks and image reconstruction, as well as makes the system immune to the noise of input data.
