Multilayer Perceptron Learning for Decreasing the Resistance of Graphene Oxide Thin Films

Abstract

The capability of the classifier to differentiate between the similarities and differences in the resistance of thin films of reduced Graphene Oxide (rGO) is put to the test via a process known as multilayer Perceptron optimisation. In particular, for the purpose of this research, we used the learning methods of scaled conjugate gradient, levenberg-marquardt, and robust back propagation. As the dataset for this investigation, the sheet resistance of rGO thin films that was given by MIMOS Bhd was used. In order for us to carry out this test, we had to gather data from several sheets of rGO thin film, each of which included a unique combination of thickness and resistance. Normalisation, randomization, and splitting were the three types of pre-processing that were applied to both the input and the output data. There was a thirty-seven percent split between the data used for training, fifteen percent split between the data used for validation, and fifteen percent split between the data used for testing. Different numbers of hidden neurons, ranging from one to ten, were used in MLP in order to enhance the effectiveness of the learning procedures. Because to their hard work, the most cutting-edge learning algorithms that have ever been built for finding MLP in rGO sheet resistance uniformity have been produced. Measurements were taken of everything from mean squared errors (MSE) to accuracy throughout all of the training, validation, and testing datasets, as well as overall performance. The whole investigation was dependent on using the MATLAB programme, version R2018a, to carry out in a mechanised manner all of the necessary analytical procedures. Throughout the course of constructing a knowledge process in MLP used for rGO sheet confrontation, it was found that the LM had a major influence on the whole process. The MSE for LM has been lowered the greatest, particularly in comparison to SCG and resilient backpropagation (RP).