Abstract. In this work, we use adaptive activation functions for regression fractional physics-informed neural networks (fPINNs) to approximate nonsmooth solutions. The adaptive activation function has better learning capabilities than the traditional one because it improves the convergence rate and solution accuracy. Our simulation results show that the adaptive parameter contributes less to the improvement of the results as the singularity becomes more strong (a decreases), because the errors incurred from the discretization and optimization of the loss function become dominant.

Abstract. In the literature, there are many works where Artificial Intelligence (AI) is used to predict postprandial blood glucose (BG). However, the following questions are asked: is it necessary to develop AI techniques to be used in the prediction of BG? How important is AI for this purpose? In this work, these questions are answered considering the use of postprandial BG prediction to obtain an optimal insulin bolus. According to the proposed model, the error obtained in the postprandial glycemia in relation to the target glycemia is in the same proportion as the error made in the prediction of this postprandial glycemia, which implies that greater errors in the prediction also lead to greater postprandial glycemia errors. In this way, it can be said that it is important to use AI to make the smallest errors in blood glucose prediction.

Case-Based Reasoning (CBR) is one of the Artificial Intelligence (AI) techniques often applied in healthcare due to its simplicity and good results. Recently, many health issues have been improved by applying this technique, many of which are related to diabetes, as it is a disease that requires strict and time-consuming daily management. Thus, in this work, the CBR technique is used to predict the postprandial blood glucose in patients with type 1 diabetes, given the importance that this prediction has, for example, in the calculation of the insulin bolus. An architecture with new metrics is developed to generate learning and obtain an increasingly intelligent CBR system.