Continuous glucose monitors (CGM) and insulin pumps are becoming increasingly important in diabetes management. Additionally, data streams from these devices enable the prospect of accurate blood glucose prediction to support patients in preventing adverse glycemic events. In this paper, we present Neural Physiological Encoder (NPE), a simple module that leverages decomposed convolutional filters to automatically generate effective features that can be used with a downstream neural network for blood glucose prediction. To our knowledge, this is the first work to investigate a decomposed architecture in the diabetes domain. Our experimental results show that the proposed NPE model can effectively capture temporal patterns and blood glucose associations with other daily activities. For predicting blood glucose 30-mins in advance, NPE+LSTM yields an average root mean square error (RMSE) of 9.18 mg/dL on an in-house diabetes dataset from 34 subjects. Additionally, it achieves state-of-the-art RMSE of 17.80 mg/dL on a publicly available diabetes dataset (OhioT1DM) from 6 subjects.
The structural complexities of grain boundaries (GBs) result in their complicated property contributions to polycrystalline metals and alloys. In this study, we propose a GB structure descriptor by linearly combining the average two-point correlation function (PCF) and standard deviation of PCF via a weight parameter, to reveal the standard deviation effect of PCF on energy predictions of Cu, Al and Ni asymmetric tilt GBs (i.e., Σ3, Σ5, Σ9, Σ11, Σ13 and Σ17), using two machine learning (ML) methods; i.e., principal component analysis (PCA)-based linear regression and recurrent neural networks (RNN). It is found that the proposed structure descriptor is capable of improving GB energy prediction for both ML methods. This suggests the discriminatory power of average PCF for different GBs is lifted since the proposed descriptor contains the data dispersion information. Meanwhile, we also show that GB atom selection methods by which PCF is evaluated also affect predictions.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.