Obstructive sleep apnea (OSA) is a common sleep disorder characterized by repeated brief interruptions of breathing during sleep. The existing methods of OSA detection using polysomnography (PSG) are inefficient and costly, and therefore the computer-aided diagnosis has become an important approach to improve the diagnostic efficiency and effectiveness. In this paper, we design a novel deep learning OSA detection model based on electrocardiograms (ECGs). Firstly, we introduce a multi-scale analysis framework for ECG signals, leveraging matrix profile-derived Euclidean distance spectra and a centralized cropping strategy to extract salient features at 1-minute, 3-minute, and 5-minute scales for robust model inputs. Secondly, a deep learning model for the OSA recognition is designed based on multi-scale convolutional neural networks, incorporating the residual attention and the channel attention to address vanishing and exploding gradients, focusing on key features of the ECG signals. Moreover, we design three different convolutional neural network (CNN) modules to process ECG signals with various segment lengths. Experimental results show that the accuracy of the segment-based OSA detection reaches 92.82% and 90.73% on two public datasets, respectively, 1.89% and 7.28% higher than the state-of-the-art models.

Since protein 3D structure prediction is very important for biochemical study and drug design, researchers have developed many machine learning algorithms to predict protein 3D structures using the sequence information only. Understanding the sequence-to-structure relationship is key for the successful structure prediction. Previous approaches including the single shallow learning model, the single deep learning model and clustering algorithms all have disadvantages to understand precise sequence-to-structure relationship. In order to further improve the performance of the local protein structure prediction, a novel deep learning model called Clustering Recurrent Neural Network (CRNN) is proposed. In this model, the whole protein dataset is divided into multiple cluster subtrees. A RNN is trained for each cluster in the subtrees so that each RNN can be used to learn the computationally simpler local sequence-to-structure relationship instead of attempting to capture the global sequence-to-structure relationship. After learning the local sequence-to-structure relationship using RNN, CRNN is designed to predict distance matrices, torsion angles and secondary structures for backbone α-carbon atoms of protein sequence segments. Our experimental analysis indicates that 3D structure prediction accuracy is comparable or better than other state-of-art approaches.

Particle filters are important techniques to support data assimilation for large-scale spatial temporal simulation systems. Distributed particle filters improve the performance of particle filtering by distributing particles to multiple processing units (PUs). While different resampling algorithms have been developed for distributed particle filters, less research has been conducted to investigate how to route particles among the PUs after resampling in effective and efficient manners. This paper develops particle routing policies in distributed particle filters with both the centralized resampling and the distributed resampling. The developed routing policies are evaluated from the aspects of the communication cost and the data assimilation accuracy based on an application of data assimilation for large-scale wildfire spread simulations.