No abstract

Background Intraoperative hypotension has an adverse impact on postoperative outcomes. However, it is difficult to predict and treat intraoperative hypotension in advance according to individual clinical parameters. Objective The aim of this study was to develop a prediction model to forecast 5-minute intraoperative hypotension based on the weighted average ensemble of individual neural networks, utilizing the biosignals recorded during noncardiac surgery. Methods In this retrospective observational study, arterial waveforms were recorded during noncardiac operations performed between August 2016 and December 2019, at Seoul National University Hospital, Seoul, South Korea. We analyzed the arterial waveforms from the big data in the VitalDB repository of electronic health records. We defined 2s hypotension as the moving average of arterial pressure under 65 mmHg for 2 seconds, and intraoperative hypotensive events were defined when the 2s hypotension lasted for at least 60 seconds. We developed an artificial intelligence–enabled process, named short-term event prediction in the operating room (STEP-OP), for predicting short-term intraoperative hypotension. Results The study was performed on 18,813 subjects undergoing noncardiac surgeries. Deep-learning algorithms (convolutional neural network [CNN] and recurrent neural network [RNN]) using raw waveforms as input showed greater area under the precision-recall curve (AUPRC) scores (0.698, 95% CI 0.690-0.705 and 0.706, 95% CI 0.698-0.715, respectively) than that of the logistic regression algorithm (0.673, 95% CI 0.665-0.682). STEP-OP performed better and had greater AUPRC values than those of the RNN and CNN algorithms (0.716, 95% CI 0.708-0.723). Conclusions We developed STEP-OP as a weighted average of deep-learning models. STEP-OP predicts intraoperative hypotension more accurately than the CNN, RNN, and logistic regression models. Trial Registration ClinicalTrials.gov NCT02914444; https://clinicaltrials.gov/ct2/show/NCT02914444.

Android application (app) stores contain a huge number of apps, which are manually classified based on the apps’ descriptions into various categories. However, the predefined categories or apps descriptions are usually not very accurate to reflect the real functionalities of apps, thereby leading to misclassify the apps, which may cause serious security issues and unreliability problem in the app store. Therefore, the automatic app classification is an important demand to construct a secure, reliable, integrated, and easy to navigate app store. In this paper, we propose an effective method called AndroClass to automatically classify apps based on their real functionalities by using rich and comprehensive features representing the actual functionalities of the apps. AndroClass performs three steps of feature extraction, feature refinement, and classification. In the feature extraction step, we extract 14 various features for each app by utilizing a unified tool suite. In the feature refinement step, we apply Random Forest algorithm to refine the features. In the classification step, we combine refined features into a single one and AndroClass is equipped with K-Nearest Neighbor, Naive Bayes, Support Vector Machine, and Deep Neural Network to classify apps. On the contrary to the existing methods, all the utilized features in AndroClass are stable and clearly represent the actual functionalities of the app, AndroClass does not pose any issues to the user privacy, and our method can be applied to classify unreleased or newly released apps. The results of extensive experiments with two real-world datasets and a dataset constructed by human experts demonstrate the effectiveness of AndroClass where the classification accuracy of AndroClass with the latter dataset is 83.5%.

As deep learning techniques advance more than ever, hyper-parameter optimization is the new major workload in deep learning clusters. Although hyper-parameter optimization is crucial in training deep learning models for high model performance, effectively executing such a computation-heavy workload still remains a challenge. We observe that numerous trials issued from existing hyper-parameter optimization algorithms share common hyper-parameter sequence prefixes, which implies that there are redundant computations from training the same hyper-parameter sequence multiple times. We propose a stage-based execution strategy for efficient execution of hyper-parameter optimization algorithms. Our strategy removes redundancy in the training process by splitting the hyper-parameter sequences of trials into homogeneous stages, and generating a tree of stages by merging the common prefixes. Our preliminary experiment results show that applying stage-based execution to hyper-parameter optimization algorithms outperforms the original trial-based method, saving required GPU-hours and end-to-end training time by up to 6.60 times and 4.13 times, respectively.1 The terms study and trial come from Vizier[9].

The rapid evolution of deep neural networks is demanding deep learning (DL) frameworks not only to satisfy the requirement of quickly executing large computations, but also to support straightforward programming models for quickly implementing and experimenting with complex network structures. However, existing frameworks fail to excel in both departments simultaneously, leading to diverged efforts for optimizing performance and improving usability. This paper presents JANUS, a system that combines the advantages from both sides by transparently converting an imperative DL program written in Python, a de-facto scripting language for DL, into an efficiently executable symbolic dataflow graph. JANUS can convert various dynamic features of Python, including dynamic control flow, dynamic types, and impure functions, into elements of a symbolic dataflow graph. Our experiments show that JANUS can achieve fast DL training by exploiting the techniques imposed by symbolic graph-based DL frameworks, while maintaining the simple and flexible programmability of imperative DL frameworks at the same time.