Efficient Learning Algorithm Research Articles

An efficient and accurate distributed learning algorithm for modeling multi-site zero-inflated count outcomes

Clinical research networks (CRNs), made up of multiple healthcare systems each with patient data from several care sites, are beneficial for studying rare outcomes and increasing generalizability of results. While CRNs encourage sharing aggregate data across healthcare systems, individual systems within CRNs often cannot share patient-level data due to privacy regulations, prohibiting multi-site regression which requires an analyst to access all individual patient data pooled together. Meta-analysis is commonly used to model data stored at multiple institutions within a CRN but can result in biased estimation, most notably in rare-event contexts. We present a communication-efficient, privacy-preserving algorithm for modeling multi-site zero-inflated count outcomes within a CRN. Our method, a one-shot distributed algorithm for performing hurdle regression (ODAH), models zero-inflated count data stored in multiple sites without sharing patient-level data across sites, resulting in estimates closely approximating those that would be obtained in a pooled patient-level data analysis. We evaluate our method through extensive simulations and two real-world data applications using electronic health records: examining risk factors associated with pediatric avoidable hospitalization and modeling serious adverse event frequency associated with a colorectal cancer therapy. In simulations, ODAH produced bias less than 0.1% across all settings explored while meta-analysis estimates exhibited bias up to 12.7%, with meta-analysis performing worst in settings with high zero-inflation or low event rates. Across both applied analyses, ODAH estimates had less than 10% bias for 18 of 20 coefficients estimated, while meta-analysis estimates exhibited substantially higher bias. Relative to existing methods for distributed data analysis, ODAH offers a highly accurate, computationally efficient method for modeling multi-site zero-inflated count data.

Distributed Deep Reinforcement Learning-Based Energy and Emission Management Strategy for Hybrid Electric Vehicles

Advanced algorithms can promote the development of energy management strategies (EMSs) as a key technology in hybrid electric vehicles (HEVs). Reinforcement learning (RL) with distributed structure can significantly improve training efficiency in complex environments, and multi-threaded parallel computing provides a reliable algorithm basis for promoting adaptability. Dedicated to trying more efficient deep reinforcement learning (DRL) algorithms, this paper proposed a deep q-network (DQN)-based energy and emission management strategy (E&EMS) at first. Then, two distributed DRL algorithms, namely asynchronous advantage actor-critic (A3C) and distributed proximal policy optimization (DPPO), were adopted to propose EMSs, respectively. Finally, emission optimization was taken into account and then distributed DRL-based E&EMSs were proposed. Regarding dynamic programming (DP) as the optimal benchmark, simulation results show that three DRL-based control strategies can achieve near-optimal fuel economy and outstanding computational efficiency, and compared with DQN, two distributed DRL algorithms have improved the learning efficiency by four times.

Machine learning the nuclear mass

The masses of $$\sim$$ 2500 nuclei have been measured experimentally; however, >7000 isotopes are predicted to exist in the nuclear landscape from H ( $$Z=1$$ ) to Og ( $$Z=118$$ ) based on various theoretical calculations. Exploring the mass of the remaining isotopes is a popular topic in nuclear physics. Machine learning has served as a powerful tool for learning complex representations of big data in many fields. We use Light Gradient Boosting Machine (LightGBM), which is a highly efficient machine learning algorithm, to predict the masses of unknown nuclei and to explore the nuclear landscape on the neutron-rich side from learning the measured nuclear masses. Several characteristic quantities (e.g., mass number and proton number) are fed into the LightGBM algorithm to mimic the patterns of the residual $$\delta (Z,A)$$ between the experimental binding energy and the theoretical one given by the liquid-drop model (LDM), Duflo–Zucker (DZ, also dubbed DZ28) mass model, finite-range droplet model (FRDM, also dubbed FRDM2012), as well as the Weizsäcker–Skyrme (WS4) model to refine these mass models. By using the experimental data of 80 $$\%$$ of known nuclei as the training dataset, the root mean square deviations (RMSDs) between the predicted and the experimental binding energy of the remaining 20% are approximately $$0.234\pm 0.022$$ , $$0.213\pm 0.018$$ , $$0.170\pm 0.011$$ , and $$0.222\pm 0.016$$ MeV for the LightGBM-refined LDM, DZ model, WS4 model, and FRDM, respectively. These values are approximately 90%, 65%, 40%, and 60% smaller than those of the corresponding origin mass models. The RMSD for 66 newly measured nuclei that appeared in AME2020 was also significantly improved. The one-neutron and two-neutron separation energies predicted by these refined models are consistent with several theoretical predictions based on various physical models. In addition, the two-neutron separation energies of several newly measured nuclei (e.g., some isotopes of Ca, Ti, Pm, and Sm) predicted with LightGBM-refined mass models are also in good agreement with the latest experimental data. LightGBM can be used to refine theoretical nuclear mass models and predict the binding energy of unknown nuclei. Moreover, the correlation between the input characteristic quantities and the output can be interpreted by SHapley additive exPlanations (a popular explainable artificial intelligence tool), which may provide new insights for developing theoretical nuclear mass models.

Cognitive multi-modal consistent hashing with flexible semantic transformation

Multi-modal hashing can encode the large-scale social geo-media multimedia data from multiple sources into a common discrete hash space, in which the heterogeneous correlations from multiple modalities could be well explored and preserved into the objective semantic-consistent hash codes. The current researches on multi-modal hashing mainly focus on performing common data reconstruction, but they fail to effectively distill the intrinsic and consensus structures of multi-modal data and fully exploit the inherent semantic knowledge to capture semantic-consistent information across multiple modalities, leading to unsatisfactory retrieval performance. To facilitate this problem and develop an efficient multi-modal geographical retrieval method, in this article, we propose a discriminative multi-modal hashing framework named Cognitive Multi-modal Consistent Hashing (CMCH), which can progressively pursue the structure consensus over heterogeneous multi-modal data and simultaneously explore the informative transformed semantics. Specifically, we construct a parameter-free collaborative multi-modal fusion module to incorporate and excavate the underlying common components from multi-source data. Particularly, our formulation seeks for a joint multi-modal compatibility among multiple modalities under a self-adaptive weighting manner, which can take full advantages of their complementary properties. Moreover, a cognitive self-paced learning policy is further leveraged to conduct progressive feature aggregation, which can coalesce multi-modal data onto the established common latent space in a curriculum learning mode. Furthermore, deep semantic transform learning is elaborated to generate flexible semantics for interactively guiding collaborative hash codes learning. An efficient discrete learning algorithm is devised to address the resulting optimization problem, which obtains stable solutions when dealing with large-scale multi-modal retrieval tasks. Sufficient experiments performed on four large-scale multi-modal datasets demonstrate the encouraging performance of the proposed CMCH method in comparison with the state-of-the-arts over multi-modal information retrieval and computational efficiency. The source codes of this work could be available at https://github.com/JunfengAn1998a/CMCH .

Machine learning is a powerful tool to study the effect of cancer on species and ecosystems

Abstract Cancer is an understudied but important process in wildlife. Cancerous cells are proposed to have had significant effect on the evolution of metazoan species due to their negative effect on host fitness. However, gaining knowledge on the impact of cancer on species and ecosystems is currently relatively slow as it requires expertise in both ecology and oncology. The field can greatly benefit from automation to reduce the need of excessive manpower and analyse complex ecological datasets. In this commentary, we examine how machine learning has been used to gain knowledge on oncogenic processes in wildlife. Using a landscape ecology approach, we explore spatial scales ranging from the size of a molecule up to whole ecosystems and detail, for each level, how machine learning has been used, or could contribute to obtain insights on cancer in wildlife populations and ecosystems. We illustrate how machine learning is a powerful toolbox to conduct studies at the interface of ecology and oncology. We provide guidance for the readers of both fields on how to implement machine learning tools in their research and identify directions to move the field forward using this promising technology. We demonstrate how applying machine learning to complex ecological datasets will (a) contribute to quantitating the effect of cancer at different life stages in wildlife; (b) allow the mining of long‐term datasets to understand the spatiotemporal variability of cancer risk factors and (c) contribute to mitigating cancer risk factors and the conservation of endangered species. With this study, we aim to facilitate the use of machine learning to wildlife species and to encourage discussion between the scientists of the fields of oncology and ecology. We highlight the importance of international and pluridisciplinary collaborations to collect high‐quality datasets on which efficient machine learning algorithms can be trained.

Real Time Face Recognition using Effective Supervised Machine Learning Algorithms

In security authentication systems, face recognition has become an emerging new trend. Modern FR processes can detect whether the individual is real (live) or not through face recognition, thus preventing the systems from being compromised by displaying an actual person’s image. These new systems of face recognition are the result of recent advances in the field of computer vision and efficient algorithms for machine learning. This thesis describes in depth how the identification of a real time face can be order to create a safety alert framework for the workplace, the machine learning algorithm Haarcascade classifier was used to build four distinct classes for identification of security equipment and eventually identify the faces in both images and video using python open CV. Initially, face detection technique is performed to scan the face identity marks through.

Unfolding the multiscale structure of networks with dynamical Ollivier-Ricci curvature

Describing networks geometrically through low-dimensional latent metric spaces has helped design efficient learning algorithms, unveil network symmetries and study dynamical network processes. However, latent space embeddings are limited to specific classes of networks because incompatible metric spaces generally result in information loss. Here, we study arbitrary networks geometrically by defining a dynamic edge curvature measuring the similarity between pairs of dynamical network processes seeded at nearby nodes. We show that the evolution of the curvature distribution exhibits gaps at characteristic timescales indicating bottleneck-edges that limit information spreading. Importantly, curvature gaps are robust to large fluctuations in node degrees, encoding communities until the phase transition of detectability, where spectral and node-clustering methods fail. Using this insight, we derive geometric modularity to find multiscale communities based on deviations from constant network curvature in generative and real-world networks, significantly outperforming most previous methods. Our work suggests using network geometry for studying and controlling the structure of and information spreading on networks.

Flood Inundation Mapping with Limited Observations Based on Physics-Aware Topography Constraint.

Spatial classification with limited observations is important in geographical applications where only a subset of sensors are deployed at certain spots or partial responses are collected in field surveys. For example, in observation-based flood inundation mapping, there is a need to map the full flood extent on geographic terrains based on earth imagery that partially covers a region. Existing research mostly focuses on addressing incomplete or missing data through data cleaning and imputation or modeling missing values as hidden variables in the EM algorithm. These methods, however, assume that missing feature observations are rare and thus are ineffective in problems whereby the vast majority of feature observations are missing. To address this issue, we recently proposed a new approach that incorporates physics-aware structural constraint into the model representation. We design efficient learning and inference algorithms. This paper extends our recent approach by allowing feature values of samples in each class to follow a multi-modal distribution. Evaluations on real-world flood mapping applications show that our approach significantly outperforms baseline methods in classification accuracy, and the multi-modal extension is more robust than our early single-modal version. Computational experiments show that the proposed solution is computationally efficient on large datasets.

Smart Crop Monitoring and Automation Irrigation System Using Machine Learning and IOT

Agriculture is still a major occupation in rural areas of India. Increase in technology creates many opportunities in different fields and attracts human resources from rural areas. Farmers are facing sever human and natural resource problems. Monitoring crops with low man power is the major problem. Smart crop monitoring and automation irrigation system deals with these problems by developing a mobile application which helps farmer to get detailed information about plant diseases and to use irrigation system efficiently. This model uses image processing techniques to identify the picture of the leaf and also provides information about temperature and moisture on field. The Raspberry Pi is the project's control unit, which controls and executes the entire system's operation. Pi camera is placed at the face of the moving vehicle to take the pictures of the leaves. These pictures are analyzed using convolution neural network which is an efficient machine learning algorithm. If the captured leaf image has a disease that is already in the given dataset then farmer will get output message which contains disease cause and pesticide or fertilizers we need to provide to eradicate the disease. Mobile application also sends the data which is sensed using sensors.

Capped [formula omitted]-norm metric based robust least squares twin support vector machine for pattern classification

Least squares twin support vector machine (LSTSVM) is an effective and efficient learning algorithm for pattern classification. However, the distance in LSTSVM is measured by squared L2-norm metric that may magnify the influence of outliers. In this paper, a novel robust least squares twin support vector machine framework is proposed for binary classification, termed as CL2,p-LSTSVM, which utilizes capped L2,p-norm distance metric to reduce the influence of noise and outliers. The goal of CL2,p-LSTSVM is to minimize the capped L2,p-norm intra-class distance dispersion, and eliminate the influence of outliers during training process, where the value of the metric is controlled by the capped parameter, which can ensure better robustness. The proposed metric includes and extends the traditional metrics by setting appropriate values of p and capped parameter. This strategy not only retains the advantages of LSTSVM, but also improves the robustness in solving a binary classification problem with outliers. However, the nonconvexity of metric makes it difficult to optimize. We design an effective iterative algorithm to solve the CL2,p-LSTSVM. In each iteration, two systems of linear equations are solved. Simultaneously, we present some insightful analyses on the computational complexity and convergence of algorithm. Moreover, we extend the CL2,p-LSTSVM to nonlinear classifier and semi-supervised classification. Experiments are conducted on artificial datasets, UCI benchmark datasets, and image datasets to evaluate our method. Under different noise settings and different evaluation criteria, the experiment results show that the CL2,p-LSTSVM has better robustness than state-of-the-art approaches in most cases, which demonstrates the feasibility and effectiveness of the proposed method.

Prediction of the academic performance of slow learners using efficient machine learning algorithm

Maintaining of immense measure of data has always been a great concern. With expansion in awareness towards educational data, the amount of data in the educational institutes is additionally expanded. To deal with increasing growth of data leads to the usage of a new approach of machine learning. Predicting student’s performance before the final examination can help management, faculty, as well as students to make timely decisions and avoid failing of students. In addition to this, the usage of sentimental analysis can gain insight to improve their performance on the student’s next term. We have used various machine learning techniques such as XGboost, K-Nearest Neighbor (K-NN), and Support Vector Machine (SVM) to build predictive models. We have evaluated the performance of these techniques in terms of the performance indicators such as accuracy, precision and recall to determine the better technique that gives accurate results. The evaluation shows that XGBoost is superior in the prediction of poor academic performers than SVM and K-NN with large dataset.

EDiaPredict: An Ensemble-based Framework for Diabetes Prediction

Medical systems incorporate modern computational intelligence in healthcare. Machine learning techniques are applied to predict the onset and reoccurrence of the disease, identify biomarkers for survivability analysis depending upon certain health conditions of the patient. Early prediction of diseases like diabetes is essential as the number of diabetic patients of all age groups is increasing rapidly. To identify underlying reasons for the onset of diabetes in its early stage has become a challenging task for medical practitioners. Continuously increasing diabetic patient data has necessitated for the applications of efficient machine learning algorithms, which learns from the trends of the underlying data and recognizes the critical conditions in patients. In this article, an ensemble-based framework named e DiaPredict is proposed. It uses ensemble modeling, which includes an ensemble of different machine learning algorithms comprising XGBoost, Random Forest, Support Vector Machine, Neural Network, and Decision tree to predict diabetes status among patients. The performance of eDiaPredict has been evaluated using various performance parameters like accuracy, sensitivity, specificity, Gini Index, precision, area under curve, area under convex hull, minimum error rate, and minimum weighted coefficient. The effectiveness of the proposed approach is shown by its application on the PIMA Indian diabetes dataset wherein an accuracy of 95% is achieved.

Learning subtree pattern importance for Weisfeiler-Lehman based graph kernels

Graph is an usual representation of relational data, which are ubiquitous in manydomains such as molecules, biological and social networks. A popular approach to learningwith graph structured data is to make use of graph kernels, which measure the similaritybetween graphs and are plugged into a kernel machine such as a support vector machine.Weisfeiler-Lehman (WL) based graph kernels, which employ WL labeling scheme to extract subtree patterns and perform node embedding, are demonstrated to achieve great performance while being efficiently computable. However, one of the main drawbacks of ageneral kernel is the decoupling of kernel construction and learning process. For moleculargraphs, usual kernels such as WL subtree, based on substructures of the molecules, consider all available substructures having the same importance, which might not be suitable inpractice. In this paper, we propose a method to learn the weights of subtree patterns in the framework of WWL kernels, the state of the art method for graph classification task [14]. To overcome the computational issue on large scale data sets, we present an efficient learning algorithm and also derive a generalization gap bound to show its convergence. Finally, through experiments on synthetic and real-world data sets, we demonstrate the effectiveness of our proposed method for learning the weights of subtree patterns.

Big data driven Internet of Things for credit evaluation and early warning in finance

The big data technology framework has been successfully used in the Internet of Things, and the financial industry also hopes to use the advanced technology of big data to integrate and improve internal and external data related to credit risks. Relying on more efficient machine learning algorithms to get a reasonable prediction of credit risk can reduce the self-generated losses of the Internet of Things finance and increase profits. This article uses distributed search engine technology to customize web crawlers to obtain the required bank card and transaction data from the multi-source heterogeneous data of the Internet of Things financial industry, design the corresponding Spark parallel algorithm to preprocess the data, and establish an inverted table and two Level index file provides data source for big data analysis platform. After the data source is determined, the Mutually Exclusive Collectively Exhaustive (MECE) analysis method is combined with the scores of many financial business experts in the industry to obtain a set of candidate indicators and quantification methods for the financial credit risk evaluation of the Internet of Things, and analyze the correlation of indicators and risk grading. The random forest algorithm in the big data machine learning library is used to select the feature of the candidate index set, and a multi-level spatial association rule algorithm based on the Hash structure is designed to mine the financial risk information of the Internet of Things, and build a credit risk assessment and intelligent early warning model. This paper selects 26 indicators of Internet of Things finance as the research objects, uses SPSS26.0 software to perform sample Kaiser–Meyer–Olkin (KMO) test and Bartlett sphere test on the original data, and describes the results of factor analysis in detail. The particle swarm algorithm is introduced into the parameter optimization of random forest, and the financial credit risk assessment model of the Internet of Things is established. The results show that this method can significantly reduce the probability of banks making the first and second error rates when evaluating the credit risk of financing the Internet of Things Finance. This is conducive to the smooth development of the Internet of Things financial business for banks, which enables banks to enhance their own profitability while effectively reducing losses due to incorrect credit provision.

An integrative scoring system for survival prediction following gefitinib therapy in non-small cell lung cancer: From a long-term real-world study.

e21041 Background: Gefitinib is first-line therapy for patients with EGFR-activating mutation in non-small cell lung cancer (NSCLC). Meanwhile, ̃30% patients were progression within 12 months even harboring EGFR-activating mutation. Furthermore, integrative scoring models for prediction of gefitinib outcomes are still lacking. Variations in genes of DNA repair and cell cycle pathway may have potential impacts on proliferation, migration and invasion of tumor cells. Therefore, we investigated a series of single nucleotide polymorphisms (SNP) in DNA repair and cell cycle genes based on an efficient algorithm machine learning algorithm for establishing an integrative scoring system for survival prediction following gefitinib therapy in NSCLC patients. Methods: A total of 282 patients with activating EGFR mutations were enrolled and divided into train cohort and test cohort, randomly. 125 SNPs in 47 candidate genes were selected by Heploview 4.2 and sequenced by Agena MassARRAY System. Predictive features were selected by randomForestSRC, an efficient algorithm machine learning algorithm to analyze survival data, for PFS following gefitinib therapy. The top candidate variables were presented into COX regression models with hazard ratio as weight for each predictor (HR = 1-1.49 = 1; HR = 1.5-2.49 = 2; HR = 2.5-3.49 = 3), providing p < 0.1. This study was approved by the ethical committee of Sun Yat-Sen University Cancer Center. Results: The progression risk predictive model was established by random forest survival and validated by COX regression analysis, included XRCC1 rs3213263(CT vs CC, RR = 1.997, 95%CI = 1.314-3.036, p = 0.001), FOXO3 rs75544369(GA vs GG, RR = 1.942, 95%CI = 1.139-3.309, p = 0.012), FOXM1 rs2302257(CC vs GG, RR = 2.889, 95%CI = 1.477-5.691, p = 0.002) and ERCC1 rs10408989(TT vs GT, RR = 1.735, 95%CI = 1.152-2.613, p = 0.008; GG vs GT, RR = 3.788, 95%CI = 1.625-8.832, p = 0.002). The progression risk score ranged from 0-10. 49 patients (25.52%) and 20 patients (22.22%) were scored 0, 18(9.38%) and 10(11.11%) patients were scored more than 5 in train cohort and test cohort, respectively. According to the risk score of patients, we divided the patients into four groups, the median PFS were 26.23(95%CI:14.23-35.37), 16.80(95%CI:12.87-23.60), 11.13(95%CI:8.87-16.70) and 6.25 (95% CI:4.43-11.67) months for 0, 2, 3&4 and ≥5 group (p < 0.0001) in the train cohort. Meanwhile, the median PFS were 18.80(95%CI:11.80-64.53),13.60(95%CI:9.17-27.33), 13.47(95%CI:9.83-18.40), 13.47(95%CI:9.83-18.40) and 7.40(95%CI:3.33- NA) months for 0, 2, 3&4 and ≥5 group in validation cohort (p = 0.0017). Conclusions: The risk score system is a simple tool for risk stratification in patients undergoing gefitinib. Widespread application of the score system will require further independent validation.

Communication-Aware Collaborative Learning

Algorithms for noiseless collaborative PAC learning have been analyzed and optimized in recent years with respect to sample complexity. In this paper, we study collaborative PAC learning with the goal of reducing communication cost at essentially no penalty to the sample complexity. We develop communication efficient collaborative PAC learning algorithms using distributed boosting. We then consider the communication cost of collaborative learning in the presence of classification noise. As an intermediate step, we show how collaborative PAC learning algorithms can be adapted to handle classification noise. With this insight, we develop communication efficient algorithms for collaborative PAC learning robust to classification noise.

Improving dynamic stroke risk prediction in non-anticoagulated patients with and without atrial fibrillation: comparing common clinical risk scores and machine learning algorithms.

AimsDiversified cardiovascular/non-cardiovascular multi-morbid risk and efficient machine learning algorithms may facilitate improvements in stroke risk prediction, especially in newly diagnosed non-anticoagulated atrial fibrillation (AF) patients where initial decision-making on stroke prevention is needed. Therefore the aims of this article are to study common clinical risk assessment for stroke risk prediction in AF/non-AF cohorts together with cardiovascular/ non-cardiovascular multi-morbid conditions; to improve stroke risk prediction using machine learning approaches; and to compare the improved clinical prediction rules for multi-morbid conditions using machine learning algorithmsMethods and resultsWe used cohort data from two health plans with 6 457 412 males/females contributing 14,188,679 person-years of data. The model inputs consisted of a diversified list of comorbidities/demographic/ temporal exposure variables, with the outcome capturing stroke event incidences. Machine learning algorithms used two parametric and two nonparametric techniques. The best prediction model was derived on the basis of non-linear formulations using machine learning criteria, with the highest c-index was obtained for logistic regression [0.892; 95% confidence interval (CI) 0.886–0.898] with consistency on external validation (0.891; 95% CI 0.882–0.9). These were significantly higher than those based on the conventional stroke risk scores (CHADS2: 0.7488, 95% CI 0.746–0.7516; CHA2DS2-VASc: 0.7801, 95% CI 0.7772–0.7831) and multi-morbid index (0.8508, 95% CI 0.8483–0.8532). The machine learning algorithm had good internal and external calibration and net benefit values.ConclusionIn this large cohort of newly diagnosed non-anticoagulated AF/non-AF patients, large improvements in stroke risk prediction can be shown with cardiovascular/non-cardiovascular multi-morbid index and a machine learning approach accounting for dynamic changes in risk factors.

Res2-UNeXt: a novel deep learning framework for few-shot cell image segmentation

Recently, developing more accurate and more efficient deep learning algorithms for medical images segmentation attracts more and more attentions of researchers. Most of methods increase the depth of the network to replace with acquiring multi-information. The costs of training images annotation are too expensive to label by hand. In this paper, we propose a multi-scale and better performance deep architecture for medical image segmentation, named Res2-UNeXt. Our architecture is an encoder-decoder network with Res2XBlocks. The Res2XBlocks aim at acquiring multi-scale information better in images. To cooperate with Res2-UNeXt, we put forward a simple and efficient method of data augmentation. The data augmentation method, based on the process of cell movement and deformation, has biological implications in away. We evaluate Res2-UNeXt in comparison with recent variants of U-Net: UNet++, CE-Net and LadderNet and the method that different from U-Net architecture: FCN and DFANet on the dataset of ISBI cell tracking challenge 2019 via four different cell images. The experimental results demonstrate that Res2-UNeXt can achieve better performance than both recent variants of U-Net and non-U-Net architecture methods. Besides, the proposed architecture and the data augmentation method have been proven efficiently by the ablation experiments.

An Artificial Neural Network for Predicting the Near-fault Directivity-pulse Period

ABSTRACT The velocity pulses produced by forward-directivity effects in the near-fault regions can have destructive effects on structures. Proper estimation of the duration of such velocity pulses is an essential step in the near-fault seismic hazard analysis and mitigating potential damage. In this study, the effects of different source, path, source-to-site geometry, and local site parameters on the duration of directivity pulse (T p) are investigated based on the mutual information (MI) concept. A dataset of near-fault pulse-like ground motions from the NGA-West2 database including 135 observations from 17 strike-slip events and 14 non-strike-slip events is utilized for the purpose of this study. The selected ground motion variables are the magnitude, hypocentral distance, depth, D and VS30 that are further applied in an artificial neural network (ANN) to predict T p. The ANN estimates are verified by support vector regression (SVR) as one of the most efficient machine learning algorithms. High correlation between observations and predictions and low error functions reveals the good predictive ability of both ANN and SVR for estimating directivity pulse period. The predictions made by ANN and SVR are further compared with those provided by the empirical and physical models.

An Efficient Learning Algorithm for Direct Training Deep Spiking Neural Networks

It is challenging to train deep spiking neural networks (SNNs) directly due to the difficulties associated with the nondifferentiable neuron model. In this work, an end-to-end learning algorithm based on discrete current-based leaky integrate-and-fire (C-LIF) neuron model and surrogate gradient is proposed to leverage the encoder–decoder network architecture to train deep SNNs directly. The proposed algorithm is capable of learning deep spatiotemporal features relying on current time step only, and several acceleration techniques including backward phase skipping and layerwise FreezeOut are proposed to accelerate the training. Experimental results show that the proposed learning algorithm achieved the classification accuracies of 98.40% and 95.83% on the dynamic neuromorphic data sets MNIST-DVS and DVS-Gestures, respectively, and of 99.58% and 95.97% on static vision data sets MNIST and SVHN, respectively, which are comparable to the existing state-of-the-art results. The training speed was accelerated by up to 49.5% on MNIST and 36.6% on DVS-Gestures with the proposed acceleration techniques while maintaining the same level of accuracy.