Journal articles: 'Kernel warping'

1

Zhou, Zhengyi, and David S. Matteson. "Predicting Melbourne ambulance demand using kernel warping." Annals of Applied Statistics 10, no. 4 (December 2016): 1977–96. http://dx.doi.org/10.1214/16-aoas961.

Full text

APA, Harvard, Vancouver, ISO, and other styles

2

Pilario, Karl Ezra, Alexander Tielemans, and Elmer-Rico E. Mojica. "Geographical discrimination of propolis using dynamic time warping kernel principal components analysis." Expert Systems with Applications 187 (January 2022): 115938. http://dx.doi.org/10.1016/j.eswa.2021.115938.

Full text

APA, Harvard, Vancouver, ISO, and other styles

3

Mishra, Piyush, and Piyush Lotia. "Speaker Recognition Using Dynamic Time Warping Polynomial Kernel SVM with Confusion Matrix." i-manager's Journal on Computer Science 3, no. 3 (November 15, 2015): 23–27. http://dx.doi.org/10.26634/jcom.3.3.3662.

Full text

APA, Harvard, Vancouver, ISO, and other styles

4

Chen, Zhicheng, Yuequan Bao, Hui Li, and Billie F. Spencer. "A novel distribution regression approach for data loss compensation in structural health monitoring." Structural Health Monitoring 17, no. 6 (December 8, 2017): 1473–90. http://dx.doi.org/10.1177/1475921717745719.

Full text

Abstract:

Structural health monitoring has arisen as an important tool for managing and maintaining civil infrastructure. A critical problem for all structural health monitoring systems is data loss or data corruption due to sensor failure or other malfunctions, which bring into question in subsequent structural health monitoring data analysis and decision-making. Probability density functions play a very important role in many applications for structural health monitoring. This article focuses on data loss compensation for probability density function estimation in structural health monitoring using imputation methods. Different from common data, continuous probability density functions belong to functional data; the conventional distribution-to-distribution regression technique has significant potential in missing probability density function imputation; however, extrapolation and directly borrowing shape information from the covariate probability density function are the main challenges. Inspired by the warping transformation of distributions in the field of functional data analysis, a new distribution regression approach for imputing missing correlated probability density functions is proposed in this article. The warping transformation for distributions is a mapping operation used to transform one probability density function to another by deforming the original probability density function with a warping function. The shape mapping between probability density functions can be characterized well by warping functions. Given a covariate probability density function, the warping function is first estimated by a kernel regression model; then, the estimated warping function is used to transform the covariate probability density function and obtain an imputation for the missing probability density function. To address issues with poor performance when the covariate probability density function is contaminated, a hybrid approach is proposed that fuses the imputations obtained by the warping transformation approach with the conventional distribution-to-distribution regression approach. Experiments based on field monitoring data are conducted to evaluate the performance of the proposed approach. The corresponding results indicate that the proposed approach can outperform the conventional method, especially in extrapolation. The proposed approach shows good potential to provide more reliable estimation of distributions of missing structural health monitoring data.

APA, Harvard, Vancouver, ISO, and other styles

5

Kamycki, Krzysztof, Tomasz Kapuscinski, and Mariusz Oszust. "Data Augmentation with Suboptimal Warping for Time-Series Classification." Sensors 20, no. 1 (December 23, 2019): 98. http://dx.doi.org/10.3390/s20010098.

Full text

Abstract:

In this paper, a novel data augmentation method for time-series classification is proposed. In the introduced method, a new time-series is obtained in warped space between suboptimally aligned input examples of different lengths. Specifically, the alignment is carried out constraining the warping path and reducing its flexibility. It is shown that the resultant synthetic time-series can form new class boundaries and enrich the training dataset. In this work, the comparative evaluation of the proposed augmentation method against related techniques on representative multivariate time-series datasets is presented. The performance of methods is examined using the nearest neighbor classifier with the dynamic time warping (NN-DTW), LogDet divergence-based metric learning with triplet constraints (LDMLT), and the recently introduced time-series cluster kernel (NN-TCK). The impact of the augmentation on the classification performance is investigated, taking into account entire datasets and cases with a small number of training examples. The extensive evaluation reveals that the introduced method outperforms related augmentation algorithms in terms of the obtained classification accuracy.

APA, Harvard, Vancouver, ISO, and other styles

6

Ahmed, Rehan, Andriy Temko, William P. Marnane, Geraldine Boylan, and Gordon Lightbody. "Exploring temporal information in neonatal seizures using a dynamic time warping based SVM kernel." Computers in Biology and Medicine 82 (March 2017): 100–110. http://dx.doi.org/10.1016/j.compbiomed.2017.01.017.

Full text

APA, Harvard, Vancouver, ISO, and other styles

7

Nasonov, A., A. Krylov, and D. Lyukov. "IMAGE SHARPENING WITH BLUR MAP ESTIMATION USING CONVOLUTIONAL NEURAL NETWORK." ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences XLII-2/W12 (May 9, 2019): 161–66. http://dx.doi.org/10.5194/isprs-archives-xlii-2-w12-161-2019.

Full text

Abstract:

<p><strong>Abstract.</strong> We propose a method for choosing optimal values of the parameters of image sharpening algorithm for out-of-focus blur based on grid warping approach. The idea of the considered sharpening algorithm is to move pixels from the edge neighborhood towards the edge centerlines. Compared to traditional deblurring algorithms, this approach requires only scalar blur level value rather than a blur kernel. We propose a convolutional neural network based algorithm for estimating the blur level value.</p>

APA, Harvard, Vancouver, ISO, and other styles

8

Ren, Zhiming, Qianzong Bao, and Bingluo Gu. "Joint wave-equation traveltime inversion of diving/direct and reflected waves for P- and S-wave velocity macromodel building." GEOPHYSICS 86, no. 4 (July 1, 2021): R603—R621. http://dx.doi.org/10.1190/geo2020-0762.1.

Full text

Abstract:

Full-waveform inversion (FWI) suffers from the local minima problem and requires a sufficiently accurate starting model to converge to the correct solution. Wave-equation traveltime inversion (WETI) is an effective tool to retrieve the long-wavelength components of the velocity model. We have developed a joint diving/direct and reflected wave WETI (JDRWETI) method to build P- and S-wave velocity macromodels. We estimate the traveltime shifts of seismic events (diving/direct waves and PP- and PS-reflections) through the dynamic warping scheme and construct a misfit function using the time shifts of diving/direct and reflected waves. We derive the adjoint wave equations and the gradients with respect to the background models based on the joint misfit function. We apply the kernel decomposition scheme to extract the kernel of the diving/direct wave and the tomography kernels of PP- and PS-reflections. For an explosive source, the kernels of the diving/direct wave and PP-reflections and the kernel of the PS-reflections are used to compute the P- and S-wave gradients of the background models, respectively. We implement JDRWETI by a two-stage inversion workflow: First, we invert the P- and S-wave velocity models using the P-wave gradients, and then we improve the S-wave velocity model using the S-wave gradients. Numerical tests on synthetic and field data sets reveal that the JDRWETI method successfully recovers the long-wavelength components of P- and S-wave velocity models, which can be used for an initial model for the subsequent elastic FWI. Moreover, the JDRWETI method prevails over the existing reflection WETI method and the cascaded diving/direct and reflected wave WETI method, especially when large velocity errors are present in the shallow part of the starting models. The JDRWETI method with the two-stage inversion workflow can give rise to reasonable inversion results even for the model with different P- and S-wave velocity structures.

APA, Harvard, Vancouver, ISO, and other styles

9

Jeong, Young-Seon. "Semiconductor Wafer Defect Classification Using Support Vector Machine with Weighted Dynamic Time Warping Kernel Function." Industrial Engineering & Management Systems 16, no. 3 (September 30, 2017): 420–26. http://dx.doi.org/10.7232/iems.2017.16.3.420.

Full text

APA, Harvard, Vancouver, ISO, and other styles

10

Jeong, Young-Seon, and Raja Jayaraman. "Support vector-based algorithms with weighted dynamic time warping kernel function for time series classification." Knowledge-Based Systems 75 (February 2015): 184–91. http://dx.doi.org/10.1016/j.knosys.2014.12.003.

Full text

APA, Harvard, Vancouver, ISO, and other styles

11

Pierri, Rocco, and Raffaele Moretta. "Asymptotic Study of the Radiation Operator for the Strip Current in Near Zone." Electronics 9, no. 6 (May 29, 2020): 911. http://dx.doi.org/10.3390/electronics9060911.

Full text

Abstract:

In this paper, we address the problem of how to efficiently sample the radiated field in the framework of near-field measurement techniques. In particular, the aim of the article is to find a sampling strategy for which the discretized model exhibits the same singular values of the continuous problem. The study is done with reference to a strip current whose radiated electric field is observed in the near zone over a bounded line parallel to the source. Differently from far zone configurations, the kernel of the related eigenvalue problem is not of convolution type, and not band-limited. Hence, the sampling-theory approach cannot be directly applied to establish how to efficiently collect the data. In order to surmount this drawback, we first use an asymptotic approach to explicit the kernel of the eigenvalue problem. After, by exploiting a warping technique, we recast the original eigenvalue problem in a new one. The latter, if the observation domain is not too large, involves a convolution operator with a band-limited kernel. Hence, in this case the sampling-theory approach can be applied, and the optimal locations of the sampling points can be found. Differently, if the observation domain is very extended, the kernel of the new eigenvalue problem is still not convolution. In this last case, in order to establish how to discretize the continuous model, we perform a numerical analysis.

APA, Harvard, Vancouver, ISO, and other styles

12

Maisto, Maria, Rocco Pierri, and Raffaele Solimene. "Near-Field Warping Sampling Scheme for Broad-Side Antenna Characterization." Electronics 9, no. 6 (June 24, 2020): 1047. http://dx.doi.org/10.3390/electronics9061047.

Full text

Abstract:

In this paper the problem of sampling the field radiated by a planar source observed over a finite planar aperture located in the near-field is addressed. The problem is cast as the determination of the spatial measurement positions which allow us to discretize the radiation problem so that the singular values of the radiation operator are well-approximated. More in detail, thanks to a suitably warping transformation of the observation variables, the kernel function of the relevant operator is approximated by a band-limited function and hence the sampling theorem applied to achieved discretization. It results in the sampling points having to be non-linearity arranged across the measurement aperture and their number can be considerably lowered as compared to more standard sampling approach. It is shown that the proposed sampling scheme works well for measurement apertures that are not too large as compared to the source’s size. As a consequence, the method appears better suited for broad-side large antenna whose radiated field is mainly concentrated in front of the antenna. A numerical analysis is included to check the theoretical findings and to study the trade-off between the field accuracy representation (over the measurement aperture) and the truncation error in the estimated far-field radiation pattern.

APA, Harvard, Vancouver, ISO, and other styles

13

Ding, Ing-Jr, and Yen-Ming Hsu. "An HMM-Like Dynamic Time Warping Scheme for Automatic Speech Recognition." Mathematical Problems in Engineering 2014 (2014): 1–8. http://dx.doi.org/10.1155/2014/898729.

Full text

Abstract:

In the past, the kernel of automatic speech recognition (ASR) is dynamic time warping (DTW), which is feature-based template matching and belongs to the category technique of dynamic programming (DP). Although DTW is an early developed ASR technique, DTW has been popular in lots of applications. DTW is playing an important role for the known Kinect-based gesture recognition application now. This paper proposed an intelligent speech recognition system using an improved DTW approach for multimedia and home automation services. The improved DTW presented in this work, called HMM-like DTW, is essentially a hidden Markov model- (HMM-) like method where the concept of the typical HMM statistical model is brought into the design of DTW. The developed HMM-like DTW method, transforming feature-based DTW recognition into model-based DTW recognition, will be able to behave as the HMM recognition technique and therefore proposed HMM-like DTW with the HMM-like recognition model will have the capability to further perform model adaptation (also known as speaker adaptation). A series of experimental results in home automation-based multimedia access service environments demonstrated the superiority and effectiveness of the developed smart speech recognition system by HMM-like DTW.

APA, Harvard, Vancouver, ISO, and other styles

14

Candelieri, Antonio, Stanislav Fedorov, and Enza Messina. "Efficient Kernel-Based Subsequence Search for Enabling Health Monitoring Services in IoT-Based Home Setting." Sensors 19, no. 23 (November 27, 2019): 5192. http://dx.doi.org/10.3390/s19235192.

Full text

Abstract:

This paper presents an efficient approach for subsequence search in data streams. The problem consists of identifying coherent repetitions of a given reference time-series, also in the multivariate case, within a longer data stream. The most widely adopted metric to address this problem is Dynamic Time Warping (DTW), but its computational complexity is a well-known issue. In this paper, we present an approach aimed at learning a kernel approximating DTW for efficiently analyzing streaming data collected from wearable sensors, while reducing the burden of DTW computation. Contrary to kernel, DTW allows for comparing two time-series with different length. To enable the use of kernel for comparing two time-series with different length, a feature embedding is required in order to obtain a fixed length vector representation. Each vector component is the DTW between the given time-series and a set of “basis” series, randomly chosen. The approach has been validated on two benchmark datasets and on a real-life application for supporting self-rehabilitation in elderly subjects has been addressed. A comparison with traditional DTW implementations and other state-of-the-art algorithms is provided: results show a slight decrease in accuracy, which is counterbalanced by a significant reduction in computational costs.

APA, Harvard, Vancouver, ISO, and other styles

15

Ho-Chul Shin, Jae Hee Park, and Seong-Dae Kim. "Combination of Warping Robust Elastic Graph Matching and Kernel-Based Projection Discriminant Analysis for Face Recognition." IEEE Transactions on Multimedia 9, no. 6 (October 2007): 1125–36. http://dx.doi.org/10.1109/tmm.2007.898933.

Full text

APA, Harvard, Vancouver, ISO, and other styles

16

Chaovalitwongse, W. A., and P. M. Pardalos. "On the time series support vector machine using dynamic time warping kernel for brain activity classification." Cybernetics and Systems Analysis 44, no. 1 (January 2008): 125–38. http://dx.doi.org/10.1007/s10559-008-0012-y.

Full text

APA, Harvard, Vancouver, ISO, and other styles

17

Huang, Lin, Li Gong, Yutao Chen, Dongliang Li, and Guoqing Zhu. "Trajectory Similarity Matching and Remaining Useful Life Prediction Based on Dynamic Time Warping." Mathematical Problems in Engineering 2022 (October 22, 2022): 1–15. http://dx.doi.org/10.1155/2022/5344461.

Full text

Abstract:

Remaining useful life prediction based on trajectory similarity is a typical example of instance-based learning. Hence, trajectory similarity prediction based on Euclidean distance has the problems of matching and low prediction accuracy. Therefore, an engine remaining useful life (RUL) prediction method based on dynamic time warping (DTW) is proposed. First, aiming at the problem of engine structure complexity and multiple monitoring parameters, the principal component analysis is used to reduce the dimension of multisensor signals. Then, the system performance degradation trajectory is extracted based on kernel regression. After obtaining the degradation trajectory database, the similarity matching of the degradation trajectory is carried out based on DTW. After finding the best matching curve, the RUL can be predicted. Finally, the proposed method is verified by the public aeroengine simulation dataset of NASA, and compared with several representatives and high-precision literature methods based on the same dataset, which verifies the effectiveness of the method.

APA, Harvard, Vancouver, ISO, and other styles

18

Marteau, Pierre-Francois. "Times Series Averaging and Denoising from a Probabilistic Perspective on Time–Elastic Kernels." International Journal of Applied Mathematics and Computer Science 29, no. 2 (June 1, 2019): 375–92. http://dx.doi.org/10.2478/amcs-2019-0028.

Full text

Abstract:

Abstract In the light of regularized dynamic time warping kernels, this paper re-considers the concept of a time elastic centroid for a set of time series. We derive a new algorithm based on a probabilistic interpretation of kernel alignment matrices. This algorithm expresses the averaging process in terms of stochastic alignment automata. It uses an iterative agglomerative heuristic method for averaging the aligned samples, while also averaging the times of their occurrence. By comparing classification accuracies for 45 heterogeneous time series data sets obtained by first nearest centroid/medoid classifiers, we show that (i) centroid-based approaches significantly outperform medoid-based ones, (ii) for the data sets considered, our algorithm, which combines averaging in the sample space and along the time axes, emerges as the most significantly robust model for time-elastic averaging with a promising noise reduction capability. We also demonstrate its benefit in an isolated gesture recognition experiment and its ability to significantly reduce the size of training instance sets. Finally, we highlight its denoising capability using demonstrative synthetic data. Specifically, we show that it is possible to retrieve, from few noisy instances, a signal whose components are scattered in a wide spectral band.

APA, Harvard, Vancouver, ISO, and other styles

19

Moretta, Raffaele, Giovanni Leone, Fortuna Munno, and Rocco Pierri. "Optimal Field Sampling of Arc Sources via Asymptotic Study of the Radiation Operator." Electronics 11, no. 2 (January 14, 2022): 270. http://dx.doi.org/10.3390/electronics11020270.

Full text

Abstract:

In this paper, the question of how to efficiently sample the field radiated by a circumference arc source is addressed. Classical sampling strategies require the acquisition of a redundant number of field measurements that can make the acquisition time prohibitive. For such reason, the paper aims at finding the minimum number of basis functions representing the radiated field with good accuracy and at providing an interpolation formula of the radiated field that exploits a non-redundant number of field samples. To achieve the first task, the number of relevant singular values of the radiation operator is computed by exploiting a weighted adjoint operator. In particular, the kernel of the related eigenvalue problem is first evaluated asymptotically; then, a warping transformation and a proper choice of the weight function are employed to recast such a kernel as a convolution and bandlimited function of sinc type. Finally, the number of significant singular values of the radiation operator is found by invoking the Slepian–Pollak results. The second task is achieved by exploiting a Shannon sampling expansion of the reduced field. The analysis is developed for both the far and the near fields radiated by a 2D scalar arc source observed on a circumference arc.

APA, Harvard, Vancouver, ISO, and other styles

20

Qiu, Kepeng, Jianlin Wang, Rutong Wang, Yongqi Guo, and Liqiang Zhao. "Soft sensor development based on kernel dynamic time warping and a relevant vector machine for unequal-length batch processes." Expert Systems with Applications 182 (November 2021): 115223. http://dx.doi.org/10.1016/j.eswa.2021.115223.

Full text

APA, Harvard, Vancouver, ISO, and other styles

21

Nguyen, An Hung, Thomas Guillemette, Andrew J. Lambert, Mark R. Pickering, and Matthew A. Garratt. "Increasing feasibility of the field-programmable gate array implementation of an iterative image registration using a kernel-warping algorithm." Journal of Electronic Imaging 26, no. 05 (September 15, 2017): 1. http://dx.doi.org/10.1117/1.jei.26.5.053010.

Full text

APA, Harvard, Vancouver, ISO, and other styles

22

Lin, Luotao, Jiaqi Guo, Yitao Li, Saul B. Gelfand, Edward J. Delp, Anindya Bhadra, Elizabeth A. Richards, Erin Hennessy, and Heather A. Eicher-Miller. "The Discovery of Data-Driven Temporal Dietary Patterns and a Validation of Their Description Using Energy and Time Cut-Offs." Nutrients 14, no. 17 (August 24, 2022): 3483. http://dx.doi.org/10.3390/nu14173483.

Full text

Abstract:

Data-driven temporal dietary patterning (TDP) methods were previously developed. The objectives were to create data-driven temporal dietary patterns and assess concurrent validity of energy and time cut-offs describing the data-driven TDPs by determining their relationships to BMI and waist circumference (WC). The first day 24-h dietary recall timing and amounts of energy for 17,915 U.S. adults of the National Health and Nutrition Examination Survey 2007–2016 were used to create clusters representing four TDPs using dynamic time warping and the kernel k-means clustering algorithm. Energy and time cut-offs were extracted from visualization of the data-derived TDPs and then applied to the data to find cut-off-derived TDPs. The strength of TDP relationships with BMI and WC were assessed using adjusted multivariate regression and compared. Both methods showed a cluster, representing a TDP with proportionally equivalent average energy consumed during three eating events/day, associated with significantly lower BMI and WC compared to the other three clusters that had one energy intake peak/day at 13:00, 18:00, and 19:00 (all p < 0.0001). Participant clusters of the methods were highly overlapped (>83%) and showed similar relationships with obesity. Data-driven TDP was validated using descriptive cut-offs and hold promise for obesity interventions and translation to dietary guidance.

APA, Harvard, Vancouver, ISO, and other styles

23

Youssfi Alaoui, Abdessamad, Youness Tabii, Rachid Oulad Haj Thami, Mohamed Daoudi, Stefano Berretti, and Pietro Pala. "Fall Detection of Elderly People Using the Manifold of Positive Semidefinite Matrices." Journal of Imaging 7, no. 7 (July 6, 2021): 109. http://dx.doi.org/10.3390/jimaging7070109.

Full text

Abstract:

Falls are one of the most critical health care risks for elderly people, being, in some adverse circumstances, an indirect cause of death. Furthermore, demographic forecasts for the future show a growing elderly population worldwide. In this context, models for automatic fall detection and prediction are of paramount relevance, especially AI applications that use ambient, sensors or computer vision. In this paper, we present an approach for fall detection using computer vision techniques. Video sequences of a person in a closed environment are used as inputs to our algorithm. In our approach, we first apply the V2V-PoseNet model to detect 2D body skeleton in every frame. Specifically, our approach involves four steps: (1) the body skeleton is detected by V2V-PoseNet in each frame; (2) joints of skeleton are first mapped into the Riemannian manifold of positive semidefinite matrices of fixed-rank 2 to build time-parameterized trajectories; (3) a temporal warping is performed on the trajectories, providing a (dis-)similarity measure between them; (4) finally, a pairwise proximity function SVM is used to classify them into fall or non-fall, incorporating the (dis-)similarity measure into the kernel function. We evaluated our approach on two publicly available datasets URFD and Charfi. The results of the proposed approach are competitive with respect to state-of-the-art methods, while only involving 2D body skeletons.

APA, Harvard, Vancouver, ISO, and other styles

24

Jia, Xuetao, Ying Huang, Yanhua Wang, and Daozong Sun. "Research on water and fertilizer irrigation system of tea plantation." International Journal of Distributed Sensor Networks 15, no. 3 (March 2019): 155014771984018. http://dx.doi.org/10.1177/1550147719840182.

Full text

Abstract:

In this article, water and fertilizer irrigation system of tea plantation was developed to ensure stable yield, quality of tea, and appropriate irrigation and fertilization. Wireless sensor network node and gateway node are designed and deployed in the tea plantation for collecting soil moisture in real time. Each of the sensor nodes was composed of a STM32F103ZET6 microprocessor and a CC2420 transceiver module. Microcontroller S3C2410X is the kernel of hardware platform in the gateway nodes. All data were transmitted by wireless sensor network. The images of the tea leaves are collected by the high-speed camera loaded onto the unmanned aerial vehicle for analysis of the tea deficiency. The monitoring system is mainly used to display the humidity information about each node, the switch status of the water pump, the battery valve, and so on. It is convenient for user to manage the tea plantation. System configuration and parameter modification can be configured through the monitoring system. The dynamic time warping algorithm is used to judge the abnormal situation of the system. Diamond deployment is used in the network, and the networking experiments were conducted comparing with random deployment. Result showed that both time delay and congestion of network increased as the network scale varied from 5, 10, and 15 to 20 nodes. The topology stabilization time gets prolonged simultaneously. Packet loss rate decreased while data transmission interval varied from 10, 20, and 30 to 40 s. Packet loss rate values in diamond deployment are lower than those in random deployment. In order to improve the accuracy of tea deficiency detection, data fusion technology was adopted. A block-based histogram is employed and similarity distance is adopted to confirm the diagnosis of absence of one or more nutrients of tea. The image information and the spectral information are acquired. The principal component factor is extracted and input into the back propagation neural network to judge the quality of the tea. The number of principal component factors of image information and spectral information is set to six and three; the overall recognition rate reached 97.8%. Therefore, the level of the system abnormality can be determined by the dynamic time warping distance. Test results indicate that it is possible to accurately determine whether tea is deficient and accurately determine whether the system data acquisition is abnormal.

APA, Harvard, Vancouver, ISO, and other styles

25

Lin, Luotao, Marah Aqeel, Jiaqi Guo, Saul Gelfand, Edward Delp, Anindya Bhadra, Elizabeth Richards, Erin Hennessy, and Heather Eicher-Miller. "Joint Temporal Dietary and Physical Activity Patterns: Associations with Health Status Indicators and Chronic Diseases." Current Developments in Nutrition 4, Supplement_2 (May 29, 2020): 590. http://dx.doi.org/10.1093/cdn/nzaa047_010.

Full text

Abstract:

Abstract Objectives The aims of this study are to (1) create clusters of joint temporal patterns of diet and physical activity (PA), and (2) determine the association of these clusters with health status indicators including body mass index (BMI), waist circumference (WC), fasting plasma glucose, hemoglobin A1c, triglyceride, high-density lipoprotein cholesterol, total cholesterol, blood pressure and health outcomes including obesity, Type 2 Diabetes (T2DM) and metabolic syndrome (MS) in U.S. adults 20–65 years. Methods A random day of PA from accelerometry data and the first day 24 hour dietary recall collected in the National Health and Nutrition Examination Survey 2003–2006 were used to determine absolute PA intensity, absolute energy intake, and the time of these activities. Dietary and PA data from 1,627 U.S. adults were Z-normalized. Dynamic time warping (DTW) coupled with kernel-k means clustering algorithm was used to develop joint temporal dietary and PA patterns that maximally partition individuals with similar temporal behaviors into mutually exclusive clusters derived from the data rather than predefined standards. Multivariate regression models adjusted for ptential confounders, multiple comparisons and survey design factors determined associations between joint temporal patterns and health status indicators along with health outcomes (P < 0.05/6). Results Significant differences of health status indicators and health outcomes were discovered among four clusters. A cluster, representing a joint temporal dietary and PA pattern with proportionally equivalent average energy consumed with two energy intake peaks at 1 PM and 8 PM and the lowest PA intensity compared to all other clusters, was associated with significantly higher BMI (β: 3.5, P < 0.0001), WC (β: 9.5, P < 0.0001), and significantly higher odds of obesity (Odds Ratio = 4.685, P < 0.0001) compared to a cluster with similar energy and intake peaks and the highest PA intensity compared to all other clusters. Conclusions The joint temporal dietary and PA patterns discovered support previous evidence of the link of energy intake and PA on health outcomes. DTW coupled with kernel-k means clustering algorithm can be used to capture differences in temporal dietary and PA behaviors and hold promise for the future development of lifestyle patterns. Funding Sources National Cancer Institute & Purdue University

APA, Harvard, Vancouver, ISO, and other styles

26

Belkhouja, Taha, Yan Yan, and Janardhan Rao Doppa. "Training Robust Deep Models for Time-Series Domain: Novel Algorithms and Theoretical Analysis." Proceedings of the AAAI Conference on Artificial Intelligence 36, no. 6 (June 28, 2022): 6055–63. http://dx.doi.org/10.1609/aaai.v36i6.20552.

Full text

Abstract:

Despite the success of deep neural networks (DNNs) for real-world applications over time-series data such as mobile health, little is known about how to train robust DNNs for time-series domain due to its unique characteristics compared to images and text data. In this paper, we fill this gap by proposing a novel algorithmic framework referred as RObust Training for Time-Series (RO-TS) to create robust deep models for time-series classification tasks. Specifically, we formulate a min-max optimization problem over the model parameters by explicitly reasoning about the robustness criteria in terms of additive perturbations to time-series inputs measured by the global alignment kernel (GAK) based distance. We also show the generality and advantages of our formulation using the summation structure over time-series alignments by relating both GAK and dynamic time warping (DTW). This problem is an instance of a family of compositional min-max optimization problems, which are challenging and open with unclear theoretical guarantee. We propose a principled stochastic compositional alternating gradient descent ascent (SCAGDA) algorithm for this family of optimization problems. Unlike traditional methods for time-series that require approximate computation of distance measures, SCAGDA approximates the GAK based distance on-the-fly using a moving average approach. We theoretically analyze the convergence rate of SCAGDA and provide strong theoretical support for the estimation of GAK based distance. Our experiments on real-world benchmarks demonstrate that RO-TS creates more robust deep models when compared to adversarial training using prior methods that rely on data augmentation or new definitions of loss functions. We also demonstrate the importance of GAK for time-series data over the Euclidean distance.

APA, Harvard, Vancouver, ISO, and other styles

27

Lozano, Antonio, Cristina Soto-Sánchez, Javier Garrigós, J. Javier Martínez, J. Manuel Ferrández, and Eduardo Fernández. "A 3D Convolutional Neural Network to Model Retinal Ganglion Cell’s Responses to Light Patterns in Mice." International Journal of Neural Systems 28, no. 10 (December 2018): 1850043. http://dx.doi.org/10.1142/s0129065718500430.

Full text

Abstract:

Deep Learning offers flexible powerful tools that have advanced our understanding of the neural coding of neurosensory systems. In this work, a 3D Convolutional Neural Network (3D CNN) is used to mimic the behavior of a population of mice retinal ganglion cells in response to different light patterns. For this purpose, we projected homogeneous RGB flashes and checkerboards stimuli with variable luminances and wavelength spectrum to mimic a more naturalistic stimuli environment onto the mouse retina. We also used white moving bars in order to localize the spatial position of the recorded cells. Then recorded spikes were smoothed with a Gaussian kernel and used as the output target when training a 3D CNN in a supervised way. To find a suitable model, two hyperparameter search stages were performed. In the first stage, a trial and error process allowed us to obtain a system that is able to fit the neurons firing rates. In the second stage, a systematic procedure was used to compare several gradient-based optimizers, loss functions and the model’s convolutional layers number. We found that a three layered 3D CNN was able to predict the ganglion cells firing rates with high correlations and low prediction error, as measured with Mean Squared Error and Dynamic Time Warping in test sets. These models were either competitive or outperformed other models used already in neuroscience, as Feed Forward Neural Networks and Linear-Nonlinear models. This methodology allowed us to capture the temporal dynamic response patterns in a robust way, even for neurons with high trial-to-trial variable spontaneous firing rates, when providing the peristimulus time histogram as an output to our model.

APA, Harvard, Vancouver, ISO, and other styles

28

Aqeel, Marah M., Jiaqi Guo, Luotao Lin, Saul B. Gelfand, Edward J. Delp, Anindya Bhadra, Elizabeth A. Richards, Erin Hennessy, and Heather A. Eicher-Miller. "Temporal Dietary Patterns Are Associated with Obesity in US Adults." Journal of Nutrition 150, no. 12 (October 22, 2020): 3259–68. http://dx.doi.org/10.1093/jn/nxaa287.

Full text

Abstract:

ABSTRACT Background The integration of time with dietary patterns throughout a day, or temporal dietary patterns (TDPs), have been linked with dietary quality but relations to health are unknown. Objective The association between TDPs and selected health status indicators and obesity, type 2 diabetes (T2D), and metabolic syndrome (MetS) was determined. Methods The first-day 24-h dietary recall from 1627 nonpregnant US adult participants aged 20–65 y from the NHANES 2003–2006 was used to determine timing, amount of energy intake, and sequence of eating occasions (EOs). Modified dynamic time warping (MDTW) and kernel k-means algorithm clustered participants into 4 groups representing distinct TDPs. Multivariate regression models determined associations between TDPs and health status, controlling for potential confounders, and adjusting for the survey design and multiple comparisons (P <0.05/6). Results A cluster representing a TDP with evenly spaced, energy balanced EOs reaching ≤1200 kcal between 06:00 to 10:00, 12:00 to 15:00, and 18:00 to 22:00, had statistically significant and clinically meaningful lower mean BMI (P <0.0001), waist circumference (WC) (P <0.0001), and 75% lower odds of obesity compared with 3 other clusters representing patterns with much higher peaks of energy: 1000–2400 kcal between 15:00 and 18:00 (OR: 5.3; 95% CI: 2.8, 10.1), 800–2400 kcal between 11:00 and 15:00 (OR: 4.4; 95% CI: 2.5, 7.9), and 1000–2600 kcal between 18:00 and 23:00 (OR: 6.7; 95% CI: 3.9, 11.6). Conclusions Individuals with a TDP characterized by evenly spaced, energy balanced EOs had significantly lower mean BMI, WC, and odds of obesity compared with the other patterns with higher energy intake peaks at different times throughout the day, providing evidence that incorporating time with other aspects of a dietary pattern may be important to health status.

APA, Harvard, Vancouver, ISO, and other styles

29

Lin, Luotao, Jiaqi Guo, Marah Aqeel, Anindya Bhadra, Saul Gelfand, Edward Delp, Elizabeth Richards, Erin Hennessy, and Heather Eicher-Miller. "Temporal Patterning Integrating Diet and Physical Activity Shows Stronger Links to Health Indicators Compared to Patterning of Either Diet or Physical Activity Alone." Current Developments in Nutrition 5, Supplement_2 (June 2021): 469. http://dx.doi.org/10.1093/cdn/nzab039_005.

Full text

Abstract:

Abstract Objectives Daily temporal patterns of energy intake (temporal dietary patterns, TDPs) and physical activity (temporal PA patterns, TPAPs) have been independently and jointly (joint temporal dietary and PA patterns, TDPAPs) associated with health indicators. The strength of the association between clusters of each pattern and health indicators including body mass index (BMI), waist circumference (WC), fasting plasma glucose (FPG), hemoglobin A1c (A1c), triglyceride (TAG), high-density lipoprotein cholesterol (HDL-C), total cholesterol (Total-C), blood pressure, type 2 diabetes (T2D), metabolic syndrome (MetS), and obesity, were compared. Methods The reported energy throughout a day from one reliable 24-hour weekday dietary recall and activity counts from a random weekday of PA accelerometer data of 1,836 U.S. adults from the National Health and Nutrition Examination Survey (2003–2006) were used to create TDP and TPAP respectively, and jointly for TDPAP. Constrained dynamic time warping distances computed over the time series were partitioned to four clusters using kernel-k means clustering algorithm. Measured BMI, WC, FPG, A1c, TAG, HDL-C, Total-C, and classified T2DM, MetS, and obesity were outcomes in multivariate regression models to determine associations with the clusters representing each pattern, controlling for potential confounders and adjusting for multiple comparisons (P < 0.05/6). Adjusted R2 and Akaike information criterion (AIC) compared the strength of the associations between clusters and continuous or categorical health indicators. Results All temporal patterns were significantly associated with BMI, WC, and obesity. Adjusted R2 of BMI and WC models for significant predictors’ effects were higher for TDPAPs (0.129 and 0.194) than TDPs (0.117 and 0.186) or TPAPs (0.077 and 0.143), and AIC of obesity for the TDPAPs (234,752,082) was smaller than for TDPs (236,650,170) or TPAPs (239,810,423). Conclusions TDPAPs incorporating time of day with energy intake and PA had the strongest associations with BMI, WC, and obesity compared with either independent temporal dietary or PA patterns. Patterns representing the integration of multiple daily behavioral habits hold promise for early detection of obesity. Funding Sources NIH (R21CA224764) and Purdue University.

APA, Harvard, Vancouver, ISO, and other styles

30

Lin, Luotao, Jiaqi Guo, Marah M. Aqeel, Saul B. Gelfand, Edward J. Delp, Anindya Bhadra, Elizabeth A. Richards, Erin Hennessy, and Heather A. Eicher-Miller. "Joint temporal dietary and physical activity patterns: associations with health status indicators and chronic diseases." American Journal of Clinical Nutrition 115, no. 2 (October 7, 2021): 456–70. http://dx.doi.org/10.1093/ajcn/nqab339.

Full text

Abstract:

ABSTRACT Background Diet and physical activity (PA) are independent risk factors for obesity and chronic diseases including type 2 diabetes mellitus (T2DM) and metabolic syndrome (MetS). The temporal sequence of these exposures may be used to create patterns with relations to health status indicators. Objectives The objectives were to create clusters of joint temporal dietary and PA patterns (JTDPAPs) and to determine their association with health status indicators including BMI, waist circumference (WC), fasting plasma glucose, glycated hemoglobin, triglycerides, HDL cholesterol, total cholesterol, blood pressure, and disease status including obesity, T2DM, and MetS in US adults. Methods A 24-h dietary recall and random day of accelerometer data of 1836 participants from the cross-sectional NHANES 2003–2006 data were used to create JTDPAP clusters by constrained dynamic time warping, coupled with a kernel k-means clustering algorithm. Multivariate regression models determined associations between the 4 JTDPAP clusters and health and disease status indicators, controlling for potential confounders and adjusting for multiple comparisons. Results A JTDPAP cluster with proportionally equivalent energy consumed at 2 main eating occasions reaching ≤1600 and ≤2200 kcal from 11:00 to 13:00 and from 17:00 to 20:00, respectively, and the highest PA counts among 4 clusters from 08:00 to 20:00, was associated with significantly lower BMI (P < 0.0001), WC (P = 0.0001), total cholesterol (P = 0.02), and odds of obesity (OR: 0.2; 95% CI: 0.1, 0.5) than a JTDPAP cluster with proportionally equivalent energy consumed reaching ≤1600 and ≤1800 kcal from 11:00 to 14:00 and from 17:00 to 21:00, respectively, and high PA counts from 09:00 to 12:00. Conclusions The joint temporally patterned sequence of diet and PA can be used to cluster individuals with meaningful associations to BMI, WC, total cholesterol, and obesity. Temporal patterns hold promise for future development of lifestyle patterns that integrate additional temporal and contextual activities.

APA, Harvard, Vancouver, ISO, and other styles

31

Lin, Luotao, Jiaqi Guo, Yitao Li, Saul Gelfand, Edward Delp, Anindya Bhadra, Elizabeth Richards, Erin Hennessy, and Heathero Eicher-Miller. "The Discovery of Data-Driven Temporal Dietary Patterns and a Validation of Their Description Using Energy and Time Cut-Offs." Current Developments in Nutrition 6, Supplement_1 (June 2022): 377. http://dx.doi.org/10.1093/cdn/nzac054.032.

Full text

Abstract:

Abstract Objectives Data-driven methods were recently applied to create temporal dietary patterns (TDPs) incorporating timing and amount of energy intake over 24-hours; their relationships to obesity were determined. However, description of the data-driven TDPs using energy and time cut-offs were not validated against obesity. Aims were to (1) create data-driven TDPs, describe pattern characteristics using energy and time cut-offs, and determine relationships to BMI and waist circumference (WC); (2) assess the concurrent validity of TDPs derived using the cut-offs by determining relationships with BMI and WC. Methods Amount and timing of energy intake from the first day 24-hour dietary recall of 17,916 U.S. adults in NHANES 2007–2016 was used to pattern 4 TDPs. Clusters were created using data-driven methods: dynamic time warping coupled with kernel-k means clustering algorithm. Relationships with BMI and WC were assessed using multivariate regression. Heat maps plotting the cluster proportion by energy amount throughout the day were used to visualize the data and find energy and time cut-offs for mutually exclusive clusters. Next, the cut-off-based descriptions were used to create new clusters and multivariate regression determined their associations with BMI and WC. Strength to predict obesity was evaluated by comparing both inferential model results. Percent of participant overlap between data-driven and cut-off derived clusters was also calculated. Results Both cut-off and data-driven methods showed a cluster, representing a TDP with proportionally equivalent average energy consumed during three eating events throughout a day, was associated with significantly lower BMI (R2 = 0.12 for both methods) and WC (R2 = 0.17 for both methods) compared to the other 3 clusters that had one energy peak throughout a day (all P < 0.0001). Participant membership of ≥ 82% overlapped between the cut-off and data-driven TDP clusters. Conclusions Four cut-off derived clusters highly overlapped with data-driven clusters and showed no differences in strength or pattern relationships with obesity. TDP discovery using a data-driven method was validated through practically interpretable descriptions of energy intake and timing cut-offs. TDPs hold promise for the prediction of obesity and translation to dietary guidance. Funding Sources Clifford B. Kinley Trust, Purdue University.

APA, Harvard, Vancouver, ISO, and other styles

32

Eicher-Miller, Heather, Marah Aqeel, Jiaqi Guo, Saul Gelfand, Edward Delp, Anindya Bhadra, Elizabeth Richards, Erin Hennessy, and Luotao Lin. "Temporal Dietary Patterns Are Associated with Body Mass Index, Waist Circumference and Obesity." Current Developments in Nutrition 4, Supplement_2 (May 29, 2020): 518. http://dx.doi.org/10.1093/cdn/nzaa046_018.

Full text

Abstract:

Abstract Objectives The integration of time with dietary patterns throughout a day, or temporal dietary patterns (TDP), have been linked with dietary quality but links to health outcomes are unknown. TDP were created with the objective to examine their association with health status indicators including body mass index (BMI), waist circumference (WC), fasting plasma glucose, hemoglobin A1c, triglyceride, high-density lipoprotein cholesterol, total cholesterol, blood pressure, and chronic diseases obesity, diabetes, and metabolic syndrome in US adults 20–65 years. Methods The first-day 24-hour dietary recall from 1627 non-pregnant US adult participants of the cross-sectional National Health and Nutrition Examination Survey 2003–2006 was used to determine energy intake (kcal), time of intake (min), and sequence of intake occasions throughout the 24-hour day. Modified dynamic time warping coupled with kernel k-means algorithm, clustered participants into four groups representing distinct TDP. Multivariate regression models determined associations between TDP clusters and all outcomes, controlling for potential confounders, energy misreporting, and adjusting for multiple comparisons and the complex survey design (P < 0.05/6). Results The cluster representing a TDP with proportionally equivalent average energy at three main eating occasions from 8:00 to 23:00 with peaks reaching 175 kcal at 9:00, 13:00, and 19:00, had statistically significant and clinically meaningful lower BMI (P < 0.0001), WC (P < 0.0001) and 75% lower odds of obesity compared to three other clusters representing distinct patterns of much higher average peak energy intake of 500 kcal at 13:00 (odds ratio (OR): 4.4; 95% confidence interval (CI)): 2.5, 7.9), 530 kcal at 18:00 (OR: 5.3; 95% CI: 2.8, 10.1), and 550 kcal at 20:00 (OR: 6.7; 95%CI: 3.9, 11.6). Conclusions A positive association of the TDP of moderate energy intake throughout the day with healthy weight outcomes supports previous findings of higher dietary quality among those with a similar TDP and provides unique evidence that incorporating time with other aspects of a dietary pattern are linked to health. Funding Sources Research reported in this publication was supported by the National Cancer Institute of the National Institutes of Health and Purdue University.

APA, Harvard, Vancouver, ISO, and other styles

33

Eicher-Miller, Heather, Marah Aqeel, Jiaqi Guo, Saul Gelfand, Edward Delp, Anindya Bhadra, Elizabeth Richards, Erin Hennessy, and Luotao Lin. "Temporal Physical Activity Patterns and Association with Health Status Indicators and Chronic Disease." Current Developments in Nutrition 4, Supplement_2 (May 29, 2020): 1166. http://dx.doi.org/10.1093/cdn/nzaa056_013.

Full text

Abstract:

Abstract Objectives Integration of time to physical activity (PA) patterns is novel and may be linked to health. Temporal PA patterns integrating time and PA were created to determine their association with health indicators: body mass index (BMI), waist circumference (WC), fasting plasma glucose, hemoglobin A1c, triglyceride, high-density lipoprotein cholesterol, total cholesterol, blood pressure, and chronic diseases obesity, diabetes, and metabolic syndrome in US adults 20–65 years. Methods Objectively measured PA accelerometry data collected from the National Health and Nutrition Examination Survey 2003–2006 was used to pattern absolute PA intensity and time of activity from 1627 non-pregnant adults with one random valid PA day. Modified dynamic time warping and kernel k-means clustering grouped individuals representing temporal PA patterns. Multivariate regression models controlling for potential confounders and adjusting for multiple comparisons (P < 0.05/6) determined associations between clusters and health outcomes. Results A cluster representing a temporal PA pattern with the highest average intensity peaking at 11:00 and tapering off through the day was associated with lower BMI (P < 0.0001), WC (P < 0.0001), and 65% lower odds of obesity compared to a cluster with lower average PA intensity peaking at 12:00 and tapering off through the day (95%CI: 0.2,0.5) and a cluster with the lowest average PA intensity with no distinct activity peaks (95%CI: 0.2,0.4). Another cluster with a temporal PA pattern with high average PA intensity peaking at 19:00 was associated with lower BMI (P = 0.0003), WC (P = 0.001), and 60% lower odds of obesity compared to a cluster with lower average PA intensity peaking at 12:00 and tapering off through the day (95%CI: 0.2,0.7) and a cluster with the lowest average PA intensity with no distinct peaks (95%CI: 0.2,0.6). Conclusions Temporal PA patterns are associated with differences in US adult health outcomes. Two clusters with the highest level of PA (peak at 11:00 or 19:00) were associated with significantly lower BMI and WC and lower odds of obesity compared to two other clusters with the lowest average PA intensity. Funding Sources Research reported in this publication was supported by the National Cancer Institute of the National Institutes of Health and Purdue University.

APA, Harvard, Vancouver, ISO, and other styles

34

Soares, Edward J., Alexandra J. Clifford, Carolyn D. Brown, Ryan R. Dean, and Amber M. Hupp. "Balancing Resolution with Analysis Time for Biodiesel–Diesel Fuel Separations Using GC, PCA, and the Mahalanobis Distance." Separations 6, no. 2 (May 27, 2019): 28. http://dx.doi.org/10.3390/separations6020028.

Full text

Abstract:

In this work, a statistical metric called the Mahalanobis distance (MD) is used to compare gas chromatography separation conditions. In the two-sample case, the MD computes the distance between the means of the multivariate probability distributions of two groups. Two gas chromatography columns of the same polarity but differing length and film thickness were utilized for the analysis of fatty acid methyl esters in biodiesel fuels. Biodiesel feedstock samples representing classes of canola, coconut, flaxseed, palm kernal, safflower, soy, soyabean, sunflower, tallow, and waste grease were used in our experiments. Data sets measured from each column were aligned with the correlated optimized warping (COW) algorithm prior to principal components analysis (PCA). The PC scores were then used to compute the MD. Differences between the data produced by each column were determined by converting the MD to its corresponding p-value using the F-distribution. The combination of COW parameters that maximized the p-value were determined for each feedstock separately. The results demonstrate that chromatograms from each column could be optimally aligned to minimize the MD derived from the PC-transformed data. The corresponding p-values for each feedstock type indicated that the two column conditions could produce data that were not statistically different. As a result, the slight loss of resolution using a faster column may be acceptable based on the application for which the data are used.

APA, Harvard, Vancouver, ISO, and other styles

35

Badiane, Mourtadha, and Pádraig Cunningham. "An empirical evaluation of kernels for time series." Artificial Intelligence Review, July 27, 2021. http://dx.doi.org/10.1007/s10462-021-10050-y.

Full text

Abstract:

AbstractThere exist a variety of distance measures which operate on time series kernels. The objective of this article is to compare those distance measures in a support vector machine setting. A support vector machine is a state-of-the-art classifier for static (non-time series) datasets and usually outperforms k-Nearest Neighbour, however it is often noted that that 1-NN DTW is a robust baseline for time-series classification. Through a collection of experiments we determine that the most effective distance measure is Dynamic Time Warping and the most effective classifier is kNN. However, a surprising result is that the pairing of kNN and DTW is not the most effective model. Instead we have discovered via experimentation that Dynamic Time Warping paired with the Gaussian Support Vector Machine is the most accurate time series classifier. Finally, with good reason we recommend a slightly inferior (in terms of accuracy) model Time Warp Edit Distance paired with the Gaussian Support Vector Machine as it has a better theoretical basis. We also discuss the reduction in computational cost achieved by using a Support Vector Machine, finding that the Negative Kernel paired with the Dynamic Time Warping distance produces the greatest reduction in computational cost.

APA, Harvard, Vancouver, ISO, and other styles

36

Adamyan, Garik. "Comparison of Model-Free Algorithms For Clustering GARCH Processes." Mathematical Problems of Computer Science 58 (December 1, 2022). http://dx.doi.org/10.51408/1963-0090.

Full text

Abstract:

In this paper, we evaluate several model-free algorithms for clustering time series datasets generated by GARCH processes. In extensive experiments, we generate synthetic datasets in different scenarios. Then, we compare K-Means (for Euclidian and dynamic time warping distance), K-Shape, and Kernel K-Means models with different clustering metrics. Several experiments show that the K-Means model with dynamic time warping distance archives comparably better results. However, the considered models have significant shortcomings in improving the clustering accuracy when the amount of information (the minimum length of the time series) increases, and in performing accurate clustering when data is unbalanced or clusters are overlapping.

APA, Harvard, Vancouver, ISO, and other styles

37

Baek, Jaeseung, Taha J. Alhindi, Young-Seon Jeong, Myong K. Jeong, Seongho Seo, Jongseok Kang, Jaekyung Choi, and Hyunsang Chung. "Real-Time Fire Detection Algorithm Based on Support Vector Machine with Dynamic Time Warping Kernel Function." Fire Technology, January 30, 2021. http://dx.doi.org/10.1007/s10694-020-01062-1.

Full text

APA, Harvard, Vancouver, ISO, and other styles

38

Luo, Jia, Jingying Huang, Jiancheng Ma, and Siyuan Liu. "Application of self-attention conditional deep convolutional generative adversarial networks in the fault diagnosis of planetary gearboxes." Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and Reliability, January 12, 2023, 1748006X2211477. http://dx.doi.org/10.1177/1748006x221147784.

Full text

Abstract:

The Generative Adversarial Network (GAN) can generate samples similar to the original data to solve the problem of fault sample imbalance in planetary gearbox fault diagnosis. Most of models rely heavily on convolution to model the dependencies across feature vectors of vibration signals. However, the characterization ability of convolution operator is limited by the size of convolution kernel and it cannot capture the long-distance dependence in the original data. In this paper, self-attention is introduced into Conditional Deep Convolutional Generative Adversarial Networks (C-DCGAN). In the model, vibration features are dynamically weighted and merged, so that it can adaptively focus “attention” on different times to solve the problem of sample differences caused by time-varying vibration signals. Finally, the proposed method is verified on the planetary gearbox experiment and the quality of the generated signal samples is evaluated with Dynamic Time Warping (DTW) algorithm. The visual experimental results indicated that the proposed model performed better than conditional deep convolutional generative adversarial networks (C-DCGAN) and could accurately diagnose various working states of planetary gearboxes.

APA, Harvard, Vancouver, ISO, and other styles

39

Tompkins, Anthony, and Fabio Ramos. "Fourier Feature Approximations for Periodic Kernels in Time-Series Modelling." Proceedings of the AAAI Conference on Artificial Intelligence 32, no. 1 (April 29, 2018). http://dx.doi.org/10.1609/aaai.v32i1.11696.

Full text

Abstract:

Gaussian Processes (GPs) provide an extremely powerful mechanism to model a variety of problems but incur an O(N3) complexity in the number of data samples. Common approximation methods rely on what are often termed inducing points but still typically incur an O(NM2) complexity in the data and corresponding inducing points. Using Random Fourier Feature (RFF) maps, we overcome this by transforming the problem into a Bayesian Linear Regression formulation upon which we apply a Bayesian Variational treatment that also allows learning the corresponding kernel hyperparameters, likelihood and noise parameters. In this paper we introduce an alternative method using Fourier series to obtain spectral representations of common kernels, in particular for periodic warpings, which surprisingly have a convergent, non-random form using special functions, requiring fewer spectral features to approximate their corresponding kernel to high accuracy. Using this, we can fuse the Random Fourier Feature spectral representations of common kernels with their periodic counterparts to show how they can more effectively and expressively learn patterns in time-series for both interpolation and extrapolation. This method combines robustness, scalability and equally importantly, interpretability through a symbolic declarative grammar that is both functionally and humanly intuitive — a property that is crucial for explainable decision making. Using probabilistic programming and Variational Inference we are able to efficiently optimise over these rich functional representations. We show significantly improved Gram matrix approximation errors, and also demonstrate the method in several time-series problems comparing other commonly used approaches such as recurrent neural networks.

APA, Harvard, Vancouver, ISO, and other styles

40

Bai, Lu, Lixin Cui, Zhihong Zhang, Lixiang Xu, Yue Wang, and Edwin R. Hancock. "Entropic Dynamic Time Warping Kernels for Co-Evolving Financial Time Series Analysis." IEEE Transactions on Neural Networks and Learning Systems, 2021, 1–15. http://dx.doi.org/10.1109/tnnls.2020.3006738.

Full text

APA, Harvard, Vancouver, ISO, and other styles

Journal articles on the topic 'Kernel warping'

Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles