Surrogate Rehabilitative Time Series Data for Image-based Deep Learning

Eileen, K. M. Tracey Lee; Kuah, Y.L.; Leo, K. H.; Sanei, Saeid; Chew, Effie; Zhao, Ling

doi:10.23919/eusipco.2019.8903012

Cited by 13 publications

(6 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…That makes these approaches a appropriate candidate for augmentation of time-series. In the experiments of [32], the authors conducted two types of data augmentation by extending the data by 10 then 100 times through AAFT and IAAFT methods, and demonstrated promising classification accuracy improvements compared to the original time series without data augmentation. Furthermore, the same approach was tested on [17], where the usage of IAAFT method yields promising results as timeseries augmentation technique.…”

Section: Surrogate Datamentioning

confidence: 99%

Deep learning enhancing banking services: a hybrid transaction classification and cash flow prediction approach

et al. 2022

View full text Add to dashboard Cite

Small Medium Enterprises (SMEs) are vital to the global economy and all societies. However, they face a complex and challenging environment, as in most sectors they are lagging behind in their digital transformation. Banks, retaining a variety of data of their SME customers to perform their main activities, could offer a solution by leveraging all available data to provide a Business Financial Management (BFM) toolkit to their customers, providing value added services on top of their core business. In this direction, this paper revolves around the development of a smart, highly personalized hybrid transaction categorization model, interconnected with a cash flow prediction model based on Recurrent Neural Networks (RNNs). As the classification of transactions is of great significance, this research is extended towards explainable AI, where LIME and SHAP frameworks are utilized to interpret and illustrate the ML classification results. Our approach shows promising results on a real-world banking use case and acts as the foundation for the development of further BFM banking microservices, such as transaction fraud detection and budget monitoring.

show abstract

Section: Surrogate Datamentioning

confidence: 99%

Deep learning enhancing banking services: a hybrid transaction classification and cash flow prediction approach

et al. 2022

View full text Add to dashboard Cite

show abstract

“…QUANTRAFFIC and CQR require setting aside some data from the model training dataset as the calibration data. If the calibration set is not representative of the test set, the performance of CQR and QUANTRAFFIC may suffer, as shown in Section VII-D. As such, techniques for data augmentation [51] like basic data augmentation methods (e.g., window cropping [52]), Deep Generative Models [53] and data selection [54] like active learning and scoring functions [55] are orthogonal to our approach.…”

Section: Discussionmentioning

confidence: 99%

Robust Design Optimization of the Cogging Torque for a PMSM Based on Manufacturing Uncertainties Analysis and Approximate Modeling

Chen

et al. 2023

Energies

View full text Add to dashboard Cite

A permanent magnet synchronous motor (PMSM) is a crucial device for power conversion in an energy system. The cogging torque of the PMSM is a crucial output characteristic, the robustness of which affects the operational reliability of the energy system. Therefore, the robust design optimization (RDO) of cogging torque has aroused widespread concern. There are several challenges in designing a robust cogging torque PMSM. In particular, some design parameters contain repetitive units, and the finite element analysis (FEA) method is time-consuming. State-of-the-art RDO methods usually treat these uncertainties from repetitive units as the same parameter, which neglects the fluctuation of the manufacturing process and cannot obtain a robust solution for the cogging torque of the motor efficiently and accurately. In order to solve this issue, an approximate modeling method based on manufacturing uncertainties analysis for RDO is proposed in this paper. First, the peak-to-peak value of cogging torque (Tcpp) is used to characterize the cogging torque, which is decoupled to an ideal component and fluctuation component produced by the center values and manufacturing tolerances of design parameters. The design of experiments (DoE) and simulation of the two components are carried out. Then, these two components are approximated separately, and the approximate model of Tcpp is obtained by adding the two components. Finally, the proposed approximate model is embedded into the RDO algorithm, and the PMSM design scheme for good Tcpp robustness is obtained. The effectiveness of the proposed method is verified through a case study of the PMSM.

show abstract

“…However, we notice that no method based on the time-frequency domain or frequency domain alone was used for Challenge 2020. In recent years, data augmentation from these two perspectives has drawn considerable attention in many fields (Lee et al, 2019;Park et al, 2019;Gao et al, 2020), including ECG classification tasks. Moreover, handling the severe class imbalance problem in ECG through data augmentation can be a future research direction.…”

Section: Data Augmentation Should Be Employed and Adapted To Specific...mentioning

confidence: 99%

Practical Lessons on 12-Lead ECG Classification: Meta-Analysis of Methods From PhysioNet/Computing in Cardiology Challenge 2020

et al. 2022

View full text Add to dashboard Cite

Cardiovascular diseases (CVDs) are one of the most fatal disease groups worldwide. Electrocardiogram (ECG) is a widely used tool for automatically detecting cardiac abnormalities, thereby helping to control and manage CVDs. To encourage more multidisciplinary researches, PhysioNet/Computing in Cardiology Challenge 2020 (Challenge 2020) provided a public platform involving multi-center databases and automatic evaluations for ECG classification tasks. As a result, 41 teams successfully submitted their solutions and were qualified for rankings. Although Challenge 2020 was a success, there has been no in-depth methodological meta-analysis of these solutions, making it difficult for researchers to benefit from the solutions and results. In this study, we aim to systematically review the 41 solutions in terms of data processing, feature engineering, model architecture, and training strategy. For each perspective, we visualize and statistically analyze the effectiveness of the common techniques, and discuss the methodological advantages and disadvantages. Finally, we summarize five practical lessons based on the aforementioned analysis: (1) Data augmentation should be employed and adapted to specific scenarios; (2) Combining different features can improve performance; (3) A hybrid design of different types of deep neural networks (DNNs) is better than using a single type; (4) The use of end-to-end architectures should depend on the task being solved; (5) Multiple models are better than one. We expect that our meta-analysis will help accelerate the research related to ECG classification based on machine-learning models.

show abstract

Surrogate Rehabilitative Time Series Data for Image-based Deep Learning

Cited by 13 publications

References 11 publications

Deep learning enhancing banking services: a hybrid transaction classification and cash flow prediction approach

Deep learning enhancing banking services: a hybrid transaction classification and cash flow prediction approach

Robust Design Optimization of the Cogging Torque for a PMSM Based on Manufacturing Uncertainties Analysis and Approximate Modeling

Practical Lessons on 12-Lead ECG Classification: Meta-Analysis of Methods From PhysioNet/Computing in Cardiology Challenge 2020

Contact Info

Product

Resources

About