Nonlinear Decomposition Method Research Articles

Machine Learning Mitigants for Speech Based Cyber Risk

Statistical analysis of speech is an emerging area of machine learning. In this paper, we tackle the biometric challenge of Automatic Speaker Verification (ASV) of differentiating between samples generated by two distinct populations of utterances, those of an authentic human voice and those generated by a synthetic one. Solving such an issue through a statistical perspective foresees the definition of a decision rule function and a learning procedure to identify the optimal classifier. Classical state-of-the-art countermeasures rely on strong assumptions such as stationarity or local-stationarity of speech that may be atypical to encounter in practice. We explore in this regard a robust non-linear and non-stationary signal decomposition method known as the Empirical Mode Decomposition combined with the Mel-Frequency Cepstral Coefficients in a novel fashion with a refined classifier technique known as multi-kernel Support Vector machine. We undertake significant real data case studies covering multiple ASV systems using different datasets, including the ASVSpoof 2019 challenge database. The obtained results overwhelmingly demonstrate the significance of our feature extraction and classifier approach versus existing conventional methods in reducing the threat of cyber-attack perpetrated by synthetic voice replication seeking unauthorised access.

A dual‐primal finite element tearing and interconnecting method for nonlinear variational inequalities utilizing linear local problems

AbstractWe propose a novel dual‐primal finite element tearing and interconnecting method for nonlinear variational inequalities. The proposed method is based on a particular Fenchel–Rockafellar dual formulation of the target problem, which yields linear local problems despite the nonlinearity of the target problem. Since local problems are linear, each iteration of the proposed method can be done very efficiently compared with usual nonlinear domain decomposition methods. We prove that the proposed method is linearly convergent with the rate while the convergence rate of relevant existing methods is , where r is proportional to the condition number of the dual operator. The spectrum of the dual operator is analyzed for the cases of two representative variational inequalities in structural mechanics: the obstacle problem and the contact problem. Numerical experiments are conducted in order to support our theoretical results.

Discriminatory effects of gender disparities in improved seed and fertilizer use at the plot-level in Malawi and Tanzania

This article adopts non-linear decomposition methods to evaluate gender disparities in improved seed and fertilizer use in the smallholder farming systems of Malawi and Tanzania. Using rich and representative plot-level data, results point to inequalities in access and use of improved seed and fertilizer between male and female plot managers. The gender gap in modern input use in Malawi is more pronounced than in Tanzania. The decomposition result for Malawi attributes differences in endowment factors as the key contributor to disparities in modern input use except for improved maize. These results suggest improving access to complementary farming resources and agricultural asset wealth among women farmers are important in reducing the agricultural input use gap. Although both endowment and structural factors significantly explained the disparities in modern input use in Tanzania, access to government extension contributed considerably to the non-discriminatory gap. These results suggest the need to enhance current efforts in empowering women farmers with vital resources, education, and gender-sensitive agriculture advisory services. The heterogeneous analysis revealed disparities in modern input use among women groups suggesting that policy intervention should also target vulnerable women groups, particularly single women in Malawi and poorer women in Tanzania. The structural component also underlined social norms and policies that do not address women's farmer needs in crop varieties, such as consumption and processing traits.

Exploring racial and ethnic disparities in medication adherence among Medicare comprehensive medication review recipients

BackgroundThere has been a lack of evidence on whether there are racial and ethnic disparities in medication nonadherence among individuals receiving comprehensive medication review (CMR), a required component of the Medicare Part D medication therapy management (MTM) services. ObjectivesTo explore racial/ethnic disparities in medication nonadherence among older MTM enrollees who received a CMR and to determine how much the identified disparities can be explained by observed characteristics. MethodsThe retrospective study used 100% of the 2017 Medicare claims, including MTM data. Linked Area Health Resources Files provided community characteristics. Nonadherence was defined as proportion of days covered <80%, and was measured for diabetes, hypertension, and hyperlipidemia medications. Racial/ethnic disparities were examined by logistic regressions that included racial/ethnic minority dummy variables. A nonlinear Blinder-Oaxaca decomposition method was applied to decompose the identified disparities. ResultsCompared with non-Hispanic Whites (Whites), Blacks were respectively 39% (odds ratio [OR] = 1.39, 95% confidence interval [CI] = 1.33–1.45), 27% (OR = 1.27, 95% CI = 1.22–1.32), and 43% (OR = 1.43, 95% CI = 1.39–1.47) more likely to be nonadherent to diabetes, hypertension, and hyperlipidemia medications; Hispanics were 20% (OR = 1.20, 95% CI = 1.14–1.27) more likely to be nonadherent to hyperlipidemia medications. The total portion of disparity explained was 13.42%, 7.66%, 14.87%, and 10.69% respectively for disparities in Black-White (B–W) diabetes, B–W hypertension, B–W hyperlipidemia, and Hispanic-White hyperlipidemia. The top three contributors were the proportion of married-couple families, census region, and male gender. ConclusionsA lower level of community affluence and social support, regional variations, and a lower proportion of males in Blacks and Hispanics may contribute to the disparities in medication nonadherence. The large unexplained portion of the disparity attests that nonadherence is a complex issue. The Medicare MTM program needs to implement measures to reduce disparities in medication adherence.

Disparities in Guideline-Concordant Initial Systemic Treatment in Women with HER2-Negative Metastatic Breast Cancer: A SEER-Medicare Analysis

BackgroundData on guideline-concordant initial systemic treatment among women with HER2-negative metastatic breast cancer (MBC) are limited. We determined the proportion of women with HER2-negative MBC who received guideline-concordant treatment and the extent to which independent variables explained differences in guideline-concordant treatment by hormone receptor (HR) status.MethodsWe conducted a retrospective cohort study using the SEER-Medicare database. We included women age >65 years diagnosed with HER2-negative MBC during 2010–2013. We used the National Comprehensive Cancer Network treatment guidelines to determine guideline-concordant initial treatment within the first 6 months of a cancer diagnosis. We conducted a multivariable logistic regression to identify the significant predictors of guideline-concordant treatment and a non-linear decomposition method to examine disparities by HR status.ResultsAmong 1089 eligible women, 72.3% received guideline-concordant treatment. Compared to women who did not receive guideline-concordant treatment, women who received guideline-concordant treatment were more like to be comparatively older (p<0.05), married (p=0.0171), resided in areas with higher proportion of people age ≥25 years with at least four years of college education, and had positive HR status (p<0.0001). Approximately 8% of the disparity in guideline-concordant treatment by HR status was explained by their observed characteristics. Need-related factors explained the highest proportion (66.9%) of the disparity.ConclusionOur findings indicate improvement of care for older women, who are single/divorced, have negative HR status, and who live in area with lower education levels. Unexplained disparities in guideline-concordant treatment by HR status can be attributed to patient preferences for treatment, physician-level factors, and perceptions.

What explains the inequalities in health care utilization between immigrants and non-migrants in Switzerland?

BackgroundInequalities in health care use between immigrants and non-migrants are an important issue in many countries, with potentially negative effects on population health and welfare. The aim of this study is to understand the factors that explain these inequalities in Switzerland, a country with one of the highest percentages of foreign-born population.MethodsUsing health survey data, we compare non-migrants to four immigrant groups, differentiating between first- and second-generation immigrants, and culturally different and similar immigrants. To retrieve the relative contribution of each inequality-associated factor, we apply a non-linear decomposition method and categorize the factors into demographic, socio-economic, health insurance and health status factors.ResultsWe find that non-migrants are more likely to visit a doctor compared to first-generation and culturally different immigrants and are less likely to visit the emergency department. Inequalities in doctor visits are mainly attributed to the explained component, namely to socio-economic factors (such as occupation and income), while inequalities in emergency visits are mainly attributed to the unexplained component. We also find that despite the universal health care coverage in Switzerland systemic barriers might exist.ConclusionsOur results indicate that immigrant-specific policies should be developed in order to improve access to care and efficiently manage patients in the health system.

Machine Learning Mitigants for Speech Based Cyber Risk

Statistical analysis of speech is an emerging area of machine learning. In this paper, we tackle the biometric challenge of Automatic Speaker Verification (ASV) of differentiating between samples generated by two distinct populations of utterances, those of an authentic human voice and those generated by a synthetic one. Solving such an issue through a statistical perspective foresees the definition of a decision rule function and a learning procedure to identify the optimal classifier. Classical state-of-the-art countermeasures rely on strong assumptions such as stationarity or local-stationarity of speech that may be atypical to encounter in practice. We explore in this regard a robust non-linear and non-stationary signal decomposition method known as the Empirical Mode Decomposition combined with the Mel-Frequency Cepstral Coefficients in a novel fashion with a refined classifier technique known as multi-kernel Support Vector machine. We undertake significant real data case studies covering multiple ASV systems using different datasets, including the ASVSpoof 2019 challenge database. The obtained results overwhelmingly demonstrate the significance of our feature extraction and classifier approach versus existing conventional methods in reducing the threat of cyber-attack perpetrated by synthetic voice replication seeking unauthorised access.

Energy efficiency of nonlinear domain decomposition methods

A nonlinear domain decomposition (DD) solver is considered with respect to improved energy efficiency. In this method, nonlinear problems are solved using Newton’s method on the subdomains in parallel and in asynchronous iterations. The method is compared to the more standard Newton-Krylov approach, where a linear domain decomposition solver is applied to the overall nonlinear problem after linearization using Newton’s method. It is found that in the nonlinear domain decomposition method, making use of the asynchronicity, some processor cores can be set to sleep to save energy and to allow better use of the power and thermal budget. Energy savings on average for each socket up to 77% (due to the RAPL hardware counters) are observed compared to the more traditional Newton-Krylov approach, which is synchronous by design, using up to 5120 Intel Broadwell (Xeon E5-2630v4) cores. The total time to solution is not affected. On the contrary, remaining cores of the same processor may be able to go to turbo mode, thus reducing the total time to solution slightly. Last, we consider the same strategy for the ASPIN (Additive Schwarz Preconditioned Inexact Newton) nonlinear domain decomposition method and observe a similar potential to save energy.

Recent trends in the gender gap in the labor market in the Philippines

We use linear and nonlinear decomposition methods to explore recent trends in male-female differentials in labor market performance in the Philippines. Using the 2018 Labor Force Survey, we calculate an unadjusted wage gap of 5.77 percent, an employment rate gap of 0.36 percentage points, and a labor force participation rate gap of 19 percentage points. We provide further context by showing the evolution of these figures over time (2002-2018), which indicates that women generally perform worse in the labor market than men as measured by the pay gap and the gap in labor force participation ( LFP ). The decomposition exercise reveals that the pay and LFP -rate differentials are largely due to women receiving lower returns to their observable characteristics relative to men. The gender gap in employment status is not significant. JEL classification: J16, J21, J71

Vibration Signal Processing-Based Detection of Short-Circuited Turns in Transformers: A Nonlinear Mode Decomposition Approach

Transformers are vital and indispensable elements in electrical systems, and therefore, their correct operation is fundamental; despite being robust electrical machines, they are susceptible to present different types of faults during their service life. Although there are different faults, the fault of short-circuited turns (SCTs) has attracted the interest of many researchers around the world since the windings in a transformer are one of the most vulnerable parts. In this regard, several works in literature have analyzed the vibration signals that generate a transformer as a source of information to carry out fault diagnosis; however this analysis is not an easy task since the information associated with the fault is embedded in high level noise. This problem becomes more difficult when low levels of fault severity are considered. In this work, as the main contribution, the nonlinear mode decomposition (NMD) method is investigated as a potential signal processing technique to extract features from vibration signals, and thus, detect SCTs in transformers, even in early stages, i.e., low levels of fault severity. Also, the instantaneous root mean square (RMS) value computed using the Hilbert transform is proposed as a fault indicator, demonstrating to be sensitive to fault severity. Finally, a fuzzy logic system is developed for automatic fault diagnosis. To test the proposal, a modified transformer representing diverse levels of SCTs is used. These levels consist of 0 (healthy condition), 5, 10, 15, 20, and 25 SCTs. Results demonstrate the capability of the proposal to extract features from vibration signals and perform automatic fault diagnosis.

Formula omitted]-norm low-rank linear approximation for accelerating deep neural networks

In this paper, we develop a L1-norm based low-rank matrix approximation method to decompose large high-complexity convolution layers into a set of low-complexity convolution layers with low-ranks to accelerate deep neural networks. Based on the alternating direction method (ADM), we derive a mathematical solution for this new L1-norm based low-rank decomposition problem. Our experimental results on public datasets, including CIFAR-10 and ImageNet, demonstrate that this new decomposition scheme outperforms the recently developed L2-norm based nonlinear decomposition method, which achieved the state-of-the-art performance.

Additive and Hybrid Nonlinear Two-Level Schwarz Methods and Energy Minimizing Coarse Spaces for Unstructured Grids

Nonlinear domain decomposition (DD) methods, such as ASPIN (additive Schwarz preconditioned inexact Newton), RASPEN (restricted additive Schwarz preconditioned inexact Newton), nonlinear FETI-DP (f...

Fast Nonlinear Chirplet Dictionary-Based Sparse Decomposition for Rotating Machinery Fault Diagnosis Under Nonstationary Conditions

Fault diagnosis under time-varying operational conditions is always the challenge in industrial systems. The chirplet transform (CT) provides an effective first-order approximation approach, but the slow operation speed and the drawback for high-order signal analysis limit its uses. Focusing on this problem, a fast nonlinear chirplet dictionary-based sparse decomposition (FNC-SD) method for nonlinear signal analysis is proposed. By replacing the time–frequency inclination parameter in the CT with a frequency bending parameter, the proposed method can track the nonstationary signals by arbitrary order polynomial law with time by an additional degree. Considering that too many parameters will slow down the operation speed, a frequency bending operator estimation algorithm is also proposed in this paper based on the calculated rotating frequency. Therefore, a laser vibrometer is needed in the experiment to collect the impulse signals, which can represent rotating speed signals. Then, the FNC-SD method is used to construct a redundant dictionary for signal sparse time–frequency representation. The high similarity between signals and atoms can make the method prefer to choose the right atoms, instead of being influenced by noise. Therefore, the proposed method also shows a strong noise robustness. Two simulation experiments are constructed to verify the performance of the proposed method. Finally, the FNC-SD method is used to diagnose a faulty motor with broken bar. The analyzed results and comparisons with respect to the state-of-the-art methods are illustrated in detail, as well as the convergence accuracy and speed, which highlight the superiority of the proposed method.

Nonlinear mode decomposition with convolutional neural networks for fluid dynamics

We present a new nonlinear mode decomposition method to visualize the decomposed flow fields, named the mode decomposing convolutional neural network autoencoder (MD-CNN-AE). The proposed method is applied to a flow around a circular cylinder at $Re_D=100$ as a test case. The flow attributes are mapped into two modes in the latent space and then these two modes are visualized in the physical space. Because the MD-CNN-AEs with nonlinear activation functions show lower reconstruction errors than the proper orthogonal decomposition (POD), the nonlinearity contained in the activation function is considered the key to improve the capability of the model. It is found by applying POD to each field decomposed using the MD-CNN-AE with hyperbolic tangent activation that a single nonlinear MD-CNN-AE mode contains multiple orthogonal bases, in contrast to the linear methods, i.e., POD and the MD-CNN-AE with linear activation. We further assess the proposed MD-CNN-AE by applying it to a transient process of a circular cylinder wake in order to examine its capability for flows containing high-order spatial modes. The present results suggest a great potential for the nonlinear MD-CNN-AE to be used for feature extraction of flow fields in lower dimension than POD, while retaining interpretable relationships with the conventional POD modes.

Gender differences in the retirement behaviour of the elderly using non-linear decomposition methods

Purpose- This paper examines factors of the gender disparity among British people on their retirement decisions using nonlinear decomposition methods and which factors cause gender differentials in retirement decisions between 1991 and 2013. In this respect, both the British Household Panel Survey and Understanding Society are utilised. Methodology- Both Fairlie (1999) and Yun (2004) methods are used to examine the impacts of observed demographic and income factors on gender differentials of retirement decisions. Findings- Age, hourly earnings, non-labour income and pension eligibility increase the gender gap, whereas education, good health conditions and being a homeowner act to reduce the gender gap. Conclusion- Even though the labour force participation of older women has consistently been increasing over the past two decades, the gender disparity influenced by demographic and financial factors has not been significantly reduced over time. The demographic and financial factors mostly explained the gender gap; however, their effects significantly change due to the explained and unexplained parts during the 20-year period.

Nonlinear response of runoff to atmospheric freezing level height variation based on hybrid prediction models

ABSTRACT We use data on the freezing level height (FLH) and summer runoff in the Hotan River, China, from 1960 to 2013, to analyse the nonlinear relationships of atmospheric and hydrological factors at different time scales, by employing three nonlinear decomposition methods. Six hybrid prediction models are established by combining linear regression and back-propagation artificial neural network (BPANN) models. The decomposition results by three nonlinear methods are compared, indicating that the extreme-point symmetric mode decomposition (ESMD) method ensures the best prediction capacity. The runoff and FLH have periods of 3 and 6 years, respectively, at the inter-annual scale, which pass the significance test of 0.05 (P < 0.05) by using the Monte Carlo method, although there were slight differences in the periods at the inter-decadal scale. Among the six models, ESMD-BPANN exhibits the highest accuracy, with good reliability and resolution, according to several performance indicators. The ESMD-BPANN model is thus selected for the simulation and prediction of runoff.

Ex-ante Inequality of Opportunity in Child Malnutrition: New Evidence from Ten Developing Countries in Asia

This study explores the inequality of opportunity in child malnutrition in ten developing countries in Asia, where a high proportion of children still remain vulnerable to food insecurity. This study takes account of multidimensional aspects of household and parental socio-economic status, and partitions children into distinct types through a data-driven clustering method. This is followed by a comparison of the malnutrition rates between types. Next, we decompose the observed disparity into the factors that are associated with the between-type disparity in malnutrition rates through a non-linear decomposition method. The results indicate that in all 10 countries, significant between-type disparities are found. We find the largest difference in Pakistan as 21.7 percentage points and the smallest difference in Maldives as 5.9 percentage points. In five of the ten countries, the difference in household affluence explains the largest part of the observed between-type disparity. All the results suggest that priority should be given to protecting children from marginalised households in order to mitigate the inequality in child health.

Decomposing the effect of drug benefit program on antihypertensive medication adherence among the elderly in urban China.

PurposeHypertension is a rapidly growing epidemic in People’s Republic of China, yet it remains inadequately controlled. This study aimed to identify the relative contributions of program effects and patients’ characteristics to the differences in antihypertensive medication nonadherence between drug benefit program enrollees and non-enrollees.Patients and methodsData were from a cross-sectional survey of 1,969 community-dwelling elderly adults with hypertension. Self-reported adherence was measured following previous studies in People’s Republic of China. The Blinder-Oaxaca nonlinear decomposition method was used to identify the relative contributions of program effects and patients’ individual characteristics.ResultsEleven percent of the drug benefit program enrollees were nonadherent to their medication, while 17% of non-enrollees were. Blinder-Oaxaca decomposition identified that over 60% of the gap between the two groups was due to the program effects (P=0.024). The rest could be explained by differences in observable characteristics (P<0.001), such as diabetic status, duration of hypertension, and blood pressure control.ConclusionThe study confirmed that drug benefit program enrollees were more likely to be adherent to their antihypertensive medication than non-enrollees in the context of People’s Republic of China.

The Evolution of Immigrants’ Homeownership in Germany

AbstractRecently, the homeownership rate of immigrants in Germany has increased by more than 20 percentage points. To shed light on this sharp rise, this paper investigates the driving forces of the trend in the homeownership rate of immigrant households in Germany between 1996 to 2005 and 2001 to 2011 using a probit-based non-linear decomposition method. Empirical findings suggest that 50 % of the change in immigrants’ homeownership rate within the first time period can be explained by characteristics such as age and educational attainment. In the second time period, the explanatory power of characteristics is almost zero, indicating that it is rather the favorable economic and institutional environment as well as changes in immigrants’ tenure choice process that contributed to the substantial increase in immigrants’ homeownership rate in Germany. We additionally find that housing quality of immigrant homeowners has slightly improved as well, but that there is still a substantial nativity gap in housing quality among tenants as well as among owners.

Solitary living in South Africa: what is driving the pattern and change?

Little is known about solitary living in sub-Saharan Africa, despite the prevalence of evidence in countries like South Africa showing that one-person households (OPHs) are on the rise. This paper examines the determinants of solitary living and the contribution of these factors to the rise in solitary living in South Africa. The analysis was based on 10% samples from three successive censuses of South Africa (1996, 2001, 2011). The multilevel binary logistic regression model was used to identify the determinants and a non-linear multivariate decomposition method was used to identify key contributing factors. The proportion of OPHs increased from 17.8% in 1996 to 27.5% in 2011. Living alone was found to be associated with various important demographic and socio-economic factors. Predominantly, the increase in solitary living in South Africa is attributed to compositional change of characteristics (68.5%) compared to change in the effects of characteristics (31.5%). The main contributing factors were compositional shifts in household ownership, household income level, migration and employment. More research on solitary living is needed to inform policy.