Subset Of Parameters Research Articles

Replicability of neural responses to speech accent is driven by study design and analytical parameters

Recent studies have reported evidence that listeners' brains process meaning differently in speech with an in-group as compared to an out-group accent. However, among studies that have used electroencephalography (EEG) to examine neural correlates of semantic processing of speech in different accents, the details of findings are often in conflict, potentially reflecting critical variations in experimental design and/or data analysis parameters. To determine which of these factors might be driving inconsistencies in results across studies, we systematically investigate how analysis parameter sets from several of these studies impact results obtained from our own EEG data set. Data were collected from forty-nine monolingual North American English listeners in an event-related potential (ERP) paradigm as they listened to semantically congruent and incongruent sentences spoken in an American accent and an Indian accent. Several key effects of in-group as compared to out-group accent were robust across the range of parameters found in the literature, including more negative scalp-wide responses to incongruence in the N400 range, more positive posterior responses to congruence in the N400 range, and more positive posterior responses to incongruence in the P600 range. These findings, however, are not fully consistent with the reported observations of the studies whose parameters we used, indicating variation in experimental design may be at play. Other reported effects only emerged under a subset of the analytical parameters tested, suggesting that analytical parameters also drive differences. We hope this spurs discussion of analytical parameters and investigation of the contributions of individual study design variables in this growing field.

A Method for Online Identification of a Subset of Synchronous Generator Fundamental Parameters from Monitoring Systems Data

Monitoring systems are common in small and large power plants, providing measurements of basic electrical, mechanical and thermal variables. These measurements can be used for the identification of the synchronous generator parameters. In this paper, a simple method for the identification of a subset of synchronous generator parameters, using basic measurements provided by monitoring systems, is proposed. The identification is performed in normal operating conditions using a synchronous machine simplified model suited for operating conditions under small disturbances. Three identification models are derived, each one depending on a subset of parameters. Three optimization subproblems are then solved by minimizing the residues between each model output and a system measurement. The resulting nonlinear least squares problems are solved by the interior-point method taking into account constraints on the parameters range. A method is proposed to estimate the load angle which is required in the identification. Results are presented for synthetic data and real data, from a 25 MVA generator in a hydroelectric power plant.

Using Reinforcement Learning for Generating Polynomial Models to Explain Complex Data

Basic oxygen steel making is a complex chemical and physical industrial process that reduces a mix of pig iron and recycled scrap into low-carbon steel. Good understanding of the process and the ability to predict how it will evolve requires long operator experience, but this can be augmented with process target prediction systems. Such systems may use machine learning to learn a model of the process based on a long process history, and have an advantage in that they can make use of vastly more process parameters than operators can comprehend. While it has become less of a challenge to build such prediction systems using machine learning algorithms, actual production implementations are rare. The hidden reasoning of complex prediction model and lack of transparency prevents operator trust, even for models that show high accuracy predictions. To express model behaviour and thereby increasing transparency we develop a reinforcement learning (RL) based agent approach, which task is to generate short polynomials that can explain the model of the process from what it has learnt from process data. The RL agent is rewarded on how well it generates polynomials that can predict the process from a smaller subset of the process parameters. Agent training is done with the REINFORCE algorithm, which enables the sampling of multiple concurrently plausible polynomials. Having multiple polynomials, process developers can evaluate several alternative and plausible explanations, as observed in the historic process data. The presented approach gives both a trained generative model and a set of polynomials that can explain the process. The performance of the polynomials is as good as or better than more complex and less interpretable models. Further, the relative simplicity of the resulting polynomials allows good generalisation to fit new instances of data. The best of the resulting polynomials in our evaluation achieves a better R^2 score on the test set in comparison to the other machine learning models evaluated.

Multirate moving horizon estimation combined with parameter subset selection

Due to the model-plant mismatch in practical applications of a nonlinear model, it is necessary to estimate both states and model parameters online. However, when the number of uncertain parameters is large, it is difficult to estimate all the parameters due to a lack of information in measurements. Under this condition, model prediction can be inaccurate although the current states are accurately estimated. To improve the accuracy of both state estimation and prediction, we propose a moving horizon estimation combined with a parameter subset selection scheme. In the proposed MHE framework, a subset of estimable parameters is selected within each horizon. Then, only the selected parameters are estimated along with the state variables. The proposed method is illustrated with the numerical example of a fed-batch bioreactor. The result shows that the proposed method improves the accuracy of model prediction, compared to the conventional MHE, while maintaining the state estimation performance.

SUPER

Feedback from active galactic nuclei (AGN) is thought to be key in shaping the life cycle of their host galaxies by regulating star-formation activity. Therefore, to understand the impact of AGN on star formation, it is essential to trace the molecular gas out of which stars form. In this paper we present the first systematic study of the CO properties of AGN hosts at z ≈ 2 for a sample of 27 X-ray selected AGN spanning two orders of magnitude in AGN bolometric luminosity (log Lbol / erg s−1 = 44.7 − 46.9) by using ALMA Band 3 observations of the CO(3-2) transition (∼1″ angular resolution). To search for evidence of AGN feedback on the CO properties of the host galaxies, we compared our AGN with a sample of inactive (i.e., non-AGN) galaxies from the PHIBSS survey with similar redshift, stellar masses, and star-formation rates (SFRs). We used the same CO transition as a consistent proxy for the gas mass for the two samples in order to avoid systematics involved when assuming conversion factors (e.g., excitation corrections and αCO). By adopting a Bayesian approach to take upper limits into account, we analyzed CO luminosities as a function of stellar masses and SFRs, as well as the ratio LCO(3–2)′/M∗ (a proxy for the gas fraction). The two samples show statistically consistent trends in the LCO(3–2)′−LFIR and LCO(3–2)′−M∗ planes. However, there are indications that AGN feature lower CO(3-2) luminosities (0.4–0.7 dex) than inactive galaxies at the 2–3σ level when we focus on the subset of parameters where the results are better constrained (i.e., LFIR ≈ 1012.2 L⊙ and M* > 1011 M⊙) and on the distribution of the mean log(LCO(3–2)′/M∗). Therefore, even by conservatively assuming the same excitation factor r31, we would find lower molecular gas masses in AGN, and assuming higher r31 would exacerbate this difference. We interpret our result as a hint of the potential effect of AGN activity (such as radiation and outflows), which may be able to heat, excite, dissociate, and/or deplete the gas reservoir of the host galaxies. Better SFR measurements and deeper CO observations for AGN as well as larger and more uniformly selected samples of both AGN and inactive galaxies are required to confirm whether there is a true difference between the two populations.

Linkage factors optimization of multi-outputs of compliant mechanism using response surface

This paper presents a linkage factors synthesis and multi-level optimization technique for bi-stable compliant mechanism. The linkage synthesis problem is modeled as multiple level factors and responses optimization problem with constraints. The bi-stable compliant mechanism is modeled as a crank slider mechanism using pseudo-rigid-body model (PRBM). The model exerts the large deflection of flexible element which explains compliant mechanism’s bi-stable performance. The design concept is applied on variable input parameters subsets. Though the effect of compliant mechanism process factors on Fmax and PRBM deflection angle (Theta-cap Θ1) are contradictory when studied individually as no response gives best process quality. The relationship model between input factors and responses characteristics were generated by ANOVA and optimized by response surface methodology (RSM). ANOVA shown more significant factors are the initial angle of link1 (θ1) and material thickness (t). The Box-Behnken design of RSM is applied with a desirability function approach to determine the optimum set of parameters for minimizing Fmax and maximizing the Theta-cap (Θ1). Thus, this technique shown flexibility based on the product application could be tested and established.

Layerwise learning for quantum neural networks

With the increased focus on quantum circuit learning for near-term applications on quantum devices, in conjunction with unique challenges presented by cost function landscapes of parametrized quantum circuits, strategies for effective training are becoming increasingly important. In order to ameliorate some of these challenges, we investigate a layerwise learning strategy for parametrized quantum circuits. The circuit depth is incrementally grown during optimization, and only subsets of parameters are updated in each training step. We show that when considering sampling noise, this strategy can help avoid the problem of barren plateaus of the error surface due to the low depth of circuits, low number of parameters trained in one step, and larger magnitude of gradients compared to training the full circuit. These properties make our algorithm preferable for execution on noisy intermediate-scale quantum devices. We demonstrate our approach on an image-classification task on handwritten digits, and show that layerwise learning attains an 8% lower generalization error on average in comparison to standard learning schemes for training quantum circuits of the same size. Additionally, the percentage of runs that reach lower test errors is up to 40% larger compared to training the full circuit, which is susceptible to creeping onto a plateau during training.

DES Y1 results: Splitting growth and geometry to test ΛCDM

We analyze Dark Energy Survey (DES) data to constrain a cosmological model where a subset of parameters -- focusing on $\Omega_m$ -- are split into versions associated with structure growth (e.g. $\Omega_m^{\rm grow}$) and expansion history (e.g. $\Omega_m^{\rm geo}$). Once the parameters have been specified for the $\Lambda$CDM cosmological model, which includes general relativity as a theory of gravity, it uniquely predicts the evolution of both geometry (distances) and the growth of structure over cosmic time. Any inconsistency between measurements of geometry and growth could therefore indicate a breakdown of that model. Our growth-geometry split approach therefore serves as both a (largely) model-independent test for beyond-$\Lambda$CDM physics, and as a means to characterize how DES observables provide cosmological information. We analyze the same multi-probe DES data as arXiv:1811.02375 : DES Year 1 (Y1) galaxy clustering and weak lensing, which are sensitive to both growth and geometry, as well as Y1 BAO and Y3 supernovae, which probe geometry. We additionally include external geometric information from BOSS DR12 BAO and a compressed Planck 2015 likelihood, and external growth information from BOSS DR12 RSD. We find no significant disagreement with $\Omega_m^{\rm grow}=\Omega_m^{\rm geo}$. When DES and external data are analyzed separately, degeneracies with neutrino mass and intrinsic alignments limit our ability to measure $\Omega_m^{\rm grow}$, but combining DES with external data allows us to constrain both growth and geometric quantities. We also consider a parameterization where we split both $\Omega_m$ and $w$, but find that even our most constraining data combination is unable to separately constrain $\Omega_m^{\rm grow}$ and $w^{\rm grow}$. Relative to $\Lambda$CDM, splitting growth and geometry weakens bounds on $\sigma_8$ but does not alter constraints on $h$.

Sports and Functional Training Improve a Subset of Obesity-Related Health Parameters in Adolescents: A Randomized Controlled Trial.

To investigate the effects of two different modes of physical activity on body composition, physical fitness, cardiometabolic risk, and psychological responses in female adolescents participating in a multi-disciplinary program. The 12-week randomized intervention included 25-adolescents with overweight divided into two groups: sports practice-SPG and functional training-FTG. The SPG intervention was divided into three sports: basketball, handball, and futsal. SPG participants performed one sport 3-times/week, over the course of 1 month. The FTG performed concurrent exercises 3-times/week. This study was registered in Clinical Trials Registry Platform under number: RBR-45ywtg and registered in Local Ethics Committee number: 2,505.200/2018. The intensity of physical exercises-PE was matched between groups by the rating of perceived exertion. The primary outcome was body composition, and secondary outcomes were physical fitness, cardiometabolic risk, and psychological responses. There was a significant time-effect for body mass, body mass index, and low-density lipoprotein (LDL-c), all being reduced. There were increases over time for musculoskeletal mass, aerobic fitness, and high-density lipoprotein (HDL-c) (p < 0.05). There was a group time interaction with body fat percentage being lower post-intervention in the SPG (p < 0.05). No significant differences were observed for the other variables. Both physical activity models were effective in improving a subset of obesity-related health parameters. The findings should be extended by further investigation using more sophisticated measures of energy expenditure.Clinical Trial Registration: https://ensaiosclinicos.gov.br/, identifier: RBR-45ywtg.

Evaluation of KDP Estimation Algorithm Performance in Rain Using a Known-Truth Framework

AbstractAccurate estimation of specific differential phase (KDP) is necessary for rain rate estimation, attenuation correction, and hydrometeor classification algorithms. There are numerous published methods to process polarimetric radar observations of propagation differential phase shift (ΦDP) and estimate KDP, but the corresponding KDP estimate uncertainty is unquantified. This study provides guidance on how commonly used KDP estimation algorithms perform in various environments. We create numerous synthetic (“true”) KDP profiles, integrate over them to obtain “smoothed” ΦDP, and then add noise typical of S-band operational weather radar measurements. Each algorithm is applied to our noisy ΦDP profiles and compared to the true KDP profile such that the errors and uncertainty are quantified. The synthetic KDP profiles are Gaussian in shape, which allows systematic variations in their magnitude and width to determine how each algorithm performs in smooth, slowly changing KDP profiles, as well as steep profiles. Results demonstrate that algorithm performance is dependent on the ΦDP field received. These results are further supported by an error analysis of each algorithm for two more complicated synthetic KDP profiles. Some KDP algorithms allow users to change various tuning parameters; a subset of these tuning parameters is tested to provide guidance on how changing these parameters impacts algorithm performance. We then provide evidence that our known-truth framework provides insight into algorithm performance in observed data through two case studies.

Differentiable Architecture Search Based on Coordinate Descent

Neural architecture search (NAS) is an automated method searching for the optimal network architecture by optimizing the combinations of edges and operations. For efficiency, recent differentiable architecture search methods adopt a one-shot network, containing all the candidate operations in each edge, instead of sampling and training individual architectures. However, a recent study doubts the effectiveness of differentiable methods by showing that random search can achieve comparable performance with differentiable methods using the same search cost. Therefore, there is a need to reduce the search cost even for previous differentiable methods. For more efficient differentiable architecture search, we propose a differentiable architecture search based on coordinate descent (DARTS-CD) that searches for optimal operation over only one sampled edge per training step. DARTS-CD is proposed based on the coordinate descent algorithm, which is an efficient learning method for resolving large-scale problems by updating only a subset of parameters. In DARTS-CD, one edge is randomly sampled, in which all the operations are performed, whereas only one operation is applied to the other edges. Weight update is also performed only at the sampled edge. By optimizing each edge separately, as in the coordinate descent that optimizes each coordinate individually, DARTS-CD converges much faster than DARTS while using the network architecture similar to that used for evaluation. We experimentally show that DARTS-CD performs comparably to the state-of-the-art efficient architecture search algorithms, with an extremely low search cost of 0.125 GPU days (1/12 of the search cost of DARTS) on CIFAR-10 and CIFAR-100. Furthermore, a warm-up regularization method is introduced to improve the exploration capability, which further enhances the performance.

An MPPT Algorithm for PV Systems Based on a Simplified Photo-Diode Model

A Maximum Power Point Tracking (MPPT) algorithm is proposed based on the assumption that a simplified three-parameter photodiode-based model can provide an excellent approximation of a PV module $i-v $ curve around its maximum power point (MPP). Procedures to obtain the MPP coordinates and the three parameters of the approximated $i-v $ curve from experimental online measurements, analytical and Newton-Raphson iterative calculations are thoroughly described. Initializing the model as well as optimizing it to operate faster by identifying only subsets of the model parameters provides excellent MPPT efficiency in both static and dynamical MPPT situations. The performance of the proposed algorithm has been verified in comparison with other well-known MPPT methods using the software-in-the-loop approach. Next, its performance has been evaluated by using the MATLAB-based hardware-in-the loop experimental setup that provides the required reproducibility of the different synthetic and real irradiance and temperature profiles considered.

State-of-Health Assessment for Aero-Engine Based on Density-Distance Clustering and Fuzzy Bayesian Risk

State-of-health (SOH) assessment for aero-engine can effectively reduce maintenance cost and operational risk, and is also a significant part of the prognostics and health management (PHM) system. However, the current SOH assessment is usually closely coordinated with other parts of PHM to achieve specific functions. This is not conducive to generalizing the function of SOH assessment. Therefore, this paper proposes a data-driven framework of SOH assessment that mainly includes data preprocessing, pseudo label generation, weight assignment and feature selection, and assessment, which enhances the systematicness of SOH assessment. A combination model based on density-distance clustering and fuzzy Bayesian risk models is designed to generate a pseudo label, select optimal parameter subset, and assign weight. Then, two assessment indicators including state membership degree and health degree are produced based on two fuzzy models for horizontal and vertical comparisons. These two indicators expand the dimensions and perspectives of SOH measurement, which can more comprehensively characterize the health state of the engine. Finally, the correctness and effectiveness of the proposed methodology are verified by the widely used Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) dataset.

Using Cadence to Predict the Walk-to-Run Transition in Children and Adolescents: A Logistic Regression Approach

ABSTRACT The natural transition from walking to running occurs in adults at ≅140 steps/min. It is unknown when this transition occurs in children and adolescents. The purpose of this study was to develop a model to predict age- and anthropometry-specific preferred transition cadences in individuals 6–20 years of age. Sixty-nine individuals performed sequentially faster 5-min treadmill walking bouts, starting at 0.22 m/s and increasing by 0.22 m/s until completion of the bout during which they freely chose to run. Steps accumulated during each bout were directly observed and converted to cadence (steps/min). A logistic regression model was developed to predict preferred transition cadences using the best subset of parameters. The resulting model, which included age, sex, height, and BMI z-score, produced preferred transition cadences that accurately classified gait behaviour (k-fold cross-validated prediction accuracy =97.02%). This transition cadence ranged from 136–161 steps/min across the developmental age range studied. The preferred transition cadence represents a simple and practical index to predict and classify gait behaviour from wearable sensors in children, adolescents, and young adults. Moreover, herein we provide an equation and an open access online R Shiny app that researchers, practitioners, or clinicians can use to predict individual-specific preferred transition cadences.

An Image-based Approach for 3D Left Atrium Functional Measurements.

There is growing interest in the assessment of function of the left atrium (LA) in patients with atrial fibrillation (AF). Existing methods of LA functional measurement only quantify a limited subset of the functional parameters from a single or biplane CINE-MRI scan through the LA. Here, we propose an image-based method for comprehensive evaluation of the function of the entire LA in 3D. 4D LA images were reconstructed from a series of CINE image stack covering the whole LA with small or no gap between thin slices. A segmentation from a high-resolution Magnetic Resonance Angiography (MRA) was registered and propagated through pairwise deformable registrations covering the cardiac cycle. Volume, LA ejection fraction and surface strain were computed for each timepoint and registered to Late Gadolinium Enhancement (LGE) scans for each of 52 patient scans. A correlation coefficient of -0.11 was calculated between LGE and strain, indicating that fibrotic tissue correlates with reduced elasticity.

A machine learning approach to emulation and biophysical parameter estimation with the Community Land Model, version 5

Abstract. Land models are essential tools for understanding and predicting terrestrial processes and climate–carbon feedbacks in the Earth system, but uncertainties in their future projections are poorly understood. Improvements in physical process realism and the representation of human influence arguably make models more comparable to reality but also increase the degrees of freedom in model configuration, leading to increased parametric uncertainty in projections. In this work we design and implement a machine learning approach to globally calibrate a subset of the parameters of the Community Land Model, version 5 (CLM5) to observations of carbon and water fluxes. We focus on parameters controlling biophysical features such as surface energy balance, hydrology, and carbon uptake. We first use parameter sensitivity simulations and a combination of objective metrics including ranked global mean sensitivity to multiple output variables and non-overlapping spatial pattern responses between parameters to narrow the parameter space and determine a subset of important CLM5 biophysical parameters for further analysis. Using a perturbed parameter ensemble, we then train a series of artificial feed-forward neural networks to emulate CLM5 output given parameter values as input. We use annual mean globally aggregated spatial variability in carbon and water fluxes as our emulation and calibration targets. Validation and out-of-sample tests are used to assess the predictive skill of the networks, and we utilize permutation feature importance and partial dependence methods to better interpret the results. The trained networks are then used to estimate global optimal parameter values with greater computational efficiency than achieved by hand tuning efforts and increased spatial scale relative to previous studies optimizing at a single site. By developing this methodology, our framework can help quantify the contribution of parameter uncertainty to overall uncertainty in land model projections.

Macroscopic modeling of bioreactors for recombinant protein producing Pichia pastoris in defined medium.

The methylotrophic yeast Pichia pastoris is widely used as a microbial host for recombinant protein production. Bioreactor models for P. pastoris can inform understanding of cellular metabolism and can be used to optimize bioreactor operation. This article constructs an extensive macroscopic bioreactor model for P. pastoris which describes substrates, biomass, total protein, other medium components, and off-gas components. Species and elemental balances are introduced to describe uptake and evolution rates for medium components and off-gas components. Additionally, a pH model is constructed using an overall charge balance, acid/base equilibria, and activity coefficients to describe production of recombinant protein and precipitation of medium components. The extent of run-to-run variability is modeled by distributions of a subset of the model parameters, which are estimated using the maximum likelihood method. Model prediction from the extensive macroscopic bioreactor model well describes experimental data with different operating conditions. The probability distributions of the model predictions quantified from the parameter distribution are quantifiably consistent with the run-to-run variability observed in the experimental data. The uncertainty description in this macroscopic bioreactor model identifies the model parameters that have large variability and provides guidance as to which aspects of cellular metabolism should be the focus of additional experimental studies. The model for medium components with pH and precipitation can be used for improving chemically defined medium by minimizing the amount of components needed while meeting cellular requirements.

Ranking-Based Parameter Subset Selection for Nonlinear Dynamics with Stochastic Disturbances under Limited Data

The modeling procedure for a production-scale plant includes a parameter estimation (PE) problem based on the measured data. However, when only the nominal operation data are available, however, it...

Identification of Key Physical Processes and Improvements for Simulating and Predicting Net Primary Production Over the Tibetan Plateau

AbstractThere are still considerable uncertainties related to the numerical simulation and prediction of net primary production (NPP) as an important part of terrestrial carbon sources and sinks over the Tibetan Plateau (TP). To reduce the uncertainty of numerical simulations and improve the ability of predictions, the key physical processes related to the uncertainty of simulated NPP are identified at nine observational stations over the TP. A sensitivity analysis of parameter combinations based on the Conditional Nonlinear Optimal Perturbation related to Parameters (CNOP‐P) approach, which can be used to assess the sensitivity of a parameter subset, is conducted for 28 target physical parameters in the Lund‐Potsdam‐Jena (LPJ) Wetland Hydrology and Methane Dynamic Global Vegetation Model (LPJ‐WHyMe v1.3.1). Firstly, the numerical results show that the uncertainties of physical parameters do lead to a large error in the simulated NPP over the TP, and the range of error varies from 72.4 (MS 3478) to 150.5 g C m−2 year−1 (Ngari station). Secondly, in areas of moderate precipitation over the TP, the photosynthesis is the main factor leading to high uncertainty in NPP modeling. In areas of low and high precipitation over the TP, the combined influences of hydrological processes and photosynthesis play a key role. Finally, eliminating the errors associated with the most sensitive and important parameter combinations led to the maximum benefit in terms of reducing the uncertainty of simulated NPP, when compared to that obtained with the traditional method. This study suggests that we should prioritize reducing the uncertainty of relatively sensitive parameter combinations among all physical parameters to improve the prediction or simulation ability of NPP over the TP.

Diagnosis and Compensation of Control Program, Sensor and Actuator Failures in Nonlinear Systems Using Hierarchical State Space Checks

Autonomous systems with nonlinear dynamics need to be extremely resilient to errors in sensors, actuators and on-board electronics for the purpose of overall vehicle safety. Prior research has focused on control-theoretic methods with significant computational burden with a focus on failures in actuation. In contrast, we propose the use of hierarchical machine learning driven state space checks that detect and diagnose errors in control program execution, sensors and actuators with high sensitivity and low latency. Each check produces a time-varying error signal that facilitates effect-cause diagnosis of the system, while allowing rapid parameter estimation from each check. Since the checks are over small subsets of system parameters, estimation is fast and accurate. The estimated parameters are then used to reconfigure the system controller parameters for rapid system recovery. We use a quadcopter system to demonstrate and validate our approach. Controller, sensor and actuator errors can be detected, diagnosed and compensated using a common checking platform with low computational overhead. The technique is validated on a quadcopter hardware test vehicle.