Accurate Threshold Research Articles

Spot urine protein-to-creatinine ratio as a diagnostic test in pre-eclampsia: A gold standard?

To determine the diagnostic accuracy and optimal threshold of the spot protein-to-creatinine ratio (PCR) compared to the gold standard, 24-hour proteinuria (24HP) in patients with suspected pre-eclampsia. A prospective observational study was performed from June 2015 to May 2017 consisting of patients hospitalized for suspected pre-eclampsia in a tertiary care referral center. To compare the two diagnostic tests, a spot urine sample was obtained to perform the PCR before starting the collection of the 24HP. Only patients who had both tests were analyzed. In total, 148 patients (216 samples) were included. The two tests were highly correlated (r=0.80, P<0.001). The receiver operating characteristic curve analysis and the area under the curve (AUC=0.92) highlighted the accuracy of PCR in diagnosing significant proteinuria and thus pre-eclampsia. The optimal cut-off using the Liu method was 56.9mg/mmol (sensitivity=79.3%, specificity=91.5%). The results suggest that PCR could replace 24HP when diagnosing proteinuria in pre-eclampsia. Moreover, it is a simple test, easy to realize and standardize, and cheap with no need for systematic hospitalization. The best cut-off should be chosen by thinking about the risks for adverse maternal and/or fetal outcomes. The test may help to optimize medical care in pre-eclampsia worldwide.

Enhanced multipactor statistical modeling for accurate threshold prediction with numerically tracking electron trajectories

Multipactor statistical modeling is one vital theoretical treatment for both multipactor mechanism analysis and threshold prediction, but its practical application is still severely constrained due to its obstinate dependence on the electron trajectory formula. Hence, this paper proposes a novel and general implementation method to improve the feasibility and accuracy of one-dimensional (1D) multipactor statistical modeling, where the joint probability density function is constructed through numerically tracking electron trajectories. On that basis, better agreement with the experimental results is obtained for multipactor threshold predictions of coaxial lines than the classical statistical calculations adopting the approximate electron trajectory formula. Besides the accuracy improvement, this approach also makes multipactor statistical modeling regardless of the field complexity and thus applicable to investigating any one-dimensional multipactor problem in microwave devices. This research work can further popularize the application of multipactor statistical modeling and provide one highly sought treatment with both good accuracy and efficiency for the optimal design of multipactor-free microwave devices, thus further promoting the development of space communication capability.

Accuracy and Reliability of Smartphone Self-Test Audiometry in Community Clinics in Low Income Settings: A Comparative Study.

There is a lack of hearing health care globally, and tele-audiology and mobile technologies have been proposed as important strategies to reduce the shortfall. To investigate the accuracy and reliability of smartphone self-test audiometry in adults, in community clinics in low-income settings. A prospective, intra-individual, repeated measurements design was used. Sixty-three adult participants (mean age 52 years, range 20-88 years) were recruited from ENT and primary health care clinics in a low-income community in Tshwane, South Africa. Air conduction hearing thresholds for octave frequencies 0.5 to 8 kHz collected with the smartphone self-test in non-sound treated environments were compared to those obtained by reference audiometry. The overall mean difference between threshold seeking methods (ie, smartphone thresholds subtracted from reference) was -2.2 dB HL (n = 467 thresholds, P = 0.00). Agreement was within 10 dB HL for 80.1% (n = 467 thresholds) of all threshold comparisons. Sensitivity for detection hearing loss >40 dB HL in one ear was 90.6% (n = 84 ears), and specificity 94.2% (n = 84 ears). Smartphone self-test audiometry can provide accurate and reliable air conduction hearing thresholds for adults in community clinics in low-income settings.

Highly efficient neoteric histogram–entropy‐based rapid and automatic thresholding method for moving vehicles and pedestrians detection

Thresholding for segmentation is an important key step and necessary process in various applications. Estimating an accurate threshold value for a complex and coarse image is computationally expensive and lacks accuracy and stability. This study is aimed at developing a general histogram–entropy-based thresholding method, referred as our HEBT method, for fast and efficient automatic threshold value evaluation. In the proposed method, the probability density function and Shannon entropy derived from 1D bimodal histogram have been used to find the optimal threshold values automatically. The proposed method implemented with a three-frame differencing segmentation technique has been tested on real-time datasets – change detection 2012, change detection 2014, and Wallflower – to identify pedestrians and vehicles. The performance of our HEBT method has been compared with six state-of-the-art automatic thresholding methods. The experimental segmented image results confirmed that our HEBT method is more adaptable and better suited for real-time systems with severe challenging conditions of great variations. Further, the new HEBT method achieved the best segmentation results with highest values of several performance parameters, i.e. recall, precision, similarity, and f-measure. Interestingly, the computation time is the lowest for the proposed method than the state-of-the-art methods, promising its application for a fast and effective image segmentation.

Feature reduction fuzzy C-Means algorithm leveraging the marginal kurtosis measure

The feature reduction fuzzy c-means (FRFCM) algorithm has been proven to be effective for clustering data with redundant/unimportant feature(s). However, the FRFCM algorithm still has the following disadvantages. 1) The FRFCM uses the mean-to-variance-ratio (MVR) index to measure the feature importance of a dataset, but this index is affected by data normalization, i.e., a large MVR value of original feature(s) may become small if the data are normalized, and vice versa. Moreover, the MVR value(s) of the important feature(s) of a dataset may not necessarily be large. 2) The feature weights obtained by the FRFCM are sensitive to the initial cluster centers and initial feature weights. 3) The FRFCM algorithm may be unable to assign the proper weights to the features of a dataset. Thus, in the feature reduction learning process, important features may be discarded, but unimportant features may be retained. These disadvantages can cause the FRFCM algorithm to discard important feature components. In addition, the threshold for the selection of the important feature(s) of the FRFCM may not be easy to determine. To mitigate the disadvantages of the FRFCM algorithm, we first devise a new index, named the marginal kurtosis measure (MKM), to measure the importance of each feature in a dataset. Then, a novel and robust feature reduction fuzzy c-means clustering algorithm called the FRFCM-MKM, which incorporates the marginal kurtosis measure into the FRFCM, is proposed. Furthermore, an accurate threshold is introduced to select important feature(s) and discard unimportant feature(s). Experiments on synthetic and real-world datasets demonstrate that the FRFCM-MKM is effective and efficient.

Characterising spatial dependence on epidemic thresholds in networks

Epidemic processes are an important security research topic for both the internet and social networks. The epidemic threshold is a fundamental metric used to evaluate epidemic spread in networks. Previous work has shown that the epidemic threshold of a network is 1/λmax(A), i.e., the inverse of the largest eigenvalue of its adjacency matrix. In this work, however, we indicate that such a theoretical threshold ignores spatial dependence among nodes and hence underestimates the actual epidemic threshold. Moreover, inspired by the Markov random field, we analytically derive a more accurate epidemic threshold based on a spatial Markov dependence assumption. Our model shows that the epidemic threshold is indeed 1/λmax(A)(1 − ρ), where ρ is the average spatial correlation coefficient between neighbouring nodes. We then apply simulations to compare the performance of these two theoretical epidemic thresholds in different networks, including regular graphs, synthesised irregular graphs, and a real topology. We find that our proposed epidemic threshold incorporates a certain spatial dependence and thus achieves greater accuracy in characterising the actual epidemic threshold in networks.

Deep Kalman Neuro Fuzzy-Based Adaptive Broadcasting Scheme for Vehicular Ad Hoc Network: A Context-Aware Approach

Vehicular Ad Hoc Networks (VANETs) are among the main enablers for future Intelligent Transportation Systems (ITSs) as they facilitate information sharing, which improves road safety, traffic efficiency, and provides passengers’ comfort. Due to the dynamic nature of VANETs, vehicles need to exchange the Cooperative Awareness Messages (CAMs) more frequently to maintain network agility and preserve applications’ performance. However, in many situations, broadcasting at a high rate leads to congest the communication channel, rendering VANET unreliable. Existing broadcasting schemes designed for VANET use partial context variables to control the broadcasting rate. Additionally, CAMs uncertainty, which is context-dependent has been neglected and a predefined fixed certainty threshold has been used instead, which is not suitable for the highly dynamic context. Consequently, vehicles disseminate a high rate of unnecessary CAMs which degrades VANET performance. A good broadcasting scheme should accurately determine which and when CAMs are broadcasted. To this end, this study proposes a Context-Aware Adaptive Cooperative Awareness Messages Broadcasting Scheme (CA-ABS) using combinations of Adaptive Kalman Filter, Autoregression, and Sequential Deep Learning and Fuzzy inference system. Four context variables have been used to represent the vehicular context, namely, individual driving behaviors, CAMs uncertainty, vehicle density, and traffic flow. Kalman Filter and Autoregression are used to estimate and predict the CAMs messages respectively. The deep learning model has been constructed to estimate the CAMs’ uncertainties which is an important context variable that has been neglected in the previous research. Fuzzy Inference System takes context variables as input and determines an accurate broadcasting threshold and broadcasting interval. Extensive simulations have been conducted to evaluate the proposed scheme. Results show that the proposed scheme improves the CAMs delivery ratio and decreases the CAMs prediction errors.

Children With Congenital Unilateral Sensorineural Hearing Loss: Effects of Late Hearing Aid Amplification-A Pilot Study.

Although children with unilateral hearing loss (uHL) have high risk of experiencing academic difficulties, speech-language delays, poor sound localization, and speech recognition in noise, studies on hearing aid (HA) outcomes are few. Consequently, it is unknown when and how amplification is optimally provided. The aim was to study whether children with mild-to-moderate congenital unilateral sensorineural hearing loss (uSNHL) benefit from HAs. All 6- to 11-year-old children with nonsyndromic congenital uSNHL and at least 6 months of HA use were invited (born in Stockholm county council, n = 7). Participants were 6 children (9.7- to 10.8-years-old) with late HA fittings (>4.8 years of age). Unaided and aided hearing was studied with a comprehensive test battery in a within-subject design. Questionnaires were used to study overall hearing performance and disability. Sound localization accuracy (SLA) and speech recognition thresholds (SRTs) in competing speech were measured in sound field to study hearing under demanding listening conditions. SLA was measured by recording eye-gaze in response to auditory-visual stimuli presented from 12 loudspeaker-video display pairs arranged equidistantly within ±55° in the frontal horizontal plane. The SRTs were measured for target sentences at 0° in spatially separated (±30° and ±150°) continuous speech. Auditory brainstem responses (ABRs) were obtained in both ears separately to study auditory nerve function at the brainstem level. The mean ± SD pure-tone average (0.5, 1, 2, and 4 kHz) was 45 ± 8 dB HL and 6 ± 4 dB HL in the impaired and normal hearing ear, respectively (n = 6). Horizontal SLA was significantly poorer in the aided compared with unaided condition. A significant relationship was found between aided SLA (quantified by an error index) and the impaired ear's ABR I to V interval, suggesting a relationship between the two. Results from questionnaires revealed aided benefit in one-to-one communication, whereas no significant benefit was found for communication in background noise or reverberation. No aided benefit was found for the SRTs in competing speech. Children with congenital uSNHL benefit from late HA intervention in one-to-one communication but not in demanding listening situations, and there is a risk of degraded SLA. The results indicate that neural transmission time from the impaired cochlea to the upper brainstem may have an important role in unilaterally aided spatial hearing, warranting further study in children with uHL receiving early HA intervention.

A Novel Illumination-Insensitive Feature Extraction Method

Image description is a great challenge in the field of computer vision under complex illumination condition. The complex illumination condition is usually unavoidable and unpredictable in real application. In this paper, we propose a novel generalized image descriptor, named as Anisotropy Weber Adapted Symmetric Ternary Pattern (AWASTP), which can overcome the directional inseparability of Weber Local Descriptor (WLD) and invariant threshold of Local Ternary Pattern (LTP). More particularly, we heighten the discriminative effectiveness of image description under complex illumination condition in several ways to restrain the effect of illumination variation. Firstly, a novel selection scheme for scale and angle is proposed. Based on this, an improved anisotropic Laplacian of Gaussian(ALOG) operator model is established by introducing the scale and angle parameter, moreover, an Anisotropic Weber Local Descriptor (AWLD) is presented, which can achieve more rich detailed information of illumination-insensitive feature. Secondly, an Adaptive Symmetric Ternary Pattern (ASTP) algorithm is proposed based on Weber criterion to generate more accurate threshold judgment according to the region characteristics. Thirdly, a two-dimensional AWASTP histogram is created to enhance the discriminative power and represent illumination-insensitive feature description. We conduct many experiments on benchmark databases, such as CMUPIE, FERET, PhoTex, RawFoot, and etc. Experimental results demonstrate the effectiveness of the proposed approach under different illumination conditions.

Dual-energy computed-tomography-based discrimination between basic calcium phosphate and calcium pyrophosphate crystal deposition in vivo.

Background:Dual-energy computed tomography (DECT) is being considered as a non-invasive diagnostic and characterization tool in calcium crystal-associated arthropathies. Our objective was to assess the potential of DECT in distinguishing between basic calcium phosphate (BCP) and calcium pyrophosphate (CPP) crystal deposition in and around joints in vivo.Methods:A total of 13 patients with calcific periarthritis and 11 patients with crystal-proven CPPD were recruited prospectively to undergo DECT scans. Samples harvested from BCP and CPP calcification types were analyzed using Raman spectroscopy and validated against synthetic crystals. Regions of interest were placed in BCP and CPP calcifications, and the following DECT attenuation parameters were obtained: CT numbers (HU) at 80 and 140 kV, dual-energy index (DEI), electron density (Rho), and effective atomic number (Zeff). These DECT attenuation parameters were compared and validated against crystal calibration phantoms at two known equal concentrations. Receiver operating characteristic (ROC) curves were plotted to determine the highest accuracy thresholds for DEI and Zeff.Results:Raman spectroscopy enabled chemical fingerprinting of BCP and CPP crystals both in vitro and in vivo. DECT was able to distinguish between HA and CPP in crystal calibration phantoms at two known equal concentrations, most notably by DEI (200 mg/cm3: 0.037 ± 0 versus 0.034 ± 0, p = 0.008) and Zeff (200 mg /cm3: 9.4 ± 0 versus 9.3 ± 0, p = 0.01) analysis. Likewise, BCP calcifications had significantly higher DEI (0.041 ± 0.005 versus 0.034 ± 0.005, p = 0.008) and Zeff (9.5 ± 0.2 versus 9.3 ± 0.2, p = 0.03) than CPP crystal deposits with comparable CT numbers in patients. With an area under the ROC curve of 0.83 [best threshold value = 0.0 39, sensitivity = 90. 9% (81.8, 97. 7%), specificity = 64.6% (50.0, 64. 6%)], DEI was the best parameter in distinguishing between BCP and CPP crystal depositions.Conclusion:DECT can help distinguish between crystal-proven BCP and CPP calcification types in vivo and, thus, aid in the diagnosis of challenging clinical cases, and in the characterization of CPP and BCP crystal deposition occurring in osteoarthritis.

Decoding Quantum Error Correction Codes With Local Variation

In this paper we investigate the role of local information in the decoding of the repetition and surface error correction codes for the protection of quantum states. Our key result is an improvement in resource efficiency when local information is taken into account during the decoding process: the code distance associated with a given logical error rate is reduced with a magnitude depending on the proximity of the physical error rate to the accuracy threshold of the code. We also briefly discuss an averaged approach with local information for table-lookup and localised decoding schemes, an expected breakdown of these effects for large-scale systems, and the importance of this resource reduction in the near-term.

A novel anomaly detection method based on adaptive Mahalanobis-squared distance and one-class kNN rule for structural health monitoring under environmental effects

Anomaly detection by Mahalanobis-squared distance (MSD) is a popular unsupervised learning approach to structural health monitoring (SHM). Despite the popularity and high applicability of the MSD-based anomaly detection method, some major challenging issues and limitations such as environmental variability, determination of an inappropriate threshold limit, estimation of an inaccurate covariance matrix, and non-Gaussianity of training data may lead to false alarms and erroneous results of damage detection. The main objective of this article is to propose a novel anomaly detection method based on adaptive Mahalanobis-squared distance and one-class kNN rule called AMSD-kNN for SHM under varying environmental conditions. The central idea behind the proposed method is to find sufficient nearest neighbors of training and testing datasets in a two-stage procedure for removing the environmental variability conditions and estimate local covariance matrices. An effective approach based on a multivariate normality hypothesis test is proposed to find sufficient nearest neighbors that guarantee the estimate of well-conditioned local covariance matrices. The great novelty of the proposed AMSD-kNN method is to create a novel unsupervised learning strategy for SHM by a new multivariate distance measure and one-class kNN rule. Generalized extreme value distribution modeling by the block maxima (BM) method is presented to determine an accurate threshold limit. Due to the importance of choosing adequate blocks in the BM method, a goodness-of-fit measure via the Kolmogorov-Smirnov hypothesis test is applied to select an optimal block number. The performance and effectiveness of the proposed methods are verified by two well-known benchmark structures. Several comparative studies are also conducted to demonstrate the superiority of the proposed methods over some state-of-the-art techniques. Results show that the proposed AMSD-kNN and BM methods highly succeed in detecting damage under environmental variability conditions.

Numerical continuation and delay equations: A novel approach for complex models of structured populations

Recently, many realistic models of structured populations are described through delay equations which involve quantities defined by the solutions of external problems. For instance, the size or survival probability of individuals may be described by ordinary differential equations, and their maturation age may be determined by a nonlinear condition. When treating these complex models with existing continuation approaches in view of analyzing stability and bifurcations, the external quantities are computed from scratch at every continuation step. As a result, the requirements from the computational point of view are often demanding. In this work we propose to improve the overall performance by investigating a suitable numerical treatment of the external problems in order to include the relevant variables into the continuation framework, thus exploiting their values computed at each previous step. We explore and test this internal continuation with prototype problems first. Then we apply it to a representative class of realistic models, demonstrating the superiority of the new approach in terms of computational time for a given accuracy threshold.

DNA computing inspired deep networks design

Deep neural networks have gained state-of-the-art results in many applications, such as pattern recognition and computer vision. However, most of the deep neural networks are designed manually by researchers. This architecture design process is generally time consuming and needs much expertise. Hence, automatic neural network design becomes an important issue. In this paper, we propose a novel method, called DNA computing inspired networks design (DNAND), to automatically learn high performance deep networks. In DNAND, we use DNA strands to represent blocks of a model, and these DNA strands are reacted to construct the overall networks according to the base pairing principle. We also present the killing strategy, with which we stop training “bad” models if they fail to reach the specific accuracy threshold on the validation set, so as to reduce the computational cost and accelerate the learning process. Extensive experiments on image classification and detection data sets demonstrate the effectiveness of the proposed method.

Gamma-Ray Sensitivity to Dark Matter Subhalo Modelling at High Latitudes

Searches for “dark” subhaloes in gamma-ray point-like source catalogues are among promising strategies for indirect dark matter detection. Such a search is nevertheless affected by uncertainties related, on the one hand, to the modelling of the dark matter subhalo distribution in Milky-Way-like galaxies, and, on the other hand, to the sensitivity of gamma-ray instruments to the dark matter subhalo signals. In the present work, we assess the detectability of dark matter subhaloes in Fermi-LAT catalogues, taking into accounts uncertainties associated with the modelling of the galactic subhalo population. We use four different halo models bracketing a large set of uncertainties. For each model, adopting an accurate detection threshold of the LAT to dark matter subhalo signals and comparing model predictions with the number of unassociated point-sources in Fermi-LAT catalogues, we derive upper limits on the annihilation cross section as a function of dark matter mass. Our results show that, even in the best-case scenario (i.e., DMonly subhalo model), which does not include tidal disruption from baryons, the limits on the dark matter parameter space are less stringent than current gamma-ray limits from dwarf spheroidal galaxies. Comparing the results obtained with the different subhalo models, we find that baryonic effects on the subhalo population are significant and lead to dark matter constraints that are less stringent by a factor of ∼2 to ∼5. This uncertainty comes from the unknown resilience of dark matter subhaloes to tidal disruption.

Kinetics and Thermodynamics of Reactions Involving Criegee Intermediates: An Assessment of Density Functional Theory and Ab Initio Methods Through Comparison with CCSDT(Q)/CBS Data.

Reactions involving Criegee intermediates (CIs, R1 R2 COO) are important in atmospheric ozonolysis models. In recent years, density functional theory (DFT) and CCSD(T)-based ab initio methods are increasingly being used for modeling reaction profiles involving CIs. We obtain highly accurate CCSDT(Q)/CBS reaction energies and barrier heights for ring-closing reactions involving atmospherically important CIs (R1 /R2 = H, Me, OH, OMe, F, CN, cyclopropene, ethylene, acetaldehyde, and acrolein). We use this benchmark data to evaluate the performance of DFT, double-hybrid DFT (DHDFT), and ab initio methods for the kinetics and thermodynamics of these reactions. We find that reaction energies are more challenging for approximate theoretical procedures than barrier heights. Overall, taking both reaction energies and barrier heights into account, only one of the 58 considered DFT methods (the meta-GGA MN12-L) attains near chemical accuracy, with root-mean-square deviations (RMSDs) of 3.5 (barrier heights) and 4.7 (reaction energies) kJ mol-1 . Therefore, MN12-L is recommended for investigations where CCSD(T)-based methods are not computationally feasible. For reaction barrier heights performance does not strictly follow Jacob's Ladder, for example, DHDFT methods do not perform better than conventional DFT methods. Of the ab initio methods, the cost-effective CCSD(T)/CBS(MP2) approach gives the best performance for both reaction energies and barrier heights, with RMSDs of 1.7 and 1.4 kJ mol-1 , respectively. All the considered Gaussian-n methods show good performance with RMSDs below the threshold of chemical accuracy for both reaction energies and barrier heights, where G4(MP2) shows the best overall performance with RMSDs of 2.9 and 1.5 kJ mol-1 , respectively. © 2019 Wiley Periodicals, Inc.

LEO-Augmented GNSS Based on Communication Navigation Integrated Signal.

Low Earth Orbit (LEO) is of great benefit for the positioning performance of Global Navigation Satellite System (GNSS). To realize the system of LEO-augmented GNSS, three methods to integrate communication and navigation signal for LEO communication system with the least influence on the communication performance are analyzed. The analysis adopts the parameters of IRIDIUM signal as restrictions. This paper gives quantitative comparison of these methods considering CN0(carrier noise power spectral density rate) margin, pseudorange accuracy, Doppler accuracy, and communication loss. For method 1, a low-power navigation signal is added to the communication signal. For method 2, the navigation signal is launched in one or more frames. For method 3, the navigation signal is launched in the frequency band separated to the communication signal. The result shows that the pseudorange accuracy of method 2 is far below method 1 and method 3. However, the difference of Doppler accuracy among the three methods can be emitted. Detailed analysis shows that method 1 is practicable when the communication and navigation signal power rate is 15 dB. It achieves the balance of pseudorange accuracy and bit error rate (BER) performance under this condition. Comprehensive comparison of these methods is given in the last. The result shows that the CN0 margin of the navigation signal for method 3 can be 13.04 dB higher than method 1, based on the accuracy threshold considered in this paper. Methods 1 and 3 have the advantage of high accuracy and high CN0 margin respectively. However, method 3 causes high communication capacity loss. Considering that the main disadvantage of GNSS signals is low CN0, method 3 is a good choice for the LEO-augmented GNSS system. Methods 1 and 3 can be combined to realize both high accuracy and high CN0 margin if possible.

Logarithmic Versus Linear Change in Step Size When Using an Adaptive Threshold-Seeking Procedure in a Frequency Discrimination Task: Does It Matter?

Different rules for changing step sizes (e.g., logarithmic, linear) are alternately used in adaptive threshold-seeking procedures, with no clear justification. We hypothesized that the linear rule may yield more accurate thresholds for poor performers because the step sizes are predetermined and fixed across listeners and thus can be small, in contrast to the logarithmic rule, in which step sizes are changed with respect to the listener's performance. Purpose The aim of this study was to test the effect of logarithmic and linear rules on frequency discrimination (FD) thresholds. Method Three experiments involving human subjects and Monte Carlo computer simulations were designed and conducted. In the 1st experiment, FD thresholds were estimated in 40 young adults with either 3-interval 2-alternative forced choice (3I2AFC; n = 19) or 2-interval 2AFC (n = 21) in a within-subject design. In the 2nd experiment, thresholds were estimated in 16 children (7-8 years old) in a within-subject design, using 3I2AFC. In the 3rd experiment, thresholds were estimated in 30 young adults in a between-subjects design using 3I2AFC. Results No significant differences were shown between the 2 rules, regardless of age group, method, or level of FD performance. Computer simulations supported the empirical findings, predicting similar FD thresholds for both rules in the majority of runs. However, they also yielded more accurate thresholds with the linear rule, but with a larger number of outliers, which increased as the listener's attention level decreased. Conclusion Overall, the use of a particular rule has little influence on FD thresholds. Possible outliers may be minimized by monitoring the participant's attention at the beginning of the run.

Gaussian Process-Based Refinement of Dispersion Corrections.

We employ Gaussian process (GP) regression to adjust for systematic errors in D3-type dispersion corrections. We refer to the associated, statistically improved model as D3-GP. It is trained on differences between interaction energies obtained from PBE-D3(BJ)/ma-def2-QZVPP and DLPNO-CCSD(T)/CBS calculations. We generated a data set containing interaction energies for 1248 molecular dimers, which resemble the dispersion-dominated systems contained in the S66 data set. Our systems represent not only equilibrium structures but also dimers with various relative orientations and conformations at both shorter and longer distances. A reparametrization of the D3(BJ) model based on 66 of these dimers suggests that two of its three empirical parameters, a1 and s8, are zero, whereas a2 = 5.6841 bohr. For the remaining 1182 dimers, we find that this new set of parameters is superior to all previously published D3(BJ) parameter sets. To train our D3-GP model, we engineered two different vectorial representations of (supra-)molecular systems, both derived from the matrix of atom-pairwise D3(BJ) interaction terms: (a) a distance-resolved interaction energy histogram, histD3(BJ), and (b) eigenvalues of the interaction matrix ordered according to their decreasing absolute value, eigD3(BJ). Hence, the GP learns a mapping from D3(BJ) information only, which renders D3-GP-type dispersion corrections comparable to those obtained with the original D3 approach. They improve systematically if the underlying training set is selected carefully. Here, we harness the prediction variance obtained from GP regression to select optimal training sets in an automated fashion. The larger the variance, the more information the corresponding data point may add to the training set. For a given set of molecular systems, variance-based sampling can approximately determine the smallest subset being subjected to reference calculations such that all dispersion corrections for the remaining systems fall below a predefined accuracy threshold. To render the entire D3-GP workflow as efficient as possible, we present an improvement over our variance-based, sequential active-learning scheme [ J. Chem. Theory Comput. 2018 , 14 , 5238 ]. Our refined learning algorithm selects multiple (instead of single) systems that can be subjected to reference calculations simultaneously. We refer to the underlying selection strategy as batchwise variance-based sampling (BVS). BVS-guided active learning is an essential component of our D3-GP workflow, which is implemented in a black-box fashion. Once provided with reference data for new molecular systems, the underlying GP model automatically learns to adapt to these and similar systems. This approach leads overall to a self-improving model (D3-GP) that predicts system-focused and GP-refined D3-type dispersion corrections for any given system of reference data.

A research on classification performance of fuzzy classifiers based on fuzzy set theory

Due to the complexities of objects and the vagueness of the human mind, it has attracted considerable attention from researchers studying fuzzy classification algorithms. In this paper, we propose a concept of fuzzy relative entropy to measure the divergence between two fuzzy sets. Applying fuzzy relative entropy, we prove the conclusion that patterns with high fuzziness are close to the classification boundary. Thus, it plays a great role in classification problems that patterns with high fuzziness are classified correctly. Meanwhile, we draw a conclusion that the fuzziness of a pattern and the uncertainty of its class label are equivalent. As is well known, entropy not only measures the uncertainty of random variable, but also represents the amount of information carried by the variable. Hence, a fuzzy classifier with high fuzziness would carry much information about training set. Therefore, in addition to some assessment criteria such as classification accuracy, we could study the classification performance from the perspective of the fuzziness of classifier. In order to try to ensure the objectivity in dealing with unseen patterns, we should make full use of information of the known pattern set and do not make too much subjective assumptions in the process of learning. Consequently, for problems with rather complex decision boundaries especially, under the condition that a certain training accuracy threshold is maintained, we demonstrate that a fuzzy classifier with high fuzziness would have a well generalization performance.