Search

Deep neural networks (DNNs) have been making progress in many ways. DNNs are typically used to model complex nonlinearity of high-dimensional data in regression or classification problems. As DNNs contain additional hidden layers, they generally improve performance but increase the number of parameters to train, thereby extending the learning time. Many studies, such as those employing Dropout and regularization methods, have undertaken to solve these problems. The method proposed in this paper is an additive ensemble neural networks (AENNs), by which a boosting mechanism of an ensemble methodology is applied to the neural networks instead of regularization techniques. That is, the model by AENNs is obtained by sequentially combining several simple shallow network models. Experiments showed that AENNs yield better results than conventional DNNs and machine learning methods for regression and classification problems, thereby alleviating the troublesome model complexity issue.

The current article dedicated to analyzing the feasibility of using conventional techniques of statistical synthesis of prognostic decisions in the conditions of dynamic chaos, which characterizes management in unstable submersion environments. We show the fundamental difference between unstable system state observation series and probabilistic descriptions of traditional models based on the statistical paradigm. We consider an additive model with a chaotic systemic component and non-stationary noise, which describes the aforementioned observation series most adequately. We propose a method for pragmatic estimation of functional efficiency of forecast techniques in the conditions of chaotic non-determinism.

Highway fog is one of the main reasons of causing a traffic accident, so the fog highway visibility detection is very important. In the existing methods, it is necessary to set an appropriate target in advance for visibility calculation and most of the methods are based on neural network for visibility prediction, which restricts the efficiency of visibility detection and prediction. In view of the above problems, this paper proposes a highway visibility detection and prediction method based on lane detection and Winters additive model in foggy days. This method is based on the dark channel prior theory to obtain the highway image after fog removal and extract the coordinates of the end points of the lane line from the image. The visibility of highway in foggy days is calculated with the lane line as the reference. The visibility time series were fitted and predicted based on Winters additive model. The experimental results show that the method can accurately calculate and predict the highway visibility in foggy days without the need of camera calibration.

The vessel traffic flow refers to the overall movement of continuously operating vessels and other objects which exhibit similar characteristics in water transportation. There are five parameters that quantitatively describe the vessel traffic flow: traffic flow direction, traffic flow, traffic flow width, traffic flow density and traffic flow velocity. The current researches of vessel traffic flow mainly focus on flow forecasting. In order to solve the problem of traffic flow velocity prediction, a time series models which based on automatic identification system (AIS) data are used to predict the traffic flow velocity of narrow waterway in Beijing-Hangzhou Grand Canal. The Prophet Model, which is an open source time series analysis model developed by Facebook, is based on the Additive Model, which uses year, month, day, etc., plus holiday effects to fit non-linear trends. This model is most suitable for fitting data with strong periodicity and several cycles, and has good support for missing values, trend offsets and outliers. The results show that the Prophet model is superior to the Long Short-term Memory (LSTM) model in terms of ship traffic flow velocity prediction accuracy and training time. It has stronger applicability and higher accuracy. The Prophet time series prediction model can provide significant scientific basis for maritime administration authorities in their daily management and program of jurisdiction waters.

Forecasting techniques are used in the various problem domains such as- sales, banking, healthcare, stock market, etc. The time-series dataset has time-related information that is useful for prediction and statistical analysis. The supermarket sales prediction helps improve sales in a business environment. The technique helps in decision making in a problem domain. Many tools are available for forecasting such as the regression model, Logistic exponential model. The Facebook (FB) Prophet is the latest tool that has shown an improved performance in terms of accuracy of prediction. This research work has proposed a FB Prophet tool for the sales prediction of the supermarket data. The proposed research work has examined few forecasting models such as- The additive model, the Autoregressive integrated moving average (ARIMA) model, FB Prophet model. From the propsoed research work, it is concluded that, FB Prophet is a better prediction model in terms of low error, better prediction, and better fitting.

This study constructs multiple models for forecasting weekly average electricity prices using weekly weather forecasts, and applies it to arbitrage trading in the forward market. In particular, we compare models that used different approaches for forecasting weekly average price and price density, and clarify the following empirical results using the data from Japan Electric Power Exchange (JEPX): 1) Instead of using forecasted temperature directly as an explanatory variable, the two-step forecast method with measured temperature is more likely to reduce the forecast error; 2) Quantile regression has better density forecast accuracy than a GARCH based model; 3) The multiplicative model using the logarithmic price series tends to have higher forecast accuracy than the additive model without logarithmic conversion; 4) Weather forecasts contribute to improving the forecast accuracy of weekly electricity prices and also play an important role in earning profits in forward market trading. The proposed arbitrage trading method can be utilized by many participants in that the strategy can be flexibly changed according to the level of risk tolerance. The existence of considerable arbitrage opportunities in the JEPX forward market, revealed by the empirical results, may have practical implications on attracting traders and stimulating the market.

Coronavirus2019 has been seen as an overall pandemic and spreads all over India. Tamilnadu is one of the most affected states in India. The study of the spread of this disease is essential to foresee the requirement in improving the health care facilities. A couple of assessments are being driven using distinctive models to foresee the probable progression of this disease. To report the present scenario, the research uses exploratory data processing and uses time series forecasting approaches to estimate the future trend. Here, the spread of the disease is analyzed in Tamil Nadu and employed Prophet Model to forecast the transmission of COVID-19 in the forthcoming months. Results show that the spread of the disease will be gradually reduced and the state will be back to normal in the forthcoming months.

The traditional gradient descent based optimization algorithms for neural network are subjected too many vulnerabilities, such as slow convergent rate, gradient vanishing and falling into local minima. Therefore, the alternative non-gradient descent learning algorithm was proposed and prevalently applied in kinds of domains, such as pseudoinverse learning algorithm (PIL). However, when a special variant of the PIL, taking the random configuration of weight parameters, is adopted, the generalization ability needs further improvement although it has excellent training efficiency. Thus, on consideration of integrating the idea of ensemble learning, we proposes two methods to enhance basic PIL. One method is equivalent to an additive model, which can raise the network’s performance by introducing boosting mechanism, and the other is to adopt a recursive way to rectify the hidden layer output of the neural network, then the relative better model is used in the subsequent prediction. Comprehensive evaluating experiments are conducted on several datasets, and the experimental results illustrate that the our proposed methods are effective on the classification accuracy.

PM2.5 monitoring data of air pollutants was calibrated and predicted. Due to the lack of monitoring data limited by equipment, the mean and linear interpolation was used to fill in the missing data. ARIMA model (A) was established based on the fluctuation characteristics of PM2.5. The difference between the monitoring value and the standard value was taken as the dependent variable, and five Meteorological factors, namely wind, pressure, precipitation, temperature and humidity, were taken as the independent variables. Multiple regression model (B) was developed. Then, the additive model y = A + B was built. By comparing the average relative error, ARIMA and Multiple Linear Regression Additive Model based on linear interpolation was the best (0.1616), followed by the model with interaction (0.1828), and the third was ARIMA and Multiple Linear Regression Additive Model based on mean filling (0.3080). The three models reduced the average relative errors and improved the effects of forecast.

PM10 monitoring data showed significant time series characteristics and its accuracy is affected by the monitoring instruments and meteorological factors. Due to the lack of monitoring data limited by equipment, the mean and linear interpolation was used to fill in the missing data. ARIMA model (A) was established based on the fluctuation characteristics of PM10. The difference between the monitoring value and the standard value was taken as the dependent variable, and five Meteorological factors, namely wind, pressure, precipitation, temperature and humidity, were taken as the independent variables. Multiple regression model (B) was developed. Then, the additive model y = A + B was built. By comparing the average relative error, ARIMA and Multiple Regression Additive Model based on linear interpolation was the best (0.3433), followed by ARIMA and Multiple Linear Regression Additive Model based on mean filling (0.3810) , and the third was ARIMA and Multiple Regression Additive Model based on mean filling (0.3974). The three models reduced the average relative errors and improved the effects of forecast.

Recent changes in power system, e.g. large-scale penetration of renewable energy source, have led to change the behavior of the electricity demand and market price drastically; this behavior is also influenced by various factors such as weather and economic conditions. The factors that affect such demand and price are complex and the analysis of these factors enables stakeholders to construct an electricity business strategy. Therefore, a framework is needed to analyze and interpret the relationships among these demand, price, and factors. Construction of a graphical model is an attractive approach to describe the statistical relationships among these factors; the directed graphical model flexibly visualizes the direct/indirect factors affecting the demand and price. In this study, the authors applied a graphical modeling scheme using the sparse partially linear additive models for factor analysis of electricity demand and price. The scheme is applied to a real-world dataset and discussed from the viewpoint of interpretability.

This paper introduces an important model that can be widely used in semi-parametric methods in recent years, the analysis of the additivity model, to more accurately describe the deep impact of information technology on national economic innovation and subdivides information technology into inflows and outflows, and examines the direction and intensity of the impact of inflows and outflows of information technology on China's national economic innovation. When the linear part is constrained, we construct Ridge Estimators Based on profile least squares method and backfitting method, discuss the properties of the estimators, and finally verify the validity of the estimators by numerical simulation. We hope to more fully reveal the reasons for the unbalanced development of China's regional development caused by the flow of information technology.

Robustness analysis framework for the high-performance data centers based on immune genetic algorithm is implemented in this paper. The server equipment of the data center mainly includes the general computing server and the storage server. At present, the most widely used core energy consumption model of server equipment in data center is mainly additive model and system utilization model. We propose the 2 primary novelties. (1) Extend the constant number of data block copies of the quantum-aware storage strategy to a flexible and variable data block storage strategy under the premise of ensuring the availability of data blocks. (2) For files with high access frequency, the frequent reading of multiple demand ends of a single active data block will cause congestion of transmission channels and reduce the accuracy of basic data reading. The simulation has proven the model’s robustness.

Among the different models investigated for SONAR image denoising, the analysis considers two noise models, the signal independent additive noise model and the multiplicative noise model. For the signal-independent additive model, in the transform domain for the denoising a mutiresolution analysis method and the sparsity of the natural sonar images are exploited. For the same model in the spatial domain a side scan sonar image is denoised by Kalman filter-based estimation method. For the multiplicative noise model, in the transform domain a mixed noise removal based on probabilistic patch-based processing is employed and in the spatial domain, the fractional integral mask based method and an unscented Kalman filter based estimation method is adopted.

User electricity consumption behavior Analysis is helpful for power supply enterprises to provide personalized services for users, and for the effective implementation of peak load staggering scheme in power supply enterprises. In this paper, the additive model is used to decompose the user electricity consumption, extract the characteristics of electricity trend, periodicity and holiday influence, and construct the modulus 1 vector to describe the user electricity consumption behavior. Then, the power consumption behavior of different users can be analyzed by the unit spherical clustering method, which is convenient for power supply enterprises to provide precision service for different user groups with different characteristics.

In the United States of America, more than 100 million adults currently live with diabetes mellitus as well as prediabetes, a condition that occurs as a result of elevated blood sugar levels. Diabetes mellitus can be identified in a series of laboratory tests where blood must be drawn to analyze a multitude of parameters such as Glucose, Hemoglobin A1C (HGBA1C), Alanine transaminase (ALT), Cholesterol (CHOL), High-density lipoprotein (HDL), Low density lipoprotein calculated (LDL_CALC), and Triglyceride (TRIG). This disease must be identified in its early stages because diabetes mellitus can lead to heart and blood vessel problems which may be fatal to the patient. This research looks at a dataset spanning ten years of a patient comprehensively to see the patterns established and how they can be useful in advising lifestyle and routine changes to prevent diabetes mellitus of a patient. In order to predict the Diabetes Mellitus parameters, statistical computation and graphical visualization was implemented using the “R” programming language. In R, different approaches were taken to explore the data in two different ways, the parameters versus the data. The parameters were looked at with the principal component analysis (PCA) approach to understand the effects of critical parameters to the patient. The data was analyzed with regards to the integrity of the patient's health as well as potential symptoms they may experience. The clustering technique was used to group the data in a way that visually represented the interactions amongst the points. Clustering serves as an easy way to organize data and detach it from details that may not be as important as the points themselves. PCA and clustering are purely descriptive statistics that can tell a physician about the patient's current health condition, but does not give them information about future symptoms. In order to create predictive models to foresee how the parameters will fluctuate over the next ten years, Vector Auto Regressive models (VARS) can be utilized. These additive models can inform a physician about how the patient's parameters are progressing within a time frame. The prediction results are within three standard deviation values which will provide the low and high range values when generating the models.

The approach of comparator identification for solving the problem of parametric synthesis of multi-criteria evaluation models for UAV construction options is considered. To normalize the values of particular criteria, it is proposed to use universal functions of their utility, which have greater accuracy in approximating expert estimates and less time complexity than the well-known Gauss, Harrington and logistic functions. Relations are given for determining the parameters of additive models and models constructed on the basis of the Kolmogorov-Gabor polynomial based on the results of qualitative assessments of alternative options by experts, presented in the form of binary relations of strict preference and equivalence. As a result of regularization, the incorrect initial problem is reduced to the linear programming problem of searching for the Chebyshev point of a system of equations and inequalities.

This work presents a novel target detector that combines a nonparametric approach for conditional probability density function (pdf) estimation and an adaptive estimation of the target strength of the additive model it is based on. The variable bandwidth kernel density estimator is employed for pdf estimation within the Generalized Likelihood Ratio Test (GLRT) framework and a closed-form solution is found. Experimental results featuring hyperspectral data of a real subpixel target detection scenario reveal the potential of the proposed approach.

We consider learning a d-dimensional parameter w through nonlinear input/output relation governed by ReLU activation. We study a supervised learning setup in which we want to decode w from input output pairs $({\text{x}},y)$ . We consider an additive model with nonlinear ReLU activation that can be represented as $ = \mathop \sum \limits_{k = 1}^d {\text{ReLU}}(w[k] + {\text{x}}[k])$ . Such a model appears in representation learning and recommendation systems where w corresponds to an unknown embedding of a user or item and the x correspond to embedding of known probe vectors. In this paper, we show that a gradient descent algorithm linearly converges with ${\mathcal{O}}(d)$ samples and quickly finds the true parameter w under mild assumptions. Our assumptions are in terms of the input distribution that captures the fundamentals of the problems. We also demonstrate the performance of our algorithm with numerical simulations.

The urgency of the problem under study is due to the need to improve the efficiency of the activities of industrial enterprises operating in constantly changing conditions. Chemical industry enterprises are enterprises with seasonal demand for products, especially in the production of mineral fertilizers. At the same time, they often face the problem of a large number of defective products and significant transportation costs. In addition, enterprises are forced to store significant volumes of purchased products. This problem arises because of inaccurate forecasts of demand and volumes of raw materials ordered from suppliers [11]. One of the ways to reduce costs is to determine the optimal number of deliveries of inventory items [3]. The purpose of the article is to develop an additive model for determining the seasonal effect on the quantity of products supplied to the chemical industry. The leading method of research is a statistical method for determining the seasonality index, which made it possible to create a trend equation that reflects the average monthly supply of raw materials and the growth rate in each month of the study period. The main results of the article: seasonality indices were determined, the seasonal component of the additive model of product supply by chemical industry enterprises was evaluated. Materials of the article can be useful for any business entities that operate in conditions of seasonal demand. The results will allow to make a forecast of the optimal number of deliveries.

The acoustic emission phenomenon arises due to local changes of dynamic structure of solid body. Main sources of emission are micro and macro cracks, friction and shifts. Active and passive acoustic emission methods are widely used to study the strength of materials, the landscape stability and various stages of seismic process. Geoacoustic emission is an acoustic emission of the sound range and it describes the interaction of micro and macro dislocations. Anomalies of geoacoustic signals may be earthquake precursors and they are of great interest to researchers. A typical geoacoustic signal consists of sequence of specific shape short pulses. Authors propose an additive model of geoacoustic signal. According to the model the signal decomposes into a sum of components described by the modulated Berlage and Gauss functions. The use of the matching pursuit method was offered to determine model coefficients. Unfortunately, this method has cubic computational complexity depending on number of functions on which the signal is decomposed. This article is devoted to ways to improve accuracy of the model offered by authors. Various numerical schemes allowed to increase the adaptive property of matching pursuit method with respect to geoacoustic emission signals are considered and compared.

Standard classification models are usually additive models, which only consider the contributions from the main effects of features. When the features are highly correlated, the interactions between features provide us not only more additional features, but also the underlying graphs between features. In this paper, we integrate into multiclass SVM a strong hierarchy regularization in order to learn the main effects and the interactions. A primal-dual proximal algorithm with epigraphical projection is proposed to minimize the objective function. The proposed algorithm is applied to face classification task on the Extended YaleB database and the results validate its effectiveness.

A popular approach to obtain a consensus ranking from ranking data is based on the probabilistic, distance-based Mallows model comprising of a modal permutation and dispersion parameters. Often, the population consists of several subpopulations. As a result, finite mixture models are used to distinguish latent sub-groups of individuals in a heterogeneous population. Given a finite number of subpopulations each based on the Mallows model, a popular inference approach is the computationally intensive expectation maximization algorithm for additive models. We address the drawbacks of this model using a novel multiplicative mixture Mallows model (M4). Given complete ranking observations from a heterogeneous population, we derive inference algorithms for the joint estimation of the parameters and the consensus rankings of the component distributions. We numerically validate the permutation estimation performance of the proposed algorithms on synthetic datasets. We also demonstrate the goodness-of-fit using the Bayesian information criterion and the integrated complete likelihood on the real-world APA and Sushi datasets.

In this work a novel target detector is proposed that is nonparametric in terms of conditional probability density function (pdf) estimation and parametric with respect to the target strength of the additive model it relies upon. The variable bandwidth kernel density estimator is employed to estimate the conditional pdfs, whereas the target strength is estimated via the Maximum Likelihood approach. Experimental results over real hyperspectral data show that the detector succeeds in detecting target objects embedded in a complex background and in providing reasonable estimates for the target strengths.

Complementary metal oxide semiconductor (CMOS) transistors are considered as dominant technology in semiconductor industry. However, downscaling these transistors, despite of many achievements, has resulted in the rise of reliability issues. Environmental parameters namely temperature and power supply as well as circuit parameters such as duty cycle and frequency also time effect can lead to variations in some circuit characteristics. The large dependency of leakage power to the variations of parameters increases the importance of modeling. This paper is intended to introduce a new aging-aware model in order to calculate the leakage power consumption. Firstly, the leakage power as a function of these parameters, making use of the proposed model, is calculated for standard cells. Next, the power consumption of the same standard cells are measured by HSPICE. Next, we apply this model which is additive to calculate the leakage power of ISCAS85 and ISCAS89 benchmark circuits for 32nm and 22nm CMOS technology. Finally, the accuracy of this model is evaluated by comparing the calculated results from the additive model and HSPICE full-chip simulation results. We demonstrate that our simple yet accurate and additive model results in the maximum 4.74% mean percentage error.

In this study, we introduce an ensemble-based approach for online machine learning. Here, instead of working on the original data, several Hoeffding tree classifiers classify and are updated on the lower dimensional projected data generated from originality by random projections. Since random projection is unstable, from one example, many diverse training data can be created to train the set of Hoeffding tree classifiers. The experiments conducted on a number of datasets chosen from different sources demonstrate that the proposed approach performs significantly better than the single Hoeffding tree and some well-known online learning algorithms including additive models and Online Bagging.

Sparse representations of sound fields have become popular in various acoustic inverse problems. The simplest models assume spatial sparsity, where a small number of sound sources are located in the near-field. However, the performance of these models deteriorates in the presence of strong reverberation. To properly treat the reverberant components, we introduce three types of reverberation models: a low-rank model, a sparse model in the plane-wave domain, and a combined low-rank+sparse model. We discuss corresponding decomposition algorithms based on ADMM convex optimization. Numerical simulations indicate that the decomposition accuracy is significantly improved by the additive model of low-rank and sparse plane wave models.

Principal component analysis (PCA) is one of the most versatile tools for unsupervised learning with applications ranging from dimensionality reduction to exploratory data analysis and visualization. While much effort has been devoted to encouraging meaningful representations through regularization (e.g. non-negativity or sparsity), underlying linearity assumptions can limit their effectiveness. To address this issue, we propose Additive Component Analysis (ACA), a novel nonlinear extension of PCA. Inspired by multivariate nonparametric regression with additive models, ACA fits a smooth manifold to data by learning an explicit mapping from a low-dimensional latent space to the input space, which trivially enables applications like denoising. Furthermore, ACA can be used as a drop-in replacement in many algorithms that use linear component analysis methods as a subroutine via the local tangent space of the learned manifold. Unlike many other nonlinear dimensionality reduction techniques, ACA can be efficiently applied to large datasets since it does not require computing pairwise similarities or storing training data during testing. Multiple ACA layers can also be composed and learned jointly with essentially the same procedure for improved representational power, demonstrating the encouraging potential of nonparametric deep learning. We evaluate ACA on a variety of datasets, showing improved robustness, reconstruction performance, and interpretability.

This paper introduces methods of deriving and computing maximal robust positively invariant sets for linear discrete time systems with additive model uncertainty. Two types of uncertainty are considered: state dependent uncertainty, which can handle multiplicative parametric model uncertainty as well as linearisation errors for nonlinear systems, and scaled sets of uncertainty. We provide a framework for analysing both types of uncertainty with illustrative examples.

Reproductive performance is important for the economic efficiency of pasture-based dairy farms. In these seasonal calving systems, a concise period of breeding is essential to ensure the alignment of peak grass availability with peak lactating cow energy demands. Trials and statistical analysis have identified the factors affecting overall reproductive performance, but few studies have analysed performance at the individual service level. In this paper, four binary models of service outcome are described, incorporating age, stage of lactation, calving events, and measures of energy balance and milk production. Random effects at the cow, sire and herd level were included. Logistic regression and generalised additive models were created, both as stand-alone predictors and using ensemble learning in the form of bagging. The four models were evaluated in terms of calibration and discrimination using an external dataset of nine dairy herds representing the typical Irish pasture-based system. Logistic regression (with and without bagging) and generalised additive modelling with bagging all performed satisfactorily and would be useful as stand-alone models or in whole-farm simulation. Logistic regression is suggested as the most useful model for farmers and their advisers due to ease of interpretation. This model will be used as part of a PhD project to create simulation software for seasonally calving dairy animals.

The inherent rate-dependent hysteresis in smart actuators hinders their wider applicability in practical engineeering. In this paper, based on the rate-dependent Hammerstein model, a 2 DOF control scheme with a robust controller combined with a hysteresis compensator is proposed for tracking control of smart actuators. The hysteresis compensator is first connected in cascade with an actuator to cancel out the static hysteresis, then the compensation errors and the modeling errors are attributed to the additive model uncertainty, and finally an H ∞ integral robust controller is designed. Compared to the H ∞ robust controller without integrator, the H ∞ integral robust controller results in smaller errors. The tracking control simulation results for a giant magnetostrictive actuator verified the effectiveness of the proposed control scheme.

A minimum variance filter for a class of discrete time systems with additive as well as multiplicative noise is investigated in this paper. We extend the results from recent work by Ponomareva and Date to account for multiplicative noise in the measurement equation. More importantly, we provide an interpretation of the multiplicative noise in both transition and measurement equations in terms of parameter perturbations in a linear additive model. The utility of the proposed filtering algorithm is demonstrated through numerical simulation experiments using models from academic literature where the parameters are estimated from real data.

This study reports the performance of 23 Electrical Distribution Divisions (EDD's) of an Indian state, Uttarakhand for the period of 2005–2006 using the concepts of Additive Model (AM), Slack Based Modelling (SBM) and Two-stage Data Envelopment Analysis (DEA). Slack Based Modelling helps in quantifying the slacks, which is a hidden source of inefficiency that might be present in BCC efficient DMU's. Following the slack quantification, a two-stage DEA is applied. The results obtained from AM and SBM indicate that numerous EDD's have scope for improvement in overall efficiency by incorporating the slacks that have been identified in the particular EDD. Further the two stage DEA modelling suggest that the inefficiency is primarily due to the intermediate measures.

In this paper, we consider problems of hedging path-dependent basket barrier options using individual options based on our previously proposed optimal hedging strategy in Yamada (2010–2013) for European options. The optimal hedging problem is formulated as follows: Find optimal smooth functions of individual options to minimize the mean square error from the payoff of down-and-out basket option. For solving the problem, we first express the necessary and sufficient condition in terms of linear equations regarding conditional expectations of basket barrier options. Based on the Brownian bridge decomposition and the Independence Lemma, we show that the computations involving conditional expectations of basket barrier options may be reduced to those of unconditional expectations. This computation involves multiple integrations in general, but is usually executed efficiently based on the standard Monte Carlo method by generating independent Gaussian random numbers. Finally, we demonstrate numerical experiments to illustrate our proposed methodology.

In this paper, we consider the problem of parameter estimation over sensor networks under data rate constraint. A general additive quantization model is introduced to capture the data rate constraint. Existing works on the effect of the additive model on standard consensus algorithms show that convergence can be guaranteed only if the quantization error variances form a convergent series. We propose to incorporate a moving average step into the consensus algorithm to smear out the randomness caused by quantization errors. It is shown that the proposed algorithm achieves the performance of the optimal centralized sample mean estimator even if the quantization error variances are not vanishing. This is guaranteed by establishing a law of the iterated logarithm for weighted sums of independent random vectors. Moreover, an explicit bound of the rate of convergence is given to quantify its almost sure performance. Finally, simulations are provided to validate the theoretical results.

In this paper we present a novel framework for cross-age face verification (FV) by seeking help from its “competitor” named cross-face age verification (AV), i.e., deciding whether two face photos are taken at similar ages. While FV and AV share some common features, FV pursues age insensitivity and AV seeks age sensitivity. Such correlation suggests that AV may be used to guide feature selection in FV, i.e., by reducing the chance of choosing age sensitive features. Driven by this intuition, we propose to learn a solution for cross-age face verification by coordinating with a solution for age verification. Specifically, a joint additive model is devised to simultaneously handling both tasks, while encoding feature coordination by a competition regularization term. Then, an alternating greedy coordinate descent (AGCD) algorithm is developed to solve this joint model. As shown in our experiments, the algorithm effectively balances feature sharing and feature exclusion between the two tasks; and, for face verification, the algorithm effectively removes distracting features used in age verification. To evaluate the proposed algorithm, we conduct cross-age face verification experiments using two benchmark cross-age face datasets, FG-Net and MORPH. In all experiments, our algorithm achieves very promising results and outperforms all previously tested solutions.

This paper studies decentralized estimation with correlated observations. Focusing on the additive model with correlated Gaussian noises, we attempt to answer several distinct yet related questions in decentralized estimation: 1. When does correlation imply redundancy, i.e., incur performance degradation compared with that of independent observations; 2. What is the optimal quantizer structure that maximizes the Fisher information at the fusion center; 3. What is the preferred communication direction in a tandem fusion network involving correlated observations? It is shown that there exist different correlation regimes whose impacts on the estimation performance are in sharp contrast with each other. For the Gaussian model, it is established that quantizing the observation is optimal regardless of the correlation coefficients; this is true despite the fact that subsequent estimators may differ at the fusion center. Finally, it is always beneficial to have the better sensor (i.e., that has a higher SNR) to serve as a fusion center in a tandem fusion network for all correlation regimes.

This paper presents a systematic methodology that rates the effectiveness of each tracker configuration or signal processing algorithm in a radar system. The approach uses a linear additive model for aggregating selected Measures of Performance (MOP) with the relative importance of each MOP determined through the application of the analytic hierarchy process (AHP). With the aggregate MOP score, track pairings based on an established baseline identifies differences in track data, and obtains the measures of effectiveness (MOE) of the algorithm/configuration being evaluated. The results are used to help determine “Return-on-Investment” in implementing signal processing or tracker parameter changes. Examples are presented that illustrate how this approach has been used to evaluate the performance of both the tracker and the mid-stage processing of a High Frequency Surface Wave Radar (HFSWR) system. The approach however is generic and applicable to evaluating updates to the signal processing schema or tracker of any radar system.

New linear prediction polymodels is described in the article. The structure of the universal linear prediction model is shown. We propose new additive linear prediction model AR1(2)+AR2(2). We also obtain spectral characteristics of additive processes and a system of equations for calculation of the additive model's parameters. Comparative analysis of resolution capability is made using four methods of statistical modeling.

The process of negotiation between buyers and suppliers has different objectives conflicting with each other beyond issues related to the price of a product or service. In this scenario, each party develops his/her own strategy, normally using a win-lose approach. In some situations the final outcome of negotiation implies in a loss for both, which can influence others issues involved in the process beyond price. The objective of this paper is to present a multicriteria additive model to support a bilateral and multi-issue negotiation between a buyer and a material supplier (seller). Moreover, it is presented an application of the model in the construction industry. The main contribution of the paper is related to structuring and standardizing a multi-issue negotiation support system which tackles how best this industry might contract suppliers. The results indicate that it is possible to develop win-win strategies and to establish a better outcome for both parties.

The maintaining of a smooth quality of research to upgrade product performance, quality and with minimum possible production cost through Off-line quality control is a effective way of optimizing product and process design in support of on-line quality control. At the off-line stage, several ideas are generated and analysis, an appropriate one is chosen and translated into concrete engineering specifications. The Taguchi method combining with the experimental design techniques with quality loss consideration is conventionally used for off-line quality control. When volume of parameters becomes large, the application of the new approach more critical, therefore, how to identify the impact variables and find out the optimal values for them is turned out to be more difficult. In this Research, a statistical Taguchi approach and additive model were devised to evaluate the effect of various parameters and identify the optimal parameter setup values in a fabrication of microfluidics by molding process.

Most state-of-the-art image quality metrics are based on the two-step approach: local distortion/fidelity measurement and pooling. During the pooling stage, many weighting strategies have been proposed incorporating properties of the distortion itself, various masking effects and visual attention. Recently, researchers have devoted great enthusiasm and effort to the improvement of image quality assessment using visual saliency models. In this research, it is noticed that visual saliency features of both the original image and the distorted one have impacts on the process of image quality assessment. To reduce the overlapping effects, a nonlinear additive model is proposed to integrate saliency features from the original and distorted images towards improved error weighting results. Our extensive experimental studies on four publicly available image databases (LIVE, TID2008, CSIQ and A57) indicate that the proposed improved nonlinear additive model based saliency map weighting strategy constantly leads to higher prediction accuracy for image quality assessment than traditional methods.