This paper proposes a novel graph-based transductive learning algorithm based on manifold regularization. First, the manifold regularization was introduced to probabilistic discriminant model for semi-supervised class...
详细信息
This paper proposes a novel graph-based transductive learning algorithm based on manifold regularization. First, the manifold regularization was introduced to probabilistic discriminant model for semi-supervised classification task. And then a variation of the expectationmaximization （EM） algorithm was derived to solve the optimization problem, which leads to an iterative algorithm. Although our method is developed in probabilistic framework, there is no need to make assumption about the specific form of data distribution. Besides, the crucial updating formula has closed form. This method was evaluated for text categorization on two standard datasets, 20 news group and Reuters-21578. Experiments show that our approach outperforms the state-of-the-art graph-based transductive learning methods.
The step stress accelerated degradation test (SSADT) is an effective tool for assessing the reliability of highly reliable products. However, conducting an SSADT is expensive and time consuming, and the obtained SSADT...
详细信息
The step stress accelerated degradation test (SSADT) is an effective tool for assessing the reliability of highly reliable products. However, conducting an SSADT is expensive and time consuming, and the obtained SSADT data has an impact on the accuracy of the subsequent product reliability index estimations. Consequently, devising a cost-constrained SSADT plan that yields high-precision reliability estimates poses a significant challenge. This paper focuses on the optimal design of SSADT for the Tweedie exponential dispersion process with random effect (TEDR), a general degradation model capable of describing product heterogeneity. Under given budget and boundary constraints, the optimal sample size, observation frequency and observation times at each stress level are obtained by minimizing the asymptotic variance of the estimated quantile life at normal operating conditions. The sensitivity and stability of the SSADT plan are also studied, and the results indicate the robustness of the optimal plan against slight parameters fluctuations. We use the expectationmaximization (EM) algorithm to estimate TEDR parameters and reliability indicators under SSADT, providing a systematic method for obtaining the optimal SSADT plan under budget constraints. The proposed framework is illustrated using the case of LED chips data, showcasing its potential for practical application.
In this paper, an online auto-calibration method for MicroElectroMechanical Systems (MEMS) triaxial accelerometer (TA) is proposed, which can simultaneously identify the time-dependent model structure and its paramete...
详细信息
In this paper, an online auto-calibration method for MicroElectroMechanical Systems (MEMS) triaxial accelerometer (TA) is proposed, which can simultaneously identify the time-dependent model structure and its parameters during the changes of the operating environment. Firstly, the model as well as its associated cost function is linearized by a new proposed linearization approach. Then, exploiting an online sparse recursive least square (SPARLS) estimation, the unknown parameters are identified. In particular, the online sparse recursive method is based on an L-1-norm penalized expectation-maximum (EM) algorithm, which can amend the model automatically by penalizing the insignificant parameters to zero. Furthermore, this method can reduce computational complexity and be implemented in a low-cost Micro-Controller-Unit (MCU). Based on the numerical analysis, it can be concluded that the proposed recursive algorithm can calculate the unknown parameters reliably and accurately for most MEMS triaxial accelerometers available in the market. Additionally, this method is experimentally validated by comparing the output estimations before and after calibration under various scenarios, which further confirms its feasibility and effectiveness for online TA calibration. (C) 2017 Elsevier B.V. All rights reserved.
This paper develops and illustrates a new maximum-likelihood based method for the identification of Hammerstein-Wiener model structures. A central aspect is that a very general situation is considered wherein multivar...
详细信息
This paper develops and illustrates a new maximum-likelihood based method for the identification of Hammerstein-Wiener model structures. A central aspect is that a very general situation is considered wherein multivariable data, non-invertible Hammerstein and Wiener nonlinearities, and colored stochastic disturbances both before and after the Wiener nonlinearity are all catered for. The method developed here addresses the blind Wiener estimation problem as a special case. (C) 2012 Elsevier Ltd. All rights reserved.
Although pricing fraud is an important issue for improving service quality of online shopping malls, research on automatic fraud detection has been limited. In this paper, we propose an unsupervised learning method ba...
详细信息
Although pricing fraud is an important issue for improving service quality of online shopping malls, research on automatic fraud detection has been limited. In this paper, we propose an unsupervised learning method based on a finite mixture model to identify pricing frauds. We consider two states, normal and fraud, for each item according to whether an item description is relevant to its price by utilizing the known number of item clusters. Two states of an observed item are modeled as hidden variables, and the proposed models estimate the state by using an expectationmaximization (EM) algorithm. Subsequently, we suggest a special case of the proposed model, which is applicable when the number of item clusters is unknown. The experiment results show that the proposed models are more effective in identifying pricing frauds than the existing outlier detection methods. Furthermore, it is presented that utilizing the number of clusters is helpful in facilitating the improvement of pricing fraud detection performances. (C) 2013 Elsevier B.V. All rights reserved.
In this paper we consider de-interleaving a finite number of stochastic parametric sources. The sources are modeled as independent autoregressive (AR) processes. Based on a Markovian switching policy, we assume that t...
详细信息
In this paper we consider de-interleaving a finite number of stochastic parametric sources. The sources are modeled as independent autoregressive (AR) processes. Based on a Markovian switching policy, we assume that the different sources transmit signals on the same single channel, The receiver records the 1-bit quantized version of the transmitted signal and aims to identify the sequence of active sources. Once the source sequence has been identified, the characteristics (parameters) of each source are estimated. We formulate the parametric pulse train de-interleaving problem as a 1-bit quantized Markov modulated AR series, The algorithm proposed in this paper combines Hidden Markov Model (HMM) and Binary Time Series (BTS) estimation techniques. Our estimation scheme generalizes Kedem's (1980) binary time series algorithm for linear time series to Markov modulated time series. (C) 1997 Elsevier Science B.V.
Energy detector (ED) is a popular choice for distributed cooperative spectrum sensing because it does not need to be cognizant of the primary user (PU) signal characteristics. However, the conventional ED-based sensin...
详细信息
We present an approach for parameter estimation with multirate measurements, with the slow measurements having variable time delays due to laboratory analysis, and also being functions of all the states during the sam...
详细信息
We present an approach for parameter estimation with multirate measurements, with the slow measurements having variable time delays due to laboratory analysis, and also being functions of all the states during the sample collection. We formulate a particle filter-based approach under the framework of the expectation maximization algorithm to develop the estimates. The effectiveness and applicability of the proposed method are demonstrated though a simulation example, a hybrid tank experiment and an industrial case study;in each case, the slow and fast measurements are for the same variable. We show that this approach results in improved parameter estimation when the information from the delayed measurements is fused with the fast measurement information.
Slope stability prediction is of primary concern in identifying terrain that is susceptible to landslides and mitigating the damages caused by landslides. In this study, a Naive Bayes Classifier (NBC) was employed to ...
详细信息
Slope stability prediction is of primary concern in identifying terrain that is susceptible to landslides and mitigating the damages caused by landslides. In this study, a Naive Bayes Classifier (NBC) was employed to predict slope stability for a slope subjected to circular failures, based on six input factors: slope height (H), slope angle (alpha), cohesion (c), friction angle (phi), unit weight (gamma), and pore pressure ratio (r (u) ). An expectation maximization algorithm was used to perform parameter learning for the NBC with an incomplete data set of 69 slope cases. The model validation with 13 new cases shows that, when compared to the existing empirical approach, the proposed NBC model yields better performance in terms of both accuracy and applicability (i.e., the NBC allows us to determine the probability of slope stability based on any subset of the six input factors).
Background: Positron emission tomography (PET) is widely used for studying dynamic processes, such as myocardial perfusion, by acquiring data over time frames. Kinetic modeling in PET allows for the estimation of phys...
详细信息
暂无评论