基于分段加权的反向稀疏跟踪算法研究

为提高稀疏表示跟踪模型性能，提出一种分段加权的反向稀疏跟踪算法，将跟踪问题转化为在贝叶斯框架下寻找概率最高的候选对象问题，构造不同的分段权重函数来分别度量候选目标与正负模板的判别特征系数。通过池化来降低跟踪结果的不确定性干扰，选择正负模板加权系数差值最大的候选表示作为跟踪结果。实验表明，在光照变化、遮挡、快速运动、运动模糊情况下，所提出的算法可以确保跟踪结果的准确性和鲁棒性。     

Mixture ratio modeling of dynamic systems

Any knowledge extraction relies (possibly implicitly) on a hypothesis about the modelled-data dependence. The extracted knowledge ultimately serves to a decision-making (DM). DM always faces uncertainty and this makes probabilistic modelling adequate. The inspected black-box modeling deals with “universal” approximators of the relevant probabilistic model. Finite mixtures with components in the exponential family are often exploited. Their attractiveness stems from their flexibility, the cluster interpretability of components and the existence of algorithms for processing high-dimensional data streams. They are even used in dynamic cases with mutually dependent data records while regression and auto-regression mixture components serve to the dependence modeling. These dynamic models, however, mostly assume data-independent component weights, that is, memoryless transitions between dynamic mixture components. Such mixtures are not universal approximators of dynamic probabilistic models. Formally, this follows from the fact that the set of finite probabilistic mixtures is not closed with respect to the conditioning, which is the key estimation and predictive operation. The paper overcomes this drawback by using ratios of finite mixtures as universally approximating dynamic parametric models. The paper motivates them, elaborates their approximate Bayesian recursive estimation and reveals their application potential.     

Transport Map sampling with PGD model reduction for fast dynamical Bayesian data assimilation

The motivation of this work is to address real-time sequential inference of parameters with a full Bayesian formulation. First, the proper generalized decomposition (PGD) is used to reduce the computational evaluation of the posterior density in the online phase. Second, Transport Map sampling is used to build a deterministic coupling between a reference measure and the posterior measure. The determination of the transport maps involves the solution of a minimization problem. As the PGD model is quasi-analytical and under a variable separation form, the use of gradient and Hessian information speeds up the minimization algorithm. Eventually, uncertainty quantification on outputs of interest of the model can be easily performed due to the global feature of the PGD solution over all coordinate domains. Numerical examples highlight the performance of the method.     

Integration of imitation learning using GAIL and reinforcement learning using task-achievement rewards via probabilistic graphical model

The integration of reinforcement learning (RL) and imitation learning (IL) is an important problem that has long been studied in the field of intelligent robotics. RL optimizes policies to maximize the cumulative reward, whereas IL attempts to extract general knowledge about the trajectories demonstrated by experts, i.e, demonstrators. Because each has its own drawbacks, many methods combining them and compensating for each set of drawbacks have been explored thus far. However, many of these methods are heuristic and do not have a solid theoretical basis. This paper presents a new theory for integrating RL and IL by extending the probabilistic graphical model (PGM) framework for RL, control as inference. We develop a new PGM for RL with multiple types of rewards, called probabilistic graphical model for Markov decision processes with multiple optimality emissions (pMDP-MO). Furthermore, we demonstrate that the integrated learning method of RL and IL can be formulated as a probabilistic inference of policies on pMDP-MO by considering the discriminator in generative adversarial imitation learning (GAIL) as an additional optimality emission. We adapt the GAIL and task-achievement reward to our proposed framework, achieving significantly better performance than policies trained with baseline methods.     

Superior foetal electrocardiogram signal elicitation using a novel artificial intelligent Bayesian methodology

Innumerable casualties due to intrauterine hypoxia are a major worry during prenatal phase besides advanced patient monitoring with latest science and technology. Hence, the analysis of foetal electrocardiogram (fECG) signals is very vital in order to evaluate the foetal heart status for timely recognition of cardiac abnormalities. Regrettably, the latest technology in the cutting edge field of biomedical signal processing does not seem to yield the desired quality of fECG signals required by physicians, which is the major cause for the pathetic condition. The focus of this work is to extort non-invasive fECG signal with highest possible quality with a motive to support physicians in utilizing the methodology for the latest intrapartum monitoring technique called STAN (ST analysis) for forecasting intrapartum foetal hypoxia. However, the critical quandary is that the non-invasive fECG signals recorded from the maternal abdomen are affected by several interferences like power line interference, baseline drift interference, electrode motion interference, muscle movement interference and the maternal electrocardiogram (mECG) being the dominant interference. A novel hybrid methodology called BANFIS (Bayesian adaptive neuro fuzzy inference system) is proposed. The BANFIS includes a Bayesian filter and an adaptive neuro fuzzy filter for mECG elimination and non-linear artefacts removal to yield high quality fECG signal. Kalman filtering frame work has been utilized to estimate the nonlinear transformed mECG component in the abdominal electrocardiogram (aECG). The adaptive neuro fuzzy filter is employed to discover the nonlinearity of the nonlinear transformed version of mECG and to align the estimated mECG signal with the maternal component in the aECG signal for annulment. The outcomes of the investigation by the proposed BANFIS system proved valuable for STAN system for efficient prediction of foetal hypoxia.     

Prediction of rock mass P wave velocity using blasthole drilling information

In this paper, the development of the models for the prediction of rock mass P wave velocity is presented. For model development, the database of 53 cases including widely used and recorded drilling parameters and P wave velocity was constructed from the field studies conducted in 13 open pit lignite mines. Both conventional linear, non-linear multiple regression and Adaptive Neuro Fuzzy Inference System (ANFIS) were used for model development. Prediction performance indicators showed that ANFIS model presented the best performance and it can successfully be used for the preliminary prediction of P wave velocities of rock masses.     

Model Selection for Unsupervised Learning of Visual Context

This study addresses the problem of choosing the most suitable probabilistic model selection criterion for unsupervised learning of visual context of a dynamic scene using mixture models. A rectified Bayesian Information Criterion (BICr) and a Completed Likelihood Akaike’s Information Criterion (CL-AIC) are formulated to estimate the optimal model order (complexity) for a given visual scene. Both criteria are designed to overcome poor model selection by existing popular criteria when the data sample size varies from small to large and the true mixture distribution kernel functions differ from the assumed ones. Extensive experiments on learning visual context for dynamic scene modelling are carried out to demonstrate the effectiveness of BICr and CL-AIC, compared to that of existing popular model selection criteria including BIC, AIC and Integrated Completed Likelihood (ICL). Our study suggests that for learning visual context using a mixture model, BICr is the most appropriate criterion given sparse data, while CL-AIC should be chosen given moderate or large data sample sizes.     

Accurate estimation of capacities and cross-traffic of all links in a path using ICMP timestamps

The estimation of the link capacity and its available bandwidth in an end-to-end path is crucial for network management, admission control and flow control for adaptive applications. This paper introduces an estimation mechanism able to accurately estimate the available bandwidth of all links in an end-to-end path, through its capacity and cross-traffic estimation. The estimation procedures resort to the dispersions of packet pairs and trains using the concepts of ICMP Timestamp and Traceroute that, efficiently used together, enable the measurement of the dispersions in all links in the path. These mechanisms were evaluated through simulation experiments, where we analyzed the influence of the several network parameters on each estimation mechanism. The results show that these estimation methods are able to accurately estimate both the capacity and the cross-traffic of all links in a path with moderate length and with low overhead.     

The Person Reference Process in Spontaneous Trait Inferences.

Five studies examined whether spontaneous trait inferences uniquely reference the person who performed a trait-implying behavior. On each study trial in 5 studies, participants saw 2 faces and a behavioral sentence referring to one of them. Later, participants saw face-trait pairs and indicated whether they had seen the trait word in the sentence presented with the face. Participants falsely recognized implied traits more when these traits were paired with actors' faces than with control faces. This effect was replicated for a large set effaces (120), after a week delay between study and recognition test, when equal attention was paid to each face, and when the orientation of the face at recognition was different from the orientation at encoding. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

基于贝叶斯理论的城市供水管网泄漏在线检测与定位

许多城市的供水管网都安装了SCADA(Supervisory Control And Data Acquisition)系统,基于SCADA系统监测到的数据可对管网泄漏进行检测。通过引进统计学概率论中一基本定理——贝叶斯定理来建立管网泄漏的在线检测与定位模型,一定程度上解决了水力模拟误差、测量误差、测点配置等因素导致的不确定性问题。模型通过实例考核取得较为满意的诊断结果。