Trading-network order formation with the help of the aggregation of specialized forecast algorithms

The problem concerning the aggregating of the forecasts of specialized expert strategies is examined using the mathematical theory of machine learning. Expert strategies are understood as the algorithms capable of successively predicting the components of a time series in the online mode. The specialized strategies can refrain from predictions at certain time instants—they make forecasts in compliance with the application area of the specific model of an object region forming their basis. An optimal algorithm whereby the forecasts of such expert strategies are aggregated into the single forecast is proposed. The algorithmic optimality consists in that, on average, its total losses are asymptotically less than those of any active prediction strategies on a set of time instants. The uppermost estimated error of the given mixing of predictions, i.e., the regret of aggregating strategies, is determined. The errors are estimated in the worst situation where no assumptions are made about the mechanism underlying the initial data source. The proposed algorithm is tested using the real information on the commodity circulation of a trading network. The numerical results and estimates of the regret are presented.     

Competitive principal component analysis for locally stationarytime series

A new unsupervised algorithm is proposed that performs competitive principal component analysis (PCA) of a time series. A set of expert PCA networks compete, through the mixture of experts (MOE) formalism, on the basis of their ability to reconstruct the original signal. The resulting network finds an optimal projection of the input onto a reduced dimensional space as a function of the input and, hence, of time. As a byproduct, the time series is both segmented and identified according to stationary regions. Examples showing the performance of the algorithm are included     

基于神经网络专家系统的推理研究

为了克服传统专家系统知识获取难、学习适应能力差、推理效率低等问题,许多专家提出将神经网络与规则专家系统相结合,构建基于神经网络的专家系统模型。文中设计了一种基于神经网络专家系统模型的混合推理机制,通过对基于神经网络推理算法、规则推理算法以及神经网络与规则的混合推理算法进行实验比较,证明本文提出的混合推理机制在改善专家系统推理准确率方面的有效性。     

Extracting and using procedural knowledge in a CAD task

The declarative and procedural knowledge, or cognitive strategies, used by expert and novice CAD operators on a 3D design task are determined in order to understand how training and management of CAD operators could affect this knowledge. Results indicate that novices were variable in performance not because of differences in declarative knowledge (on which they were trained) but because of differences in procedural knowledge (on which no training was given). The design expert could transfer procedural knowledge from other systems to the CAD system tested. The system expert could perform fast, because of the highly developed declarative knowledge, without thinking about the strategies. This research indicates the need for attention by managers of CAD systems to procedural knowledge training for system experts and novice CAD operators; declarative knowledge training can be emphasized for design experts     

Sequential prediction of individual sequences under general lossfunctions

We consider adaptive sequential prediction of arbitrary binary sequences when the performance is evaluated using a general loss function. The goal is to predict on each individual sequence nearly as well as the best prediction strategy in a given comparison class of (possibly adaptive) prediction strategies, called experts. By using a general loss function, we generalize previous work on universal prediction, forecasting, and data compression. However, here we restrict ourselves to the case when the comparison class is finite. For a given sequence, we define the regret as the total loss on the entire sequence suffered by the adaptive sequential predictor, minus the total loss suffered by the predictor in the comparison class that performs best on that particular sequence. We show that for a large class of loss functions, the minimax regret is either &thetas;(log N) or Ω(√ℒlog N), depending on the loss function, where N is the number of predictors in the comparison class andℒ is the length of the sequence to be predicted. The former case was shown previously by Vovk (1990); we give a simplified analysis with an explicit closed form for the constant in the minimax regret formula, and give a probabilistic argument that shows this constant is the best possible. Some weak regularity conditions are imposed on the loss function in obtaining these results. We also extend our analysis to the case of predicting arbitrary sequences that take real values in the interval [0,1]     

Automated breath detection on long-duration signals using feedforward backpropagation artificial neural networks

A new breath-detection algorithm is presented, intended to automate the analysis of respiratory data acquired during sleep. The algorithm is based on two independent artificial neural networks (ANN(insp) and ANN(expi)) that recognize, in the original signal, windows of interest where the onset of inspiration and expiration occurs. Postprocessing consists in finding inside each of these windows of interest minimum and maximum corresponding to each inspiration and expiration. The ANN(insp) and ANN(expi) correctly determine respectively 98.0% and 98.7% of the desired windows, when compared with 29,820 inspirations and 29,819 expirations detected by a human expert, obtained from three entire-night recordings. Postprocessing allowed determination of inspiration and expiration onsets with a mean difference with respect to the same human expert of (mean +/- SD) 34 +/- 71 ms for inspiration and 5 +/- 46 ms for expiration. The method proved to be effective in detecting the onset of inspiration and expiration in full night continuous recordings. A comparison of five human experts performing the same classification task yielded that the automated algorithm was undifferentiable from these human experts, falling within the distribution of human expert results. Besides being applicable to adult respiratory volume data, the presented algorithm was also successfully applied to infant sleep data, consisting of uncalibrated rib cage and abdominal movement recordings. A comparison with two previously published algorithms for breath detection in respiratory volume signal shows that the presented algorithm has a higher specificity, while presenting similar or higher positive predictive values.     

Privacy Based Data Publishing Model for Cloud Computing Environment

The evaluation of corporate social responsibility (CSR) performance may enhance companies’ willingness to undertake social responsibilities, so it is very important to improve the quality of CSR performance evaluation. Based on the three factors of economic performance, social performance and environmental performance, this paper proposed an improved analytic hierarchy process-back propagation (AHP-BP) neural network algorithm, and introduced the improved AHP-BP neural network algorithm into CSR performance evaluation model. In the stage of improved AHP, the model included the importance of the knowledge and experience of the experts by expert scoring, and reduced the subjective influence of expert judgment on the results by introducing a personality test scale. In the stage of BP neural network, trained models have been used for CSR performance evaluation. The results showed that the prediction result of improved AHP-BP neural network model was better than that of BP neural network model. Therefore, the improved AHP-BP neural network algorithm can be used as a good predictor for CSR performance evaluation.

     

Joint end-to-end loss-delay hidden Markov model for periodic UDP traffic over the Internet

Performance of real-time applications on network communication channels is strongly related to losses and temporal delays. Several studies showed that these network features may be correlated and exhibit a certain degree of memory such as bursty losses and delays. The memory and the statistical dependence between losses and temporal delays suggest that the channel may be well modeled by a hidden Markov model (HMM) with appropriate hidden variables that capture the current state of the network. In this paper, an HMM is proposed that shows excellent performance in modeling typical channel behaviors in a set of real packet links. The system is trained with a modified version of the Expectation-Maximization (EM) algorithm. Hidden-state analysis shows how the state variables characterize channel dynamics. State-sequence estimation is obtained by the use of Viterbi algorithm. Real-time modeling of the channel is the first step to implement adaptive communication strategies.     

基于改进DS证据理论和BN的信息安全风险评估

针对信息安全风险评估过程中专家评价意见的多样性以及不确定信息难以量化处理的问题,提出了一种基于改进的DS证据理论与贝叶斯网络(BN)结合的风险评估方法.首先,在充分研究信息安全风险评估流程和要素的基础上,建立了风险评估模型,确定风险影响因素;其次,根据评估模型并结合专家知识构建相应的贝叶斯网络模型,确定贝叶斯网络模型中的条件概率表;再次,利用基于权值分配和矩阵分析的改进DS证据理论融合多位专家对风险影响因素的评价意见;最后,根据贝叶斯网络模型的推理算法,计算被测信息系统处于不同风险等级的概率值,并对结果进行有效性分析.分析表明,将改进后的DS证据理论与贝叶斯网络应用到风险评估过程中,在一定程度上能够提高评估结果的可信度和直观性.     

基于动态云贝叶斯网络的舰艇防空目标威胁评估

针对目标威胁评估中信息表达的不确定性以及威胁评估模型专家网络结构的主观性,提出一种基于结构学习的动态云贝叶斯网络评估模型。首先,利用云模型良好的知识表达能力定量描述不确定连续性信息;其次,使用爬山算法进行结构学习,综合专家提出的网络结构构建贝叶斯网络;接着引入时间变量,将其扩展成为动态贝叶斯网络,然后用最大似然概率估计算法学习网络参数;最后,利用联合树算法对动态云贝叶斯网络进行推理评估。仿真结果表明,该模型能够有效的对观测信息进行威胁评估,具有合理性和可行性。     

数控机床多故障诊断专家系统设计

系统以数控机床的数控系统为分析对象,结合某数控公司提供的故障分析手册和积累的案例以及参照该公司故障诊断专家给出的诊断方法,学习其他的专家系统的成功经验,设计了一套数控机床图形化多故障诊断系统.文中介绍了数控机床故障诊断专家系统中的一个故障诊断的方法,它具有多故障诊断的能力,可以图形化辅助诊断过程.开发了一种新的匹配算法,它是由最近相邻算法根据本系统的实际情况而开发的一种新算法.对于多故障诊断,系统采用化多故障为单故障处理的方法.开发了图形化辅助诊断功能,提高了诊断系统的实用性.开发了针对数控机床故障诊断的自学习功能,实现了故障诊断的经验积累.     

A modular neural network vector predictor for predictive imagecoding

In this paper, we present a modular neural network vector predictor that improves the predictive component of a predictive vector quantization (PVQ) scheme. The proposed vector prediction technique consists of five dedicated predictors (experts), where each expert predictor is optimized for a particular class of input vectors. An input vector is classified into one of five classes, based on its directional variances. One expert predictor is optimized for stationary blocks, and each of the other four expert predictors are optimized to predict horizontal, vertical, 45 degrees , and 135 degrees diagonally oriented edge-blocks, respectively. An integrating unit is then used to select or combine the outputs of the experts in order to form the final output of the modular network. Therefore, no side information is transmitted to the receiver about the selected predictor or the integration of the predictors. Experimental results show that the proposed scheme gives an improvement of 1.7 dB over a single multilayer perceptron (MLP) predictor. Furthermore, if the information about the predictor selection is sent to the receiver, the improvement could be up to 3 dB over a single MLP predictor. The perceptual quality of the predicted images is also significantly improved.     

Enhancing multiple expert decision combination strategies throughexploitation of a priori information sources

The concept of combining multiple experts in a unified framework to generate a combined decision based on individual decisions delivered by the cooperating experts has been exploited in solving the problem of handwritten and machine printed character recognition. The level of performance achieved in terms of the absolute recognition performance and increased confidences associated with these decisions is very encouraging. However, the underlying philosophy behind this success is still not completely understood. The authors analyse the problem of decision combination of multiple experts from a completely different perspective. It is demonstrated that the success or failure of the decision combination strategy largely depends on the extent to which the various possible sources of information are exploited in designing the decision combination framework. Seven different multiple expert decision combination strategies are evaluated in terms of this information management issue. It is demonstrated that it is possible to treat the comparative evaluation of the multiple expert decision combination approaches based on their capability for exploiting diverse information extracted from the various sources as a yardstick in estimating the level of performance that is achievable from these combined configurations     

Expert judgment in maintenance optimization

A comprehensive method for the use of expert opinion for obtaining lifetime distributions required for maintenance optimization is proposed. The method includes procedures for the elicitation of discretized lifetime distributions from several experts, the combination of the elicited expert opinion into a consensus distribution, and the updating of the consensus distribution with failure and maintenance data. The development of the method was prompted by the lack of statistical training of the experts and the high demands on their time. The use of a discretized life distribution provides more flexibility, is more comprehendible by the experts in the elicitation stage, and greatly reduces the computation in the combination and updating stages. The methodology is Bayes, using the Dirichlet distribution as the prior distribution for the elicited discrete lifetime distribution. Methods are described for incorporating information concerning the expertise of the experts into the analysis     

Automated segmentation of multiple sclerosis lesions by model outlier detection

This paper presents a fully automated algorithm for segmentation of multiple sclerosis (MS) lesions from multispectral magnetic resonance (MR) images. The method performs intensity-based tissue classification using a stochastic model for normal brain images and simultaneously detects MS lesions as outliers that are not well explained by the model. It corrects for MR field inhomogeneities, estimates tissue-specific intensity models from the data itself, and incorporates contextual information in the classification using a Markov random field. The results of the automated method are compared with lesion delineations by human experts, showing a high total lesion load correlation. When the degree of spatial correspondence between segmentations is taken into account, considerable disagreement is found, both between expert segmentations, and between expert and automatic measurements.     

A formal methodology for acquiring and representing expert knowledge

The process of eliciting knowledge from human experts and representing that knowledge in an expert or knowledge-based system suffers from numerous problems. Not only is this process time-consuming and tedious, but the weak knowledge acquisition methods typically used (i.e., interviews and protocol analysis) are inadequate for eliciting tacit knowledge and may, in fact, lead to inaccuracies in the knowledge base. In addition, the intended knowledge representation scheme guides the acquisition of knowledge resulting in a representation-driven knowledge base as opposed to one that is knowledge-driven. In this paper, a formal methodology is proposed that employs techniques from the field of cognitive psychology to uncover expert knowledge as well as an appropriate representation of that knowledge. The advantages of such a methodology are discussed, as well as results from studies concerning the elicitation of concepts from experts and the assignment of labels to links in empirically derived semantic networks.     

Classification of anatomical structures in MR brain images using fuzzy parameters

We present an algorithm that automatically segments and classifies the brain structures in a set of magnetic resonance (MR) brain images using expert information contained in a small subset of the image set. The algorithm is intended to do the segmentation and classification tasks mimicking the way a human expert would reason. The algorithm uses a knowledge base taken from a small subset of semiautomatically classified images that is combined with a set of fuzzy indexes that capture the experience and expectation a human expert uses during recognition tasks. The fuzzy indexes are tissue specific and spatial specific, in order to consider the biological variations in the tissues and the acquisition inhomogeneities through the image set. The brain structures are segmented and classified one at a time. For each brain structure the algorithm needs one semiautomatically classified image and makes one pass through the image set. The algorithm uses low-level image processing techniques on a pixel basis for the segmentations, then validates or corrects the segmentations, and makes the final classification decision using higher level criteria measured by the set of fuzzy indexes. We use single-echo MR images because of their high volumetric resolution; but even though we are working with only one image per brain slice, we have multiple sources of information on each pixel: absolute and relative positions in the image, gray level value, statistics of the pixel and its three-dimensional neighborhood and relation to its counterpart pixels in adjacent images. We have validated our algorithm for ease of use and precision both with clinical experts and with measurable error indexes over a Brainweb simulated MR set.     

Multiexpert automatic speech recognition using acoustic and myoelectric signals

Classification accuracy of conventional automatic speech recognition (ASR) systems can decrease dramatically under acoustically noisy conditions. To improve classification accuracy and increase system robustness a multiexpert ASR system is implemented. In this system, acoustic speech information is supplemented with information from facial myoelectric signals (MES). A new method of combining experts, known as the plausibility method, is employed to combine an acoustic ASR expert and a MES ASR expert. The plausibility method of combining multiple experts, which is based on the mathematical framework of evidence theory, is compared to the Borda count and score-based methods of combination. Acoustic and facial MES data were collected from 5 subjects, using a 10-word vocabulary across an 18-dB range of acoustic noise. As expected the performance of an acoustic expert decreases with increasing acoustic noise; classification accuracies of the acoustic ASR expert are as low as 11.5%. The effect of noise is significantly reduced with the addition of the MES ASR expert. Classification accuracies remain above 78.8% across the 18-dB range of acoustic noise, when the plausibility method is used to combine the opinions of multiple experts. In addition, the plausibility method produced classification accuracies higher than any individual expert at all noise levels, as well as the highest classification accuracies, except at the 9-dB noise level. Using the Borda count and score-based multiexpert systems, classification accuracies are improved relative to the acoustic ASR expert but are as low as 51.5% and 59.5%, respectively.