基于变分模态分解和高斯过程回归的锂离子电池剩余寿命预测方法

锂离子电池在实际工作中常处于间歇工作状态,存在容量再生现象,其性能退化呈现非单调性和随机性,无法采用传统的单一模型准确进行预测。针对上述问题,研究一种基于变分模态分解（Variational Mode Decomposition, VMD）和高斯过程回归（Gaussian Process Regression, GPR）的锂离子电池剩余寿命预测方法。首先,利用VMD将锂离子电池容量退化数据分解为一系列相对平稳的分量,并获取电池退化趋势分量及容量再生分量。然后针对不同分量的具体特性,构建合适的GPR预测模型以提高单个分量预测精度。最后,将分量预测结果叠加获取容量预测结果,进而实现电池剩余寿命预测。基于NASA研究中心锂电池容量退化数据进行实验分析,结果表明本文方法相比于直接采用GPR模型,降低了容量预测误差,并有效提高了剩余寿命预测精度。     

PROBABILISTIC MODELS FOR FOCUSED WEB CRAWLING

A focused crawler is an efficient tool used to traverse the Web to gather documents on a specific topic. It can be used to build domain‐specific Web search portals and online personalized search tools. Focused crawlers can only use information obtained from previously crawled pages to estimate the relevance of a newly seen URL. Therefore, good performance depends on powerful modeling of context as well as the quality of the current observations. To address this challenge, we propose capturing sequential patterns along paths leading to targets based on probabilistic models. We model the process of crawling by a walk along an underlying chain of hidden states, defined by hop distance from target pages, from which the actual topics of the documents are observed. When a new document is seen, prediction amounts to estimating the distance of this document from a target. Within this framework, we propose two probabilistic models for focused crawling, Maximum Entropy Markov Model (MEMM) and Linear‐chain Conditional Random Field (CRF). With MEMM, we exploit multiple overlapping features, such as anchor text, to represent useful context and form a chain of local classifier models. With CRF, a form of undirected graphical models, we focus on obtaining global optimal solutions along the sequences by taking advantage not only of text content, but also of linkage relations. We conclude with an experimental validation and comparison with focused crawling based on Best‐First Search (BFS), Hidden Markov Model (HMM), and Context‐graph Search (CGS).     

Inference Methods for CRFs with Co-occurrence Statistics

The Markov and Conditional random fields (CRFs) used in computer vision typically model only local interactions between variables, as this is generally thought to be the only case that is computationally tractable. In this paper we consider a class of global potentials defined over all variables in the CRF. We show how they can be readily optimised using standard graph cut algorithms at little extra expense compared to a standard pairwise field. This result can be directly used for the problem of class based image segmentation which has seen increasing recent interest within computer vision. Here the aim is to assign a label to each pixel of a given image from a set of possible object classes. Typically these methods use random fields to model local interactions between pixels or super-pixels. One of the cues that helps recognition is global object co-occurrence statistics, a measure of which classes (such as chair or motorbike) are likely to occur in the same image together. There have been several approaches proposed to exploit this property, but all of them suffer from different limitations and typically carry a high computational cost, preventing their application on large images. We find that the new model we propose produces a significant improvement in the labelling compared to just using a pairwise model and that this improvement increases as the number of labels increases.     

面向空气质量的时空混合预测模型

由于城市中各区域空气质量同时存在时间与空间维度上的相关性,而传统深度学习模型结构比较单一,并且难以从时空角度进行建模。针对该问题提出一种可以同时提取空气质量间复杂时空关系的STAQI模型用于空气质量预测。该模型由局部组件和全局组件构成,分别用于描述本地污染物浓度和邻近站点空气质量状况对目标站点空气质量预测产生的影响,并利用加权融合组件输出获得预测结果。在全局组件中,利用图卷积网络改进门控循环单元网络的输入部分,从而提取出输入数据中的空间特征。最后将STAQI模型与多种基准模型和变体模型进行对比。其中,STAQI模型与门控循环单元模型和全局组件变体模型相比,均方根误差（RMSE）分别下降约19%和16%。结果表明STAQI模型对于任意时间窗口都具有最佳预测性能,并且对不同目标站点的预测结果验证了该模型具有较强的泛化能力。     

面向空气质量的时空混合预测模型

由于城市中各区域空气质量同时存在时间与空间维度上的相关性，而传统深度学习模型结构比较单一，并且难以从时空角度进行建模。针对该问题提出一种可以同时提取空气质量间复杂时空关系的STAQI模型用于空气质量预测。该模型由局部组件和全局组件构成，分别用于描述本地污染物浓度和邻近站点空气质量状况对目标站点空气质量预测产生的影响，并利用加权融合组件输出获得预测结果。在全局组件中，利用图卷积网络改进门控循环单元网络的输入部分，从而提取出输入数据中的空间特征。最后将STAQI模型与多种基准模型和变体模型进行对比。其中，STAQI模型与门控循环单元模型和全局组件变体模型相比，均方根误差（RMSE）分别下降约19%和16%。结果表明STAQI模型对于任意时间窗口都具有最佳预测性能，并且对不同目标站点的预测结果验证了该模型具有较强的泛化能力。     

Validity of decision mode analysis on an ROI determination problem in multichannel fNIRS data

Region of interest (ROI) determination is necessary when using functional near-infrared spectroscopy (fNIRS) data to detect brain activity. To extract ROIs from multiple fNIRS channels, we investigated the validity of applying decision mode analysis to the fNIRS dataset. This classifies a dataset into clusters with similar features. For each cluster, the dataset is decomposed into a mean vector and a linear combination of eigenvectors. Applying this to fNIRS signals, the mean vector can be used to represent change in hemoglobin (Hb), and the eigenvectors interpreted as a signal component constructing the arbitrary signal. Characterizing these vectors by correlating them with a theoretical model of brain function aids our understanding of where Hb changes occur and what type of Hb changes reflect brain activity in fNIRS data. Decision mode analysis of fNIRS data measured during viewing stereoscopic images identified ROIs around the right inferior frontal gyrus associated with attentional control, and frontal association area associated with decision on action and prediction. Our experimental results showed that information obtained from decision mode analysis can aid quantitative and qualitative ROI determination.     

Improving branch prediction accuracy by reducing pattern history table interference

A deeply pipelined superscalar processor needs an accurate branch predictor in order to approach its performance potential. The 2-level branch predictors have been shown to achieve high prediction accuracy, yet they still suffer a significant number of mispredictions. It has been shown that a number of these mispredictions are due to interference in the pattern history tables. This paper details a method for reducing the amount of pattern history table interference by dynamically identifying some easily predictable branches and inhibiting the pattern history table update for these branches. We show that inhibiting the update in this manner reduces the amount of destructive interference in the global history variation of the 2-level branch predictor, resulting in significantly improved branch prediction accuracy for the SPEC 95 benchmarks. For example, for a 2 K Byte gshare predictor, we eliminate 38% of the mispredictions for the gcc benchmark.     

Local learning regularization networks for localized regression

Local learning algorithms use a neighborhood of training data close to a given testing query point in order to learn the local parameters and create on-the-fly a local model specifically designed for this query point. The local approach delivers breakthrough performance in many application domains. This paper considers local learning versions of regularization networks (RN) and investigates several options for improving their online prediction performance, both in accuracy and speed. First, we exploit the interplay between locally optimized and globally optimized hyper-parameters (regularization parameter and kernel width) each new predictor needs to optimize online. There is a substantial reduction of the operation cost in the case we use two globally optimized hyper-parameters that are common to all local models. We also demonstrate that this global optimization of the two hyper-parameters produces more accurate models than the other cases that locally optimize online either the regularization parameter, or the kernel width, or both. Then by comparing Eigenvalue decomposition (EVD) with Cholesky decomposition specifically for the local learning training and testing phases, we also reveal that the Cholesky-based implementations are faster that their EVD counterparts for all the training cases. While EVD is suitable for validating cost-effectively several regularization parameters, Cholesky should be preferred when validating several neighborhood sizes (the number of k-nearest neighbors) as well as when the local network operates online. Then, we exploit parallelism in a multi-core system for these local computations demonstrating that the execution times are further reduced. Finally, although the use of pre-computed stored local models instead of the online learning local models is even faster, this option deteriorates the performance. Apparently, there is a substantial gain in waiting for a testing point to arrive before building a local model, and hence the online local learning RNs are more accurate than their pre-computed stored local models. To support all these findings, we also present extensive experimental results and comparisons on several benchmark datasets.

     

Multidimensional hierarchical classification

Hierarchical classification can be seen as a multidimensional classification problem where the objective is to predict a class, or set of classes, according to a taxonomy. There have been different proposals for hierarchical classification, including local and global approaches. Local approaches can suffer from the inconsistency problem, that is, if a local classifier has a wrong prediction, the error propagates down the hierarchy. Global approaches tend to produce more complex models. In this paper, we propose an alternative approach inspired in multidimensional classification. It starts by building a multi-class classifier per each parent node in the hierarchy. In the classification phase, all the local classifiers are applied simultaneously to each instance, providing a probability for each class in the taxonomy. Then the probability of the subset of classes, for each path in the hierarchy, is obtained by combining the local classifiers results. The path with highest probability is returned as the result for all the levels in the hierarchy. As an extension of the proposal method, we also developed a new technique, based on information gain, to classifies at different levels in the hierarchy. The proposed method was tested on different hierarchical classification data sets and was compared against state-of-the-art methods, resulting in superior predictive performance and/or efficiency to the other approaches in all the datasets.     

多视角融合的时空动态图卷积网络城市交通流量预测

城市交通流量预测是构建绿色低碳、安全高效的智能交通系统的重要组成部分.时空图神经网络由于具有强大的时空数据表征能力,被广泛应用于城市交通流量预测.当前时空图神经网络在城市交通流量预测中仍存在以下两方面局限性:1)直接构建静态路网拓扑图对城市空间相关性进行表示,忽略了节点的动态交通模式,难以表达节点流量之间的时序相似性,无法捕获路网节点之间在时序上的动态关联.2)只考虑路网节点的局部空间相关性,忽略节点的全局空间相关性,无法建模交通路网中局部区域和全局空间之间的依赖关系.为打破上述局限性,本文提出了一种多视角融合的时空动态图卷积模型用于预测交通流量.首先,从静态空间拓扑和动态流量模式视角出发,构建路网空间结构图和动态流量关联图,并使用动态图卷积学习节点在两种视角下的特征,全面捕获城市路网中多元的空间相关性.其次,从局部视角和全局视角出发,计算路网的全局表示,将全局特征与局部特征融合,增强路网节点特征的表现力,发掘城市交通流量的整体结构特征.接下来,设计了局部卷积多头自注意力机制来获取交通数据的动态时间相关性,实现在多种时间窗口下的准确流量预测.最后,在四种真实交通数据上的实验结果证明了本文模型的有效性和准确性.     

Recurrent neuro-fuzzy networks for nonlinear process modeling

A type of recurrent neuro-fuzzy network is proposed in this paper to build long-term prediction models for nonlinear processes. The process operation is partitioned into several fuzzy operating regions. Within each region, a local linear model is used to model the process. The global model output is obtained through the centre of gravity defuzzification which is essentially the interpolation of local model outputs. This modeling strategy utilizes both process knowledge and process input/output data. Process knowledge is used to initially divide the process operation into several fuzzy operating regions and to set up the initial fuzzification layer weights. Process I/O data are used to train the network. Network weights are such trained so that the long-term prediction errors are minimized. Through training, membership functions of fuzzy operating regions are refined and local models are learnt. Based on the recurrent neuro-fuzzy network model, a novel type of nonlinear model-based long range predictive controller can be developed and it consists of several local linear model-based predictive controllers. Local controllers are constructed based on the corresponding local linear models and their outputs are combined to form a global control action by using their membership functions. This control strategy has the advantage that control actions can be calculated analytically avoiding the time consuming nonlinear programming procedures required in conventional nonlinear model-based predictive control. The techniques have been successfully applied to the modeling and control of a neutralization process.     

全局与局部模型对QSAR/QSPR预报能力比较

针对两组数据进行了比较讨论,试图说明在QSAR／QSPR研究中经常碰到的一个基本问题。第一组为一散布度(diver- sity)很大分子结构多样化的大样本数据;第二组则是按照分子结构相似度筛选出来的散布度较小结构相似的小样本数据。对于第一组数据,因数据集分散,全局模型难以完全描述物质结构特征与其性质之间的关系,所得回归结果很差(检验集相关系数Q2=0．68、平均预报偏差(RMSEP)=40．65)。试采用新近提出的局部懒惰回归(Local lazy regression,LLR)对其进行改善,但实际结果是局部模型的效果更差(Q2=0．60、RMSEP=45．05)。继对散布度较小且相对均匀(结构相似)的数据集用LLR方法建立局部模型,此时得到的预报结果(Q2=0．90、RMSEP=24．66)却明显优于全局模型(Q2=O．86、RMSEP=29．37)。     

Fusion of EEG response and sentiment analysis of products review to predict customer satisfaction

This paper proposes a novel multimodal framework for rating prediction of consumer products by fusing different data sources, namely physiological signals, global reviews obtained separately for the product and its brand. The reviews posted by global viewers are retrieved and processed using Natural Language Processing (NLP) technique to compute compound score considered as global rating. Also, electroencephalogram (EEG) signals of the participants were recorded simultaneously while watching different products on computer’s screen. From EEG, valence scores in terms of product rating are obtained using self-report towards each viewed product for acquiring local rating. A higher valence score corresponds to intrinsic attractiveness of the participant towards a product. Random forest based regression techniques is used to model EEG data to build a rating prediction framework considered as local rating. Furthermore, Artificial Bee Colony (ABC) based optimization algorithm is used to boost the overall performance of the framework by fusing global and local ratings. EEG dataset of 40 participants including 25 male and 15 female is recorded while viewing 42 different products available on e-commerce website. Experiment results are encouraging and suggest that the proposed ABC optimization approach can achieve lower Root Mean Square Error (RMSE) in rating prediction as compared to individual unimodal schemes.     

Multiple-step-ahead prediction in control systems with Gaussian process models and TS-fuzzy models

In this paper one-step-ahead and multiple-step-ahead predictions of time series in disturbed open loop and closed loop systems using Gaussian process models and TS-fuzzy models are described. Gaussian process models are based on the Bayesian framework where the conditional distribution of output measurements is used for the prediction of the system outputs. For one-step-ahead prediction a local process model with a small past horizon is built online with the help of Gaussian processes. Multiple-step-ahead prediction requires the knowledge of previous outputs and control values as well as the future control values. A “naive” multiple-step-ahead prediction is a successive one-step-ahead prediction where the outputs in each consecutive step are used as inputs for the next step of prediction. A global TS-fuzzy model is built to generate the nominal future control trajectory for multiple-step-ahead prediction. In the presence of model uncertainties a correction of the so computed control trajectory is needed. This is done by an internal feedback between the two process models. The method is tested on disturbed time invariant and time variant systems for different past horizons. The combination of the TS-fuzzy model and the Gaussian process model together with a correction of the control trajectory shows a good performance of the multiple-step-ahead prediction for systems with uncertainties.     

Efficiency of local models ensembles for time series prediction

Many current technological challenges require the capacity of forecasting future measurements of a phenomenon. This, in most cases, leads directly to solve a time series prediction problem. Statistical models are the classical approaches for tackling this problem. More recently, neural approaches such as Backpropagation, Radial Basis Functions and recurrent networks have been proposed as an alternative. Most neural-based predictors have chosen a global modelling approach, which tries to approximate a goal function adjusting a unique model. This philosophy of design could present problems when data is extracted from a phenomenon that continuously changes its operational regime or represents distinct operational regimes in a unbalanced manner. In this paper, two alternative neural-based local modelling approaches are proposed. Both follow the divide and conquer principle, splitting the original prediction problem into several subproblems, adjusting a local model for each one. In order to check their adequacy, these methods are compared with other global and local modelling classical approaches using three benchmark time series and different sizes (medium and high) of training data sets. As it is shown, both models demonstrate to be useful pragmatic paradigms to improve forecasting accuracy, with the advantages of a relatively low computational time and scalability to data set size.     

Design of local fuzzy models using evolutionary algorithms

The application of local fuzzy models to determine the remaining life of a unit in a fleet of vehicles is described. Instead of developing individual models based on the track history of each unit or developing a global model based on the collective track history of the fleet, local fuzzy models are used based on clusters of peers—similar units with comparable utilization and performance characteristics. A local fuzzy performance model is created for each cluster of peers. This is combined with an evolutionary framework to maintain the models. A process has been defined to generate a collection of competing models, evaluate their performance in light of the currently available data, refine the best models using evolutionary search, and select the best one after a finite number of iterations. This process is repeated periodically to automatically update and improve the overall model. To illustrate this methodology an asset selection problem has been identified: given a fleet of industrial vehicles (diesel electric locomotives), select the best subset for mission-critical utilization. To this end, the remaining life of each unit in the fleet is predicted. The fleet is then sorted using this prediction and the highest ranked units are selected. A series of experiments using data from locomotive operations was conducted and the results from an initial validation exercise are presented. The approach of constructing local predictive models using fuzzy similarity with neighboring points along appropriate dimensions is not specific to any asset type and may be applied to any problem where the premise of similarity along chosen attribute dimensions implies similarity in predicted future behavior.     

What causes users to switch from a local to a global social network site? The cultural,social, economic,and motivational factors of Facebook’s globalization

This study investigates what causes local users to switch or not to switch from a domestic to a global social network site (SNS), Facebook. In the prediction model using cultural, social, economic factors, and motives for using SNS, we found in S. Korean users that, along with entertainment motives, the expected benefit of a new global SNS was a positive predictor of transition to Facebook. The western cultural values of a global SNS and the sunk costs of using a local SNS were negative predictors of the intention to use Facebook as the main platform of online social networking. Given that global SNSs force anti-localization policies related to privacy protocols and relationship styles, the results highlight the fact that cultural values are a critical factor for resisting globalization of SNSs.     

基于有向图模型的旅游领域命名实体识别

旅游领域命名实体识别是旅游知识图谱构建过程中的关键步骤,与通用领域的实体相比,旅游文本的实体具有长度长、一词多义、嵌套严重的特点,导致命名实体识别准确率低。提出一种融合词典信息的有向图神经网络（L-CGNN）模型,用于旅游领域中的命名实体识别。将预训练词向量通过卷积神经网络提取丰富的字特征,利用词典构造句子的有向图,以生成邻接矩阵并融合字词信息,通过将包含局部特征的词向量和邻接矩阵输入图神经网络（GNN）中,提取全局语义信息,并引入条件随机场（CRF）得到最优的标签序列。实验结果表明,相比Lattice LSTM、ID-CNN+CRF、CRF等模型,L-CGNN模型在旅游和简历数据集上具有较高的识别准确率,其F1值分别达到86.86%和95.02%。     

Various approaches for constructing an ensemble of metamodels using local measures

Metamodels are approximate mathematical models used as surrogates for computationally expensive simulations. Since metamodels are widely used in design space exploration and optimization, there is growing interest in developing techniques to enhance their accuracy. It has been shown that the accuracy of metamodel predictions can be increased by combining individual metamodels in the form of an ensemble. Several efforts were focused on determining the contribution (or weight factor) of a metamodel in the ensemble using global error measures. In addition, prediction variance is also used as a local error measure to determine the weight factors. This paper investigates the efficiency of using local error measures, and also presents the use of the pointwise cross validation error as a local error measure as an alternative to using prediction variance. The effectiveness of ensemble models are tested on several problems with varying dimensionality: five mathematical benchmark problems, two structural mechanics problems and an automobile crash problem. It is found that the spatial ensemble models show better performances than the global ensemble for the low-dimensional problems, while the global ensemble is a more accurate model than the spatial ensembles for the high-dimensional problems. Ensembles based on pointwise cross validation error and prediction variance provide similar accuracy. The ensemble models based on local measures reduce cross validation errors drastically, but their performances are not that impressive in reducing the error evaluated at random test points, because the pointwise cross validation error is not a good surrogate for the error at a point.