Analysis of the errors produced by the 2004 BBN speech recognition system in the DARPA EARS evaluations

This paper aims to quantify the main error types the 2004 BBN speech recognition system made in the broadcast news (BN) and conversational telephone speech (CTS) DARPA EARS evaluations. We show that many of the remaining errors occur in clusters rather than isolated, have specific causes, and differ to some extent between the BN and CTS domains. The correctly recognized words are also clustered and are highly correlated with regions where the system produces a single hypothesized choice per word. A statistical analysis of some well-known error causes (out-of-vocabulary words, word fragments, hesitations, and unlikely language constructs) was performed in order to assess their contribution to the overall word error rate (WER). We conclude with a discussion of the lower bound on the WER introduced by the human annotator disagreement.     

Genetic algorithm wrapped Bayesian network feature selection applied to differential diagnosis of erythemato-squamous diseases

This paper presents a new method for differential diagnosis of erythemato-squamous diseases based on Genetic Algorithm (GA) wrapped Bayesian Network (BN) Feature Selection (FS). With this aim, a GA based FS algorithm combined in parallel with a BN classifier is proposed.Basically, erythemato-squamous dataset contains six dermatological diseases defined with 34 features. In GA–BN algorithm, GA makes a heuristic search to find most relevant feature model that increase accuracy of BN algorithm with the use of a 10-fold cross-validation strategy. The subsets of features are sequentially used to identify six dermatological diseases via a BN fitting the corresponding data. The algorithm, in this case, produces 99.20% classification accuracy in the diagnosis of erythemato-squamous diseases. The strength of feature model generated for BN is furthermore tested with the use of Support Vector Machine (SVM), Multi-Layer Perceptron (MLP), Simple Logistics (SL) and Functional Decision Tree (FT). The resultant classification accuracies of algorithms are 98.36%, 97.00%, 98.36% and 97.81% respectively. On the other hand, BN algorithm with classification accuracy of 99.20% is quite a high diagnosis performance for erythemato-squamous diseases. The proposed algorithm makes no more than 3 misclassifications out of 366 instances. Furthermore, FS power of GA is also compared with two alternative search algorithms, i.e. Best First (BF) and Sequential Floating (SF).The obtained results have all together shown that the proposed GA–BN based FS and prediction strategy is very promising in diagnosis of erythemato-squamous diseases.     

A Bayesian belief network analysis of factors influencing wildfire occurrence in Swaziland

The impacts of wildfires on ecosystems and the factors contributing to their occurrence are increasingly receiving global attention. Advances in satellite remote sensing and information technology provide an opportunity to study these complex interrelationships. A Bayesian belief network (BBN) model was developed from a set of 12 biotic, abiotic and human variables to determine factors that influence wildfire activity in Swaziland using wildfire data from the Terra and Aqua satellites' Moderate Resolution Imaging Spectroradiometer (MODIS) for the period 2001–2007. These were geospatially integrated in the geographic information system (GIS) software ArcView and input into the software Netica for BBN analyses. Land cover, elevation, and climate (mean annual rainfall and mean annual temperature) were found to be strong predictors of wildfire occurrence, while aspect had the least influence on the wildfire occurrence. The model had a high predictive accuracy with an error rate of 9.62%, and an area under the receiver-operating characteristic (ROC) curve of 0.961. The study demonstrates how domain or field knowledge and limited empirical and GIS data can be combined within a BBN model to assist in determining key fire management interventions and lays the foundation for the future development of advanced and dynamic models.     

Use of Bayesian Belief Networks when combining disparate sources of information in the safety assessment of software-based systems

The paper discusses how disparate sources of information can be combined in the safety assessment of software-based systems. The emphasis is put on an emerging methodology, relevant for intelligent product-support systems, to combine information about disparate evidences systematically based on Bayesian Belief Networks. The objective is to show the link between basic information and the confidence one can have in a system. How one combines the Bayesian Belief Net (BBN) method with a software safety standard (RTCA/DO-178B) for safety assessment of software-based systems is also discussed. Finally, the applicability of the BBN methodology and experiences from cooperative research work together with Kongsberg Defence & Aerospace and Det Norske Veritas, and ongoing research with VTT Automation are presented.     

Semantic analysis of soccer video using dynamic Bayesian network

Video semantic analysis is formulated based on the low-level image features and the high-level knowledge which is encoded in abstract, nongeometric representations. This paper introduces a semantic analysis system based on Bayesian network (BN) and dynamic Bayesian network (DBN). It is validated in the particular domain of soccer game videos. Based on BN/DBN, it can identify the special events in soccer games such as goal event, corner kick event, penalty kick event, and card event. The video analyzer extracts the low-level evidences, whereas the semantic analyzer uses BN/DBN to interpret the high-level semantics. Different from previous shot-based semantic analysis approaches, the proposed semantic analysis is frame-based for each input frame, it provides the current semantics of the event nodes as well as the hidden nodes. Another contribution is that the BN and DBN are automatically generated by the training process instead of determined by ad hoc. The last contribution is that we introduce a so-called temporal intervening network to improve the accuracy of the semantics output.     

深度学习批归一化及其相关算法研究进展

深度学习已经广泛应用到各个领域, 如计算机视觉和自然语言处理等, 并都取得了明显优于早期机器学习算法的效果. 在信息技术飞速发展的今天, 训练数据逐渐趋于大数据集, 深度神经网络不断趋于大型化, 导致训练越来越困难, 速度和精度都有待提升. 2013年, Ioffe等指出训练深度神经网络过程中存在一个严重问题: 中间协变量迁移(Internal covariate shift), 使网络训练过程对参数初值敏感、收敛速度变慢, 并提出了批归一化(Batch normalization, BN)方法, 以减少中间协变量迁移问题, 加快神经网络训练过程收敛速度. 目前很多网络都将BN作为一种加速网络训练的重要手段, 鉴于BN的应用价值, 本文系统综述了BN及其相关算法的研究进展. 首先对BN的原理进行了详细分析. BN虽然简单实用, 但也存在一些问题, 如依赖于小批量数据集的大小、训练和推理过程对数据处理方式不同等, 于是很多学者相继提出了BN的各种相关结构与算法, 本文对这些结构和算法的原理、优势和可以解决的主要问题进行了分析与归纳. 然后对BN在各个神经网络领域的应用方法进行了概括总结, 并且对其他常用于提升神经网络训练性能的手段进行了归纳. 最后进行了总结, 并对BN的未来研究方向进行了展望.     

基于贝叶斯网络的航班保障服务时间动态估计

针对航班保障服务时间估计的问题,考虑到航班保障服务流程的特殊性、复杂性以及影响因素的不确定性,提出了一种基于贝叶斯网络（BN）的航班保障服务时间估计模型。该模型把航空领域的专家知识与历史数据的机器学习相结合,使用贝叶斯网络的增量学习特性动态地调整BN模型,使其适应新的变化,进而不断更新航班保障服务时间的估计值。使用国内某大型枢纽机场信息系统内提取的数据,通过期望最大化（EM）方法对模型进行训练,得到了测试结果。实验结果分析与模型评价表明,所提方法能有效估计航班保障服务时间且具有较高的准确度。敏感性分析表明,航班到达时段的航班密度对航班保障服务时间影响最强。     

SR-30 turbojet engine real-time sensor health monitoring using neural networks,and Bayesian belief networks

This paper describes the use of artificial intelligence-based techniques for detecting and isolating sensor failures in a turbojet engine. Specifically, three artificial intelligence (AI) techniques are employed: artificial neural networks (NNs), statistical expectations, and Bayesian belief networks (BBNs). These techniques are combined into an overall system that is capable of distinguishing between sensor failure and engine failure—a critical capability in the operation of turbojet engines. The turbojet engine used in this study is an SR-30 developed by Turbine Technologies. Initially, NNs were designed and trained to recognize sensor failure in the engine. The increased random noise output from failing sensors was used as the key indicator. Next, a Bayesian statistical method was used to recognize sensor failure based on the bias error occurring in the sensors. Finally, a BBN was developed to interpret the results of the NN and statistical evaluations. The BBN determines whether single or multiple sensor failures signify engine failure, or whether sensor failures represent separate, unrelated incidences. The BBN algorithm is also used to distinguish between bias and noise errors on sensors used to monitor turbojet performance. The overall system is demonstrated to work equally well during start-up and main-stage operation of the engine. Results show that the method can efficiently detect and isolate single or multiple sensor failures within this dynamic environment.     

An efficient concentrative photovoltaic solar system with Bayesian selection of optimal solar tracking algorithms

As the importance of sustainable energy has been rapidly growing, a concentrative photovoltaic (CPV) solar system is drawing much attention. In order for a system to operate efficiently, a deliberate solar tracking system must be equipped because an optimal tilt of solar panel is changed as the Sun orbits its trajectory. However, many existing tracking methods did not clearly consider the effect of various weather conditions, even though the performance of tracking method is subject to them. In this paper, we propose a CPV solar system that chooses the most proper solar tracking method among the group of heterogeneous tracking algorithms, based on an inference on the current weather conditions with Bayesian network (BN). We use 13 features derived from image processing and implement four tracking algorithms which have relative performance depending on nine different weather conditions. We constructed the working CPV system and collected the 1630 image data every three minutes for five hours over a period of 16 days. The proposed BN shows 93.9% accuracy for inferencing weather conditions, and the proposed system shows 16.58% higher power productivity, compared to a pinhole system and other existing methods.     

Bayesian belief network for box-office performance: A case study on Korean movies

Due to their definition as experience goods with short product lifetime cycles, it is difficult to forecast the demand for motion pictures. Nevertheless, producers and distributors of new movies need to forecast box-office results in an attempt to reduce the uncertainty in the motion picture business. Previous research demonstrated the ability of certain movie attributes such as early box-office data and release season to forecast box-office revenues. However, no previous research has focused on the causal relationship among various movie attributes, which have the potential to increase the accuracy of box-office predictions. In this paper a Bayesian belief network (BBN), which is known as a causal belief network, is constructed to investigate the causal relationship among various movie attributes in the performance prediction of box-office success. Subsequently, sensitivity analysis is conducted to determine those attributes most critically related to box-office performance. Finally, the probability of a movie’s box-office success is computed using the BBN model based on the domain knowledge from the value chain of theoretical motion pictures. The results confirm the improved forecasting accuracy of the BBN model compared to artificial neural network and decision tree.     

Evaluation of Deep Learning Models for Person Authentication Based on Touch Gesture

Touch gesture biometrics authentication system is the study of user's touching behavior on his touch device to identify him. The features traditionally used in touch gesture authentication systems are extracted using hand-crafted feature extraction approach. In this work, we investigate the ability of Deep Learning (DL) to automatically discover useful features of touch gesture and use them to authenticate the user. Four different models are investigated Long-Short Term Memory (LSTM), Gated Recurrent Unit (GRU), Convolutional Neural Network (CNN) combined with LSTM (CNN-LSTM), and CNN combined with GRU(CNN-GRU). In addition, different regularization techniques are investigated such as Activity Regularizer, Batch Normalization (BN), Dropout, and LeakyReLU. These deep networks were trained from scratch and tested using TouchAlytics and BioIdent datasets for dynamic touch authentication. The result reported in terms of authentication accuracy, False Acceptance Rate (FAR), False Rejection Rate (FRR). The best result we have been obtained was 96.73%, 96.07% and 96.08% for training, validation and testing accuracy respectively with dynamic touch authentication system on TouchAlytics dataset with CNN-GRU DL model, while the best result of FAR and FRR obtained on TouchAlytics dataset was with CNN-LSTM were FAR was 0.0009 and FRR was 0.0530. For BioIdent dataset the best results have been obtained was 84.87%, 78.28% and 78.35% for Training, validation and testing accuracy respectively with CNN-LSTM model. The use of a learning based approach in touch authentication system has shown good results comparing with other state-of-the-art using TouchAlytics dataset.     

改进型的batch normalization:BNalpha

针对提高卷积神经网络(convolutional neural network,CNN)在图像识别方向的训练速度和识别准确率进行了研究.从BN(batch normalization)着手,通过新增参数对BN的仿射变换进行具体调节,并提出一种改进型的BN——BNalpha.除去带有某些特定结构的神经网络,相对于原始的BN,BNalpha可以在不增加运算复杂度的前提下,提升神经网络的训练速度和识别准确度.通过对BN仿射变换的参数进行分析和对比,尝试解释BN在网络中的运行机理,并以此说明BNalpha相对于BN的改进为何生效.最后通过CIFAR-10和CIFAR-100数据集以及不同类型的卷积神经网络结构对BNalpha和BN进行对比实验分析,实验结果表明BNalpha能够进一步提升训练速度和识别准确度.     

平衡态分子动力学 Green-Kubo 方法计算氮化硼 单层结构热导率的模型尺寸效应研究

分子动力学模拟可以直接表征体系原子的行为,因此成为研究氮化硼相关材料微观导热机理的重要工具,但目前尚没有关于氮化硼材料模型尺寸对其热传导相关性质影响规律的研究。该文采用平衡态分子动力学并结合 Green-Kubo 方法,研究了纯净氮化硼单层结构热导率、声子色散关系以及态密度随模拟尺寸的变化规律,并解释了其内部机理。实验发现,氮化硼单层材料热导率随着模拟尺寸的增大而减小,并在单层面积约 4.1 nm×4.1 nm 时收敛于(349±22)W/(m?K),此收敛值远小于平衡态分子动力学计算中石墨烯热导率的收敛尺寸(10 nm×10 nm),这说明氮化硼单层中声子之间的散射大于石墨烯。此外,不同于热导率,氮化硼单层结构的声子色散曲线、态密度几乎不受模拟尺寸的影响。该研究结果可为采用平衡态分子动力学研究氮化硼相关材料的微观导热机理提供重要参考。     

INVESTIGATION OF THE K2 ALGORITHM IN LEARNING BAYESIAN NETWORK CLASSIFIERS

We experimentally study the K2 algorithm in learning a Bayesian network (BN) classifier for image detection of cytogenetic abnormalities. Starting from an initial BN structure, the K2 algorithm searches the BN structure space and selects the structure maximizing the K2 metric. To improve the accuracy of the K2-based BN classifier, we investigate the K2 algorithm initial ordering, search procedure, and metric. We find that BN structures learned using random initial orderings, orderings based on expert knowledge, or a scatter criterion are comparable and lead to similar classification accuracies. Replacing the K2 search with hill-climbing search improves the accuracy as does the inclusion of hidden nodes in the BN structure. Also, we demonstrate that though the maximization of the K2 metric solicits structures providing improved inference, these structures contribute to only limited classification accuracy.     

Compile-time optimization of near-neighbor communication for scalable shared-memory multiprocessors

Scalable shared-memory multiprocessor systems are typically NUMA (nonuniform memory access) machines, where the exploitation of the memory hierarchy is critical to achieving high performance. Iterative data parallel loops with near-neighbor communication account for many important numerical applications. In such loops, the communication of partial results stresses the memory system performance. In this paper, we develop data placement schemes that minimize communication time where the near-neighbor interaction is determined by a stencil. Under a given loop partition, our compile-time algorithm partitions global data into four classes for each processor, with each class requiring specific consistency maintenance requirements. The ADAPT (Automatic Data Allocation and Partitioning Tool) system was implemented to automatically partition parallel code segments for the BBN TC2000, a scalable shared-memory multiprocessor. ADAPT caches global arrays and maintains data consistency in software through instructions that flush data from private caches. Restructuring of a fluid flow code segment by ADAPT improved performance by a factor of more than 3 on the BBN TC2000. Features in current generation pipelined processors with multiple functional units permit the overlap of memory accesses with computation. Our experiments on the BBN TC2000 show that the degree of overlap is limited by architectural parameters, such as the number of CPU registers.     

基于深度学习的网络空间操作系统识别技术研究

在网络空间精准、快速、全面地进行网络资产探测是实现数字资产安全有效管理的前提,而识别操作系统是网络资产探测的基础,通过对流量中的操作系统信息的识别可以对已知漏洞进行预防范。本文主要提供了一种基于卷积神经网络的操作系统指纹快速识别方法,设计和构建了以ReLU函数作为激活函数的二层卷积模型且增加了BN层、池化层、全连接层,通过使用流量探测分析工具p0f将其指纹库操作系统指纹数据作为训练集,对收集到的流量数据作为测试集进行指纹识别测试,并将SVM方法和决策树方法与本文构建模型进行对照组实验。实验结果表明,本文操作系统识别模型具有较高的收敛速度和准确率,且平均判别准确率相比于SVM算法和C4.5决策树算法提高了13和6个百分点,证明本文研究的模型在操作系统识别方面具有良好的性能。     

基于贝叶斯网络参数迁移学习的电熔镁炉异常工况识别

在贝叶斯网络(Bayesian network,BN)参数学习中,如果数据不够充分,将无法建立准确的BN模型来分析和解决问题.针对电熔镁炉熔炼过程的异常工况识别建模,提出一种新的BN参数迁移学习方法来改进异常工况识别精度.该方法可以解决源域BN与目标域BN在结构不一致情况下的参数迁移学习问题.在实验部分,首先在著名的A...