Tibetan Question Generation Based on Sequence to Sequence Model

As the dual task of question answering, question generation (QG) is a significant and challenging task that aims to generate valid and fluent questions from a given paragraph. The QG task is of great significance to question answering systems, conversational systems, and machine reading comprehension systems. Recent sequence to sequence neural models have achieved outstanding performance in English and Chinese QG tasks. However, the task of Tibetan QG is rarely mentioned. The key factor impeding its development is the lack of a public Tibetan QG dataset. Faced with this challenge, the present paper first collects 425 articles from the Tibetan Wikipedia website and constructs 7,234 question–answer pairs through crowdsourcing. Next, we propose a Tibetan QG model based on the sequence to sequence framework to generate Tibetan questions from given paragraphs. Secondly, in order to generate answer-aware questions, we introduce an attention mechanism that can capture the key semantic information related to the answer. Meanwhile, we adopt a copy mechanism to copy some words in the paragraph to avoid generating unknown or rare words in the question. Finally, experiments show that our model achieves higher performance than baseline models. We also further explore the attention and copy mechanisms, and prove their effectiveness through experiments.     

An Improved End-to-End Memory Network for QA Tasks

At present, End-to-End trainable Memory Networks (MemN2N) has proven to be promising in many deep learning fields, especially on simple natural language-based reasoning question and answer (QA) tasks. However, when solving some subtasks such as basic induction, path finding or time reasoning tasks, it remains challenging because of limited ability to learn useful information between memory and query. In this paper, we propose a novel gated linear units (GLU) and local-attention based end-to-end memory networks (MemN2N-GL) motivated by the success of attention mechanism theory in the field of neural machine translation, it shows an improved possibility to develop the ability of capturing complex memory-query relations and works better on some subtasks. It is an improved end-to-end memory network for QA tasks. We demonstrate the effectiveness of these approaches on the 20 bAbI dataset which includes 20 challenging tasks, without the use of any domain knowledge. Our project is open source on github⁴.     

Improved neural network model for reverse engineering

While conventional engineering transforms engineering concepts into real parts, in reverse engineering real parts are transformed into engineering models. The construction of a surface from three-dimensional (3D) measuring data points is an important problem in reverse engineering. This paper presents a reconstruction method for the sculptured surfaces from the 3D measuring data points. The surface reconstruction scheme is presented based on a neural network. The reconstruction of the existing surfaces is realized by training the network. A series of measuring points from existing sculptured surfaces is used as a training set. Once the neural network has been trained, it serves as a geometric model to generate all the points that are needed. However, the learning rate for the neural network is relatively slow, and the learning accuracy is often unacceptably low. In this paper, to improve the performance of the neural network, a pre-processor is proposed before the input layer. The pre-processor maps the input into the larger space by generating a set of linearly independent values. The effect of the pre-processor is to increase modelling accuracy, and reduce learning time. Based on this method, experimental results are given to show that the reconstructed surfaces are faithful to the original data points. The proposed scheme is useful for regular or irregular digitized data.     

包装机械故障诊断的新方法——神经网络专家系统(PMNS)

提出了一种基于知识与模糊神经网络专家系统故障诊断方法,设计了包装机械故障诊断的神经网络专家系统的知识获得、知识库、推理机制等主要功能模块,并实现了基于改进型神经网络的包装机械诊断专家系统。     

Antinoise approximation of the lidar signal with wavelet neural networks

We propose a new, to our knowledge, denoising method for lidar signals based on a regression model and a wavelet neural network (WNN) that permits the regression model not only to have a good wavelet approximation property but also to make a neural network that has a self-learning and adaptive capability for increasing the quality of lidar signals. Specifically, we investigate the performance of the WNN for antinoise approximation of lidar signals by simultaneously addressing simulated and real lidar signals. To clarify the antinoise approximation capability of the WNN for lidar signals, we calculate the atmosphere temperature profile with the real signal processed by the WNN. To show the contrast, we also demonstrate the results of the Monte Carlo moving average method and the finite impulse response filter. Finally, the experimental results show that our proposed approach is significantly superior to the traditional methods.     

A model for social interaction between cognition and action through a key-word simulation of knowledge growth

The question of knowledge construction can be regarded as a question of cognition in relation to action.Callon and al. have suggested interactive processes mixing both cognitive and social aspects of knowledge or technology. Both actors and interactions can usually be described by texts, and namely, by words. Thus knowledge development can be described through key-words network development. The authors have made simulations for knowledge development according to a local positive feed-back rule within small sets of word associations. In comparison with real data, the simulation results are fairly good. This approach leads to a general and very simple interaction model describing knowledge development. In this model, as opposed to usual cybernetics, actors constantly change, building a common scenario in relation to a mutual definition rule.     

A Novel Bidirectional LSTM and Attention Mechanism Based Neural Network for Answer Selection in Community Question Answering

Deep learning models have been shown to have great advantages in answer selection tasks. The existing models, which employ encoder-decoder recurrent neural network (RNN), have been demonstrated to be effective. However, the traditional RNN-based models still suffer from limitations such as 1) high-dimensional data representation in natural language processing and 2) biased attentive weights for subsequent words in traditional time series models. In this study, a new answer selection model is proposed based on the Bidirectional Long Short-Term Memory (Bi-LSTM) and attention mechanism. The proposed model is able to generate the more effective question-answer pair representation. Experiments on a question answering dataset that includes information from multiple fields show the great advantages of our proposed model. Specifically, we achieve a maximum improvement of 3.8% over the classical LSTM model in terms of mean average precision.     

大型氦低温系统多变量控制

大型氦低温系统广泛应用于各类大科学装置中,运行中往往会产生热脉冲,通过负载端传导给制冷系统,对制冷系统产生热冲击。为了研究和应对热冲击,建立了一种多变量控制策略并得到了相关仿真和实验结果。首先以真实系统为基础建立了氦低温系统的动态仿真模型,同时建立了一个基于模糊神经网络的多变量协同控制策略,并将其应用在仿真液化器模型和一个真实的氦透平制冷系统上,得到了低温系统降温过程和控制过程的仿真和实验数据。仿真和实验结果显示本策略的偏差积分量为0.016 5,下降时间为102 s,上升时间为112 s。普通PID的的偏差积分量为0.026 9,下降时间为154 s,上升时间为170 s。通过仿真和实验过程的比较,验证了本文建立的动态仿真模型具有可用的精度,证明了本策略具有较好的控制效果。     

过程神经元网络的若干理论问题

章提出一种过程神经元模型,勘全入为与时间有关的函数或过程,它是传统人工神经元模型在时间域上的扩展。基于这种过程神经元模型,给出了一种仅含一个隐层的前馈型过程神经网络模型,即基展开过程神经元网络模型。证明了相庆的连续性定理,逼近定理,计算能力定理等。     

基于模糊神经网络的公差-成本模型研究

研究了面向制造环境的公差-成本模型。对制造环境中有关的加工因素进行分类及模糊化处理,构造了加工因素的成本模糊影响系数,并以此系数和零件公差作为输入,建立了基于模糊神经网络的公差-成本模型。此模型在表征加工成本、公差关系上的精度较高,更适应面向制造公差设计的要求。     

基于BP神经网络的压电陶瓷蠕变预测

压电陶瓷驱动器的蠕变误差随时间呈现非线性变化,难以实时修正。提出基于BP神经网络的压电陶瓷蠕变预测方法,使用压电陶瓷驱动系统采集数据,对数据进行归一化处理,通过实验设计BP神经网络的隐含层数、隐含层节点数、节点转移函数和训练函数,构建BP神经网络预测模型,建立压电陶瓷蠕变与时间的关系。用BP神经网络模型对压电陶瓷蠕变进行了预测仿真,并将结果与实测数据进行了对比。结果表明,蠕变预测结果与实验数据的最大绝对误差均小于0.1 μm,最大蠕变误差均不超过0.6%,最大均方误差仅为0.0021,可见,BP预测模型具有较高的预测精度,可作为预测压电陶瓷蠕变误差的一种有效手段。     

基于小波基模糊神经网络的漂浮基柔性空间机械臂轨迹跟踪控制

讨论了载体位置不受控、姿态受控情况下,参数未知漂浮基柔性空间机械臂关节空间协调运动的轨迹跟踪控制问题。依据系统动量守恒关系和拉格朗日第二类方程,由假设模态法,建立了漂浮基柔性空间机械臂的系统动力学方程。以此为基础,针对系统参数未知的情况,设计了一种小波基模糊神经网络的控制方案,以使柔性空间机械臂的载体姿态到达期望位置的同时,机械臂关节铰能够协调地跟踪关节空间的期望轨迹。该方案具有无需预知柔性空间机械臂的模型和系统参数的优点,且网络权值是通过在线学习进行调整,节省了离线训练的时间。系统的数值仿真表明了所提出方案的有效性和可行性。     

基于误差修正 NLSTM 神经网络的无人机航迹预测

无人机产业近年来发展迅猛,在军用和民用方面都拥有广泛的应用前景。无人机的航迹记录在其航行过程中发挥着重要作用,无人机的航迹预测也成为当前世界研究的热点,使用神经网络进行航迹预测更可以充分发挥其优势。首先对国内外学者关于航迹预测的文献进行了梳理,根据航迹预测的原理对目前飞行器航迹预测算法进行了总结和分类,针对利用神经网络模型预测无人机航迹并逐步改进模型以提高预测精度的问题进行了研究。接着对于传统神经网络模型预测精度不够高的问题,提出一种带误差修正的嵌套长短期记忆 (ENLSTM) 神经网络预测模型。ENLSTM 在嵌套长短期记忆网络模型的基础上引入了误差修正项,从而使得预测精度更高。最后使用 BP、RNN、LSTM 和 ENLSTM 四种神经网络模型分别对无人机的真实航迹数据和模拟航迹数据进行仿真实验,得出结论：循环神经网络相对 BP 神经网络在无人机航迹的预测上更具有优势,基于基础循环神经网络的逐步改进提升了模型的预测能力,ENLSTM 模型对于无人机的航迹预测具有更好的效果。     

Prediction of die casting process parameters by using an artificial neural network model for zinc alloys

Pressure die casting is an important production process. In pressure die casting, the first setting of process parameters is established through guess work. Experts use their previous experience and knowledge to develop a solution for a new application. Due to rapid expansion in the die casting process to produce better quality products in a short period of time, there is ever increasing demand to replace the time-consuming and expert-reliant traditional trial and error methods of establishing process parameters. A neural network system is developed to generate the process parameters for the pressure die casting process. The system aims to replace the existing high-cost, time-consuming and expertdependent trial and error approach for determining the process parameters. The scope of this work includes analysing a physical model of the pressure die casting filling stage based on governing equations of die cavity filling and the collection of feasible casting data for the training of the network. The training data were generated by using ZN-DA3 material on a hot chamber die casting machine with a plunger diameter of 60 mm. The present network was developed using the MATLAB application toolbox. In this work, the neural network was developed by comparing three different training algorithms: i.e. error backpropagation algorithm; momentum and adaptive learning algorithm; and Levenberg-Marquardt approximation algorithm. It was found that the Levenberg-Marquardt approximation algorithm was the preferred method for this application as it reduced the sum-squared error to a small value. The accuracy of the developed network was tested by comparing the data generated from the network with those of an expert from a local die casting industry. It was established that by using this network the selection of process parameters becomes much easier, so that it can be used by a novice user without prior knowledge of the die casting process or optimization techniques.     

基于EBP模型的服务化制造业集群知识共享风险评价

通过与传统制造业集群的对比,分析了服务化制造产业集群的知识共享过程,识别了服务化制造业集群知识共享过程中关系、利益与知识共享能力三方面的知识流并生风险,构建了基于EBP（熵值法和BP神经网络法）的服务化制造业集群知识共享风险组合评价模型。通过实证分析验证了模型的有效性与合理性,并通过与BP神经网络模型的误差对比分析,证实了该模型在知识共享风险识别方面的准确度更高。所提供的模型方法可为服务化制造业集群企业知识共享风险管理提供决策支持。     

A hybrid neural network method for simulating spatial variation in earthquake ground motion

Different excitations for supports should be considered for the analysis of long-span structures. The excitation of each support has time delay and spatial variation relative to other support excitations. The present study aims to propose a new method for simulating accelerograms for various distances considering spatial variation of earthquake records. The accelerograms are simulated based on response or design spectra using the learning capabilities of neural networks. In this method, the response spectrum, and the distance parameter (distance from fault rupture) are the input, and the corresponding accelerograms are the output of the network. There are three stages involved in this study. In the first stage, a replicator neural network is used as a data compressor to increase capability of the simulation. In the second stage, a radial basis function neural network is employed to generate a compressed accelerogram for a certain distance and a response spectrum. In the third stage, the compressed acceleration data is decompressed to resemble real earthquake records. Recorded accelerograms of the strong motion array in Taiwan are used to train the artificial neural network. The obtained results show the robustness of the applied method in producing spatially varying accelerograms. Finally, compatible accelerograms of the design spectrum, suggested in the Taiwan building seismic design code, are simulated for different distances with the proposed method.     

Deep Learning Enabled Object Detection and Tracking Model for Big Data Environment

Recently, big data becomes evitable due to massive increase in the generation of data in real time application. Presently, object detection and tracking applications becomes popular among research communities and finds useful in different applications namely vehicle navigation, augmented reality, surveillance, etc. This paper introduces an effective deep learning based object tracker using Automated Image Annotation with Inception v2 based Faster RCNN (AIA-IFRCNN) model in big data environment. The AIA-IFRCNN model annotates the images by Discriminative Correlation Filter (DCF) with Channel and Spatial Reliability tracker (CSR), named DCF-CSRT model. The AIA-IFRCNN technique employs Faster RCNN for object detection and tracking, which comprises region proposal network (RPN) and Fast R-CNN. In addition, inception v2 model is applied as a shared convolution neural network (CNN) to generate the feature map. Lastly, softmax layer is applied to perform classification task. The effectiveness of the AIA-IFRCNN method undergoes experimentation against a benchmark dataset and the results are assessed under diverse aspects with maximum detection accuracy of 97.77%.     

High-performance current control for switched reluctance motors based on on-line estimated parameters

To reduce torque ripple in a switched reluctance motor (SRM) by current profiling, a high-performance current controller is necessary. This study presents a high-performance current controller for SRM drives. A Bspline neural network is used to model the non-linearity of the SRM and estimate back electromotive force (EMF) and incremental inductance on-line in real time. The on-line modelling scheme does not require a priori knowledge of the machine?s electromagnetic characteristics. Based on the on-line estimated parameters, a current controller with adjustable PI gains and back-EMF decoupling is implemented. The performance of the current controller has been demonstrated in simulation and experimentally using a four-phase 8/6 550 W SRM drive system.     

Analog fault diagnosis of actual circuits using neural networks

We have developed a neural-network based analog fault diagnostic system for actual circuits. Our system uses a data acquisition board to excite a circuit with an impulse and sample its output to collect training data for the neural network. The collected data is preprocessed by wavelet decomposition, normalization, and principal component analysis (PCA) to generate optimal features for training the neural network. This ensures a simple architecture for the neural network and minimizes the size of the training set required for its proper training. Our studies indicate that features extracted from actual circuits lie closer to each other and exhibit more overlap across fault classes compared to SPICE simulations. This implies that the neural network architecture which can most reliably perform fault diagnosis of actual circuits is one whose outputs estimate the probabilities that input features belong to different fault classes. Our work also shows that SPICE simulations can be used to select appropriate features for training the neural network. Reliable diagnosis of faults in an actual circuit, however, requires training data from the circuit itself. Our fault diagnostic system, trained and tested using data obtained from real sample circuits, achieves 95% accuracy in classifying faulty components     

基于粒子群的神经网络测试生成算法

随着集成电路的规模和复杂度的不断提高,高效地生成测试矢量已成为数字电路板故障检测的关键所在。在对测试向量自动生成问题分析的基础上,利用神经网络对被测电路进行数学建模,将测试矢量生成转化为数学问题,提出了一种高效求解该问题的粒子群优化算法。用VC++对所提出的方法进行编程实现,并对ISCAS’85国际标准电路中的一些电路进行了实验。实验数据表明,故障覆盖率达到了100%,对于小规模电路单故障的测试时间与有关文献相比减少了13%,规模相对较大的电路的测试时间减少了61%,而且电路规模越大,时间的减少就越明显。