实用知识获取系统PKAS的知识维护

ＰＫＡＳ是一个实用化知识获取系统，它设置了背影知识库支持知识获取。ＫＢＭ是ＰＫＡＳ的背影知识库维护子系统。在ＫＢＭ的支持下，用ＰＫＡＳ获取的规则预测油井类型精度率达到７９％。也给出ＰＫＡＳ系统中背影知识的形式。     

Burgers''''方程的特征混合有限元数值模拟

本文对Burgers'方程采用特征混合有限元方法进行数值模拟,证明了特征混合元格式的稳定性。作为数值例子,我们计算了正弦波传播和冲击波传播,通过与混合元和有限元方法的比较,说明了该方法在粘性系数逐渐减小时对锋线前沿处理的有效性。     

Recent Advances in Disordered and Nanostructured Carbon Coatings for Superl Ubricity and Wearless Sliding

Increasingly more demanding and very stringent operating conditions envisioned for future mechanical and tribological systems will certainly require new materials and coatings that are superhard and at the same time self-lubricating.For example, dry machining is a much desired practice in manufacturing sector, but it is currently very difficult to realize mainly because of high friction and severe wear losses. However, recent advances in surface engineering and coating technologies may enable design and production of novel coatings architectures that can combine superhardness with self-lubricating properties in both the disordered or nanostructured forms. Recently developed nearly frictionless carbon films, ultrananocrystalline diamond and carbide derived carbon films can dramatically lower friction and at the same time reduce wear under very harsh sliding conditions. These coatings can be formulated in such a way that they can substantially increase the load-bearing capacity of sliding surfaces and hence improve their resistance to scuffing. It is also possible to design nano-composite coatings that can form self-replenishing and-lubricating tribofilms on their sliding surfaces and thus help increase the overall lubricity of these surfaces. In this paper, an overview of recent advances in disordered and nanostructured carbon films will be presented. Specific examples will be given to demonstrate the superior performance and durability of such novel coatings under a very wide range of tribological conditions. The major emphasis is placed on super low friction carbon films. The fundamental tribological mechanisms that control their exceptional friction and wear behaviors are also discussed.     

Unpaired image to image transformation via informative coupled generative adversarial networks

We consider image transformation problems, and the objective is to translate images from a source domain to a target one. The problem is challenging since it is difficult to preserve the key properties of the source images, and to make the details of target being as distinguishable as possible. To solve this problem, we propose an informative coupled generative adversarial networks (ICoGAN). For each domain, an adversarial generator-and-discriminator network is constructed. Basically, we make an approximately-shared latent space assumption by a mutual information mechanism, which enables the algorithm to learn representations of both domains in unsupervised setting, and to transform the key properties of images from source to target.Moreover, to further enhance the performance, a weightsharing constraint between two subnetworks, and different level perceptual losses extracted from the intermediate layers of the networks are combined. With quantitative and visual results presented on the tasks of edge to photo transformation, face attribute transfer, and image inpainting, we demonstrate the ICo- GAN’s effectiveness, as compared with other state-of-the-art algorithms.     

Case studies on passivity-based stabilisation of closed sets

We present a novel control design procedure for the passivity-based stabilisation of closed sets which leverages recent theoretical advances. The procedure involves using part of the control freedom in order to enforce a detectability property, while the remaining part is used for passivity-based stabilisation. The procedure is illustrated in four case studies of path following coordination for one or two kinematic unicycles, and variations of these problems. Among other things, we present a smooth global path following controller making the unicycle converge to an arbitrary closed and strictly convex curve, and a coordinated path following controller for two unicycles.     

Distillation technology – still young and full of breakthrough opportunities

Throughout history, distillation has been the most widespread separation method. However, despite its simplicity and flexibility, distillation still remains very energy inefficient. Novel distillation concepts based on process intensification, can deliver major benefits, not just in terms of significantly lower energy use, but also in reducing capital investment and improving eco‐efficiency. While very likely to remain the separation technology of choice for the next decades, there is no doubt that distillation technology needs to make radical changes in order to meet the demands of the energy‐conscious modern society. This article aims to show that in spite of its long age, distillation technology is still young and full of breakthrough opportunities. Moreover, it provides a broad overview of the recent developments in distillation based on process intensification principles, for example heat pump assisted distillation (e.g. vapor compression or compression–resorption), heat‐integrated distillation column, membrane distillation, HiGee distillation, cyclic distillation, thermally coupled distillation systems (Petlyuk), dividing‐wall column, and reactive distillation. These developments as well as the future perspectives of distillation are discussed in the context of changes towards a more energy efficient and sustainable chemical process industry. Several key examples are also included to illustrate the astonishing potential of these new distillation concepts to significantly reduce the capital and operating cost at industrial scale. © 2013 Society of Chemical Industry     

A 360-Degree Panoramic Image Inpainting Network Using a Cube Map

Inpainting has been continuously studied in the field of computer vision. As artificial intelligence technology developed, deep learning technology was introduced in inpainting research, helping to improve performance. Currently, the input target of an inpainting algorithm using deep learning has been studied from a single image to a video. However, deep learning-based inpainting technology for panoramic images has not been actively studied. We propose a 360-degree panoramic image inpainting method using generative adversarial networks (GANs). The proposed network inputs a 360-degree equirectangular format panoramic image converts it into a cube map format, which has relatively little distortion and uses it as a training network. Since the cube map format is used, the correlation of the six sides of the cube map should be considered. Therefore, all faces of the cube map are used as input for the whole discriminative network, and each face of the cube map is used as input for the slice discriminative network to determine the authenticity of the generated image. The proposed network performed qualitatively better than existing single-image inpainting algorithms and baseline algorithms.     

一种半监督对抗鲁棒模型无关元学习方法

元学习期望训练所得的元模型在学习到的“元知识”基础上利用来自新任务的少量标注样本，仅通过较少的梯度下降步骤微调模型就能够快速适应该任务。但是，由于缺乏训练样本，元学习算法在元训练期间对现有任务过度训练时所得的分类器决策边界不够准确，不合理的决策边界使得元模型更容易受到微小对抗扰动的影响，导致元模型在新任务上的鲁棒性能降低。提出一种半监督对抗鲁棒模型无关元学习（semi-ARMAML）方法，在目标函数中分别引入半监督的对抗鲁棒正则项和基于信息熵的任务无偏正则项，以此优化决策边界，其中对抗鲁棒正则项的计算允许未标注样本包含未见过类样本，从而使得元模型能更好地适应真实应用场景，降低对输入扰动的敏感性，提高对抗鲁棒性。实验结果表明，相比ADML、R-MAML-TRADES等当下主流的对抗元学习方法，semi-ARMAML方法在干净样本上准确率较高，在MiniImageNet数据集的5-way 1-shot与5-way 5-shot任务上对抗鲁棒性能分别约提升1.8%和2.7%，在CIFAR-FS数据集上分别约提升5.2%和8.1%。     

Diffusion of innovation model of consumer behaviour - Ideas to accelerate adoption of renewable energy sources by consumer communities in India

Renewable Energy (RE) sources form a minuscule portion of India’s overall Energy consumption today. India continues to rely on fast depleting fossil fuel and expensive Oil imports to satisfy the energy demands of the economy. But this is hardly sustainable and India has to quickly get RE sources to play a major role in servicing the energy needs of its population. Despite the best efforts the adoption of RE sources by consumer communities in India is patchy. This article will focus on what needs to be done to create a pull from the market for RE sources, by looking at Consumer Behaviour literature available in the area of Diffusion of Innovation[1]. Demand for RE sources from consumer communities must reach a tipping point[4] quickly; for the sector to take-off on its own and become a self-sustaining business.     

网络入侵检测系统中的漂移检测

目前基于机器学习的入侵检测系统大都建立在入侵数据始终保持统计平稳的假设之上,无法应对攻击者有意改变数据特性或新型攻击方式的出现,而导致的检测率下降的状况.对于上述问题,即攻击漂移,提出了加权Rényi距离的检测方法.在KDD Cup99数据集上的实验证明,Rényi距离可以有效地增强检测效果;在检测到漂移后,通过重新训练模型可以使得对攻击的识别率显著提高.