KNOWLEDGE-BASED FEATURE DISCOVERY FOR EVALUATION FUNCTIONS

Since Samuel's work on checkers over thirty years ago, much effort has been devoted to learning evaluation functions. However, all such methods are sensitive to the feature set chosen to represent the examples. If the features do not capture aspects of the examples significant for problem solving, the learned evaluation function may be inaccurate or inconsistent. Typically, good feature sets are carefully handcrafted and a great deal of time and effort goes into refining and tuning them. This paper presents an automatic knowledge-based method for generating features for evaluation functions. The feature set is developed iteratively: features are generated, then evaluated, and this information is used to develop new features in turn. Both the contribution of a feature and its computational expense are considered in determining whether and how to develop it further.
This method has been applied to two problem-solving domains: the Othello board game and the domain of telecommunications network management. Empirical results show that the method is able to generate many known features and several novel features and to improve concept accuracy in both domains.     

连续体结构的参数化水平集统一优化

针对传统分步式结构优化设计的不足,提出一种同时进行结构拓扑、形状和尺寸统一优化的设计方法.首先采用水平集函数描述统一的结构优化模型和几何尺寸边界,通过引入紧支径向插值基函数将结构拓扑优化变量、形状优化变量和尺寸优化变量变换为基函数的扩展系数;然后取该扩展系数为设计变量,借助一种参数的变化表达3种优化要素对结构性能的影响,将复杂的多变量优化问题变换为相对简单的参数优化问题,有利于与相对成熟的优化算法相结合提高求解效率;进一步用R函数将其融合为一个整体,构造出统一优化模型,并用最优化准则法进行求解.最后通过数值案例证明了该方法的有效性和精确性.     

Correlation properties of combiners with memory in stream ciphers

For pseudo-random generators where one or several LFSRs are combined by a memoryless function, it is known that the output sequences are correlated to certain LFSR-sequences whose correlation coefficients c _t satisfy the equation _i c ₂ ⁱ = 1. In this paper it is proved that a corresponding result also holds for generators whose LFSRs are connected to a combiner with memory.If correlation probabilities are conditioned on side information, e.g., on known output digits, it is shown that new or stronger correlations may occur. This is exemplified for the summation cipher with only two LFSRs where such correlations can be exploited in a known plaintext attack. A cryptanalytic algorithm is given which is shown to be successful for LFSRs of considerable length and with arbitrary feedback connection.A preliminary version of this paper was presented at Eurocrypt '90, May 21–24, Århus, Denmark, and has appeared in the proceedings, pp. 204–213.     

Feedrate optimization and tool profile modification for the high-efficiency ball-end milling process

Trend in die/mold machining is to produce highly quailed surface using the high-speed hard machining with the ball-end cutter. The ball-end milling is, however, less efficiency than the flat end milling. It is important to optimize the feedrate that gives the maximum material removal rate constrained by an allowable surface roughness. The state-of-art of the CBN ball-end cutter technology allows increasing the tooth feed for high-speed and high-efficiency machining. However, because the spherical shape of the cutter can result in the scallop-liked cusps on the machined surface, the surface roughness consideration makes a feedrate limitation to the CBN cutter. In this paper, the optimization of the feedrate by considering the generated-scallop effect of the ball-end cutter has been studied. It was found that the tooth feed must be kept within one third of the path pick in order to keep the feed-interval scallop height not over the path-interval scallop height. Therefore, the potential capability of the CBN cutter for the larger tooth feed (i.e. high efficient) machining can not be fully exploited. It was found a notch-cut on the center of the ball-end cutter reduced the feed-interval cusp height, thus allowing an increased feedrate of more than 50% compared with the standard ball-end cutter. If the parameters of the notch-cut profile can be optimized, it is believed that the feedrate can be further increased.     

基于马尔可夫链的自适应性神经网络训练算法

提出一种基于马尔可夫链的自适应性神经网络训练方法,对传统的S型激励函数进行了改进,建立了自适应性的神经网络分类器。在假设样本中噪声服从于正态分布的情况下建立最大似然估计,通过后验概率建立马尔科夫链对样本进行训练,提高了神经网络训练速度。在轴承故障诊断中的测试结果表明,该算法可以迅速稳定地训练出神经网络,有效提高诊断的分类结果。     

Older adults’ use of smart phones: an investigation of the factors influencing the acceptance of new functions

After adopting mobile phones, most older adults use them only for calling and SMS. The purpose of this study is to extend their usage of mobile phones to new functions. To understand older adults’ requirements of mobile phones, a questionnaire was constructed and 351 Chinese older adults were recruited to complete the questionnaires. Data collected through the questionnaires were analysed using explorative factor analysis. The results revealed that older adults’ requirements were composed of 10 factors: Find a Specific Function, Awareness and Attractiveness, Readability, Personal Concern, Soft Keys and Multi-tap, Hardware Capacity, Touch Screen, Concern of Learning, Connectivity, and Social Influence. Then, from the above 10 factors, the 6 most important factors were revealed using multiple linear regression analysis. The results imply that accepting new functions is different from accepting a product. Readability and Find a Specific Function, which are critical for older adults’ acceptance of feature phones, are not determinants of their acceptance of new functions in smart phones.     

Thermoplastic extrusion—the mechanism of the formation of extrudate structure and properties

Systems processed by thermoplastic extrusion can be regarded as heterophase polymer melts of incompatible water-plasticized biopolymers. In the process of thermoplastic extrusion, proteins and polysaccharides are melted at high pressure and temperature below the temperature region of their thermal decomposition. Dispersed particles of these systems can be deformed in flow. The mixed-melt anisotropic structure, formed in flow, is fixed by rapid conversion of the melt jet that lets the extruder die from a viscous state to a rubber-like state and then to a glassy state caused by cooling and drying. Incompatibility of proteins and polysaccharides in their water-plasticized melt mixtures impacts on structure formation and texturization during thermoplastic extrusion. Presented at the 20th ISF World Congress and 83rd AOCS Annual Meeting and Expo, May 10–14, 1992, Toronto, Ontario, Canada.     

A sum of squares approach to backstepping controller synthesis for piecewise affine and polynomial systems

This paper develops a backstepping controller synthesis methodology for piecewise polynomial (PWP) systems in strict form. The main contribution of the paper is to formulate sufficient conditions for controller design for PWP systems in strict form as a sum of squares feasibility problem under the assumption that an initial control Lyapunov function exists to start the iterative backstepping procedure. This problem can then be translated into a convex SDP problem and solved by available software packages. The controller synthesis problem for PWP systems in strict feedback form is divided into two cases. The first case consists of the construction of a sum of squares polynomial control Lyapunov function for PWP systems with discontinuous vector fields. The second case addresses the construction of a PWP control Lyapunov function for PWP systems with continuous vector fields. One major advantage of the proposed method is the fact that it can handle systems with discontinuous vector fields and sliding modes. The new synthesis method is applied to several numerical examples. One of these examples offers the first convex optimization solution to piecewise affine (PWA) control of a benchmark circuit system addressed before in the literature using non‐convex PWA control solutions. Copyright © 2013 John Wiley & Sons, Ltd.     

高维力学系统的能量方程

在低维系统中,能量方程因其明晰的物理意义而得到广泛的应用,而研究高维力学系统的能量方程也同样具有理论价值和实际意义,如位移对时间的导数是速度,速度对时间的导数是加速度,这些具有明显的物理意义,而加速度的导数其物理意义就不是很清晰了,但具有理论上的意义.本文应用广义经典力学中关于广义拉格朗日函数、广义动量和广义哈密顿函数等概念,推导了高维系统的能量方程,文中举了实例具体说明新方程的应用,为力学数学系统能量方程的推广提供了一种途经.     

Construction of highly nonlinear resilient Boolean functions satisfying strict avalanche criterion

Boolean functions with high nonlinearity, high resiliency and strict avalanche criterion （SAC） play an important role in the designs of conventional cryptographic systems. In this paper, a method is proposed to construct resilient Boolean functions on n variables （n even） satisfying SAC with nonlinearity 〉 2n-1 -2n/2. A large class of cryptographic Boolean functions that were not known earlier were obtained.