Retrieval and chaos in extremely dilutedQ-Ising neural networks

Using a probabilistic approach, the deterministic and the stochastic parallel dynamics of aQ-Ising neural network are studied at finiteQ and in the limitQ. Exact evolution equations are presented for the first time-step. These formulas constitute recursion relations for the parallel dynamics of the extremely diluted asymmetric versions of these networks. An explicit analysis of the retrieval properties is carried out in terms of the gain parameter, the loading capacity, and the temperature. The results for theQ network are compared with those for theQ=3 andQ=4 models. Possible chaotic microscopic behavior is studied using the time evolution of the distance between two network configurations. For arbitrary finiteQ the retrieval regime is always chaotic. In the limitQ the network exhibits a dynamical transition toward chaos.     

The least element property of center location on tree networks with applications to distance and precedence constrained problems

In the classicalp-center location model on a network there is a set of customers, and the primary objective is to selectp service centers that will minimize the maximum distance of a customer to a closest center. Suppose that thep centers receive their supplies from an existing central depot on the network, e.g. a warehouse. Thus, a secondary objective is to locate the centers that optimize the primary objective as close as possible to the central depot. We consider tree networks and twop-center models. We show that the set of optimal solutions to the primary objective has a semilattice structure with respect to some natural ordering. Using this property we prove that there is ap-center solution to the primary objective that simultaneously minimizes every secondary objective function which is monotone nondecreasing in the distances of thep centers from the existing central depot.Restricting the location models to a rooted path network (real line) we prove that the above results hold for the respective classicalp-median problems as well.     

Rigorous results on the thermodynamics of the dilute Hopfield model

We study the Hopfield model of an autoassociative memory on a random graph onN vertices where the probability of two vertices being joined by a link isp(N). Assuming thatp(N) goes to zero more slowly thanO(1/N), we prove the following results: (1) If the number of stored patternsm(N) is small enough such thatm(N)/Np(N) 0, asN, then the free energy of this model converges, upon proper rescaling, to that of the standard Curie-Weiss model, for almost all choices of the random graph and the random patterns. (2) If in additionm(N) < ln N/ln 2, we prove that there exists, forT< 1, a Gibbs measure associated to each original pattern, whereas for higher temperatures the Gibbs measure is unique. The basic technical result in the proofs is a uniform bound on the difference between the Hamiltonian on a random graph and its mean value.     

Crosslinking of vinyl‐terminated biphenyl and naphthalene side‐chain liquid‐crystalline poly(epichlorohydrin) derivatives

We performed the crosslinking of vinyl‐terminated biphenyl and naphthalene side‐chain liquid‐crystalline polyethers using peroxide‐type initiators with and without the addition of tertiary amine promoters. The crosslinking temperatures were chosen in the range of mesophase stability to allow the mesophase order to be frozen. The biphenyl derivatives, with a high isotropization temperature, were crosslinked to a large extent. This led to anisotropic thermosets. To crosslink naphthalene derivatives, amine promoters were needed, but degrees of crosslinking were lower, and anisotropic elastomers were obtained. Crosslinking processes were studied by differential scanning calorimetry, polarized optical microscopy, and Fourier transform infrared spectroscopy. The nature of the frozen mesophase was confirmed by X‐ray diffraction studies on mechanically oriented samples. © 2002 Wiley Periodicals, Inc. J Polym Sci Part A: Polym Chem 40: 2237–2244, 2002     

Charge calculations in molecular mechanics. Part 8 Partial atomic charges from classical calculations

Summary The CHARGE2 programme, which involves the classical calculation of both the inductive and resonance contributions to the partial atomic charges in molecules is described, and the charges and electrostatic potentials obtained presented for some illustrative examples.In substituted methanes (CH₃X, CF₃X, CCl₃X) the effects of varying the electronegativity of the substituents and the - and -substituent contributions are clearly illustrated for a variety of substituent groups X.The problems involved in the inclusion of silicon into this scheme are detailed, together with the methods of overcoming them. The partial atomic charges ( and contributions) and electrostatic potentials for some silicon oxygen compounds are presented and discussed.The partial atomic charges from CHARGE2 for all the natural amino acids as their N-acetyl, N-methyl-amides are given and compared with those obtained from the AMBER and ECEPP/2 force fields. Considerable differences in these figures are observed, with the AMBER charges consistently much larger than those from the other two methods.The CHARGE2 partial atomic charges and electrostatic potentials for the four common nucleic acids, adenine, cytosine, guanine and thymine, are given and compared with those derived from other calculations. Again there is general similarity but also there are considerable differences, with those from the AMBER force field somewhat larger than the other methods.For previous parts in this series, see Refs. 1-7.     

Miscible blends through hydrogen bonding: effects on polymer properties

Miscible blends through hydrogen bonding have been intensively studied. The effects of a variety of miscible hydrogen bonded polymer blends on properties such as thermal and thermal oxidative stability, moisture sensitivity, modulus and glass transition temperature are discussed. In addition, the preparation of semi-interpenetrating polymer networks (IPNs) and studies of the effect of crosslinking on the miscibility in hydrogen bonded polymer blends are reviewed.     

聚甲基丙烯酸甲酯单分子膜的原子力学显微镜研究

报道了利用原子力显微镜（ＡＦＭ）技术对聚甲基丙烯酸甲酯（ＰＭＭＡ）单分子膜进行分子结构水平的观察研究．由ＡＰＭ图像中得到的ＰＭＭＡ单体所占面积与用π－Ａ曲线中换算出液面上ＰＭＭＡＬａｕｇｍｕｉｒ膜中单体的面积符合得很好．从ＡＦＭ图像中可见ＰＭＭＡＬＢ膜中ＰＭＭＡ的线型碳链是与档板（ｂａｒｒｉｅｒ）平行的紧密排列，与ＬＢ提膜方向相垂直．并由此对ＰＭＭＡＬＢ膜的形成过程进行分析．此外，还观察到了在Ｓｉ基底上ＰＭＭＡ直链的卷绕伸直排列形态，初步分析了其形成原因．     

Prediction of Complexation Properties of Crown Ethers Using Computational Neural Networks

A computational neural network method was used for the prediction of stability constants of simple crown ether complexes. The essence of the method lies in the ability of a computer neural network to recognize the structure-property relationships in these host-guest systems. Testing of the computational method has demonstrated that stability constants of alkali metal cation (Na+, K+, Cs+)-crown ether complexes in methanol at 25 °C can be predicted with an average error of ±0.3 log K units based on the chemical structure of the crown ethers alone. The computer model was then used for the preliminary analysis of trends in the stabilities of the above complexes.     

基于支持向量机的羰基化合物红外光谱研究

设计了一个4层分等级分类系统，通过支持向量机技术对羰基化合物的红外光谱成功地进行了信息分类和提取，并与人工神经网络进行比较，结果表明：支持向量机对羰基类化合物红外光谱分类效果优于人工神经网络；另外详细探讨了表征酰胺类化合物的特征光谱峰片段对识圳伯仲叔酰胺的影响，证明了酰胺的N-H伸缩振动峰对识别伯仲叔酰胺贡献最大，为构建红外光谱智能解析系统提供了定量依据。     

神经网络法用于多元混酸同时测定

利用多层神经网络误差反向传播算法处理酸碱电位滴定数据，求得出多元混合酸各组分的浓度，优化了神经网络的结构和参数，测定了三组分有机酸混合样品，结果良好，平均相对偏差ＲＳＤ≤４％。