Analysis of Knowledge Representations in Cascade Correlation Networks

Feed-forward neural network models approximate nonlinear functions connecting inputs to outputs. The cascade correlation (CC) learning algorithm allows networks to grow dynamically starting from the simplest network topology to solve increasingly more difficult problems. It has been demonstrated that the CC network can solve a wide range of problems including those for which other kinds of networks (e.g., back-propagation networks) have been found to fail. In this paper we show the mechanism and characteristics of nonlinear function learning and representations in CC networks, their generalization capabilities, the effects of environmental bias, etc., using a variety of knowledge representation analysis tools.

     

Representations in hydrodynamics

Summary An elementary matrix inversion technique is employed to construct Galerkin type representations for the creeping motion equations of fluid mechanics. These representations lead to the fundamental singular solutions which together with a reciprocal theorem yield integral representations for the velocity vector and the pressure distribution. Some of the results agree with known solutions.

---

Darstellungen in der Hydrodynamik

Zusammenfassung Eine elementare Matrix-Inversions-Methode wird angewendet, um Galerkin-Darstellungen für die Gleichungen der Kriechbewegung der Flüssigkeitsmechanik zu konstruieren. Diese Darstellungen führen zu den fundamentalen singulären Lösungen, die zusammen mit einem entsprechenden Theorem Integraldarstellungen für den Geschwindigkeitsvektor und die Druckverteilung liefern. Einige der erhaltenen Resultate stimmen mit den bekannten Lösungen überein.

     

Representations in elastic dielectrics

The method of associated matrices is employed to obtain Galerkin type representations for the equations of motion of linear elastic dielectrics. Fourier transforms are applied to the equations written in terms of stress function to construct matrices of fundamental solutions for the problems of concentrated body force, electric force and the charge density. The singular matrices and the reciprocity relation yield integral representations for the displacement vector, polarization vector, and the potential fields.     

Interpreting Data

We consider Bayesian prediction intervals that contain a proportion of a finite number of observations with a specified probability. Such intervals arise in numerous applied contexts and are closely related to tolerance intervals. Several examples are provided to illustrate this methodology, and simulation studies are used to demonstrate potential pitfalls of using tolerance intervals when prediction intervals are required.     

汉英口译中重复词语的翻译

文章在比较了英汉两种语言形式的不同的基础上，重点分析了在汉英口译过程中出现的重复词语的现象。由于汉语语言形式的特殊性，汉语比英语更耐重复。文章结合大量译例，探讨了简化、变译、省略等三种处理口译中汉语重复词语的方法，并据此提出译员在平时的训练中应该注意的问题和训练的重点。     

Representations of transatlantic telegraphy

The successful laying of a telegraph cable across the Atlantic in 1866 had a major impact on Victorian society. An important part of the process of habituating the public to the new technology was the representation of submarine telegraphy in books and magazines, such as Punch and the Illustrated London News, aimed at an educated, disinterested audience. This paper examines some typical representations in prose, poetry, and pictures     

工程物体线图的计算机解释

在Sugihara方法的基础上提出一种计算机自动解释工程物体线图的迭代方法。该方法根据线图中隐含的几何元素间的约束关系,从有误差的工程物体画隐线图恢复平面立体的三维参数,并对画隐线图中有误差的节点位置进行校正,使得修正后的线图是一个工程物体的投影线图,并且尽量和原输入线图一致。实验结果表明该方法是有效的。     

New Ways of Transforming Subjective-Measurement Representations

Subjective measurements are considered, which are presented in the form of proportions of total and relative values of the utility parameter. New methods are proposed for the forward and inverse transformation of these forms of subjective measurement. They can be used in preparing data for decisions and in supporting models for correcting the subjective evaluations. They extend the region in which one can use transitive closure methods and methods of matching subjective measurements.     

奇异不可约M-矩阵的等价表示

给出Z-矩阵为奇异不可约M-矩阵的若干充要条件。     

Interpreting the Evidence: Elemental Analysis in the SEM

Failure investigation often involves artifact examination in a scanning electron microscope (SEM). In turn, SEM examination often includes elemental analysis of the surfaces examined, e.g., deposits, a fracture surface, a polished cross section, or other surface. However, elemental data can be misinterpreted without a basic understanding of how they are obtained and the inherent limitations and assumptions of the methods employed. Recent experiences of the authors reinforce the importance of these points.     

Use of Half-Normal Plots in Interpreting Factorial Two-Level Experiments

Plotting the empirical cumulative distribution of the usual set of orthogonal contrasts computed from a 2 ^p experiment on a special grid may aid in its criticism and interpretation. Bad values, heteroscedasticity, dependence of variance on mean, and some types of defective randomization, all leave characteristic stigmata. The halfnormal plot can be used to estimate the error standard deviation and to make judgments about the reality of the observed effects. An accompanying paper by A. Birnbaum gives some operating characteristics of these judgments. Examples are given of the use of half-normal plots in each of these ways.     

Object Recognition in X-ray Testing Using Adaptive Sparse Representations

In recent years, X-ray screening systems have been used to safeguard environments in which access control is of paramount importance. Security checkpoints have been placed at the entrances to many public places to detect prohibited items such as handguns and explosives. Human operators complete these tasks because automated recognition in baggage inspection is far from perfect. Research and development on X-ray testing is, however, ongoing into new approaches that can be used to aid human operators. This paper attempts to make a contribution to the field of object recognition by proposing a new approach called Adaptive Sparse Representation (XASR+). It consists of two stages: learning and testing. In the learning stage, for each object of training dataset, several patches are extracted from its X-ray images in order to construct representative dictionaries. A stop-list is used to remove very common words of the dictionaries. In the testing stage, test patches of the test image are extracted, and for each test patch a dictionary is built concatenating the ‘best’ representative dictionary of each object. Using this adapted dictionary, each test patch is classified following the Sparse Representation Classification methodology. Finally, the test image is classified by patch voting. Thus, our approach is able to deal with less constrained conditions including some contrast variability, pose, intra-class variability, size of the image and focal distance. We tested the effectiveness of our method for the detection of four different objects. In our experiments, the recognition rate was more than 97 % in each class, and more than 94 % if the object is occluded less than 15 %. Results show that XASR+ deals well with unconstrained conditions, outperforming various representative methods in the literature.     

Interpreting anomalously low voltage noise in two-channel measurement systems

In this work we 1) analyze and give a theoretical explanation for the anomalously low cross-spectral density of voltage fluctuations that is observed when two thermal noise sources with matched intensities are coupled to the inputs of two-channel phase modulation (PM) or amplitude modulation (AM) noise measurement systems (NMS), 2) empirically evaluate spectral resolutions of different types of measurement systems, and 3) discuss noise measurement techniques involving cross-correlation signal processing. Our work shows that the statistical uncertainty, which sets the ultimate spectral resolution in the thermal noise limited regime, is approximately the same for both systems. However, in practical terms, the non-stationary nature of the noise, the temporal separation of calibration and measurement, and the difficulty of reproducing the calibrations for two measurements make it extremely difficult to resolve noise that is more than 10 dB below the noise floor in a single channel NMS. In a two-channel NMS, however, the calibrations of the two channels are carried out simultaneously, and one can take full advantage of a large number of averages and make reproducible noise measurements with resolution 10 dB below the noise floor of a single channel NMS     

Interpreting top-down mass spectra using spectral alignment

Recent advances in mass spectrometry instrumentation, such as FTICR and OrbiTrap, have made it possible to generate high-resolution spectra of entire proteins. While these methods offer new opportunities for performing "top-down" studies of proteins, the computational tools for analyzing top-down data are still scarce. In this paper we investigate the application of spectral alignment to the problem of identifying protein forms in top-down mass spectra (i.e., identifying the modifications, mutations, insertions, and deletions). We demonstrate how spectral alignment efficiently discovers protein forms even in the presence of numerous modifications and how the algorithm can be extended to discover positional isomers from spectra of mixtures of isobaric protein forms.