Clustering in Concept Association Networks

We view association of concepts as a complex network and present a heuristic for clustering concepts by taking into account the underlying network structure of their associations. Clusters generated from our approach are qualitatively better than clusters generated from the conventional spectral clustering mechanism used for graph partitioning.

Scalable non-linear Support Vector Machine using hierarchical clustering

This paper discusses a method for scaling SVM with Gaussian kernel function to handle large data sets by using a selective sampling strategy for the training set. It employs a scalable hierarchical clustering algorithm to construct cluster indexing structures of the training data in the kernel induced feature space. These are then used for selective sampling of the training data for SVM to impart scalability to the training process. Empirical studies made on real world data sets show that the proposed strategy performs well on large data sets.

Agglomerative clustering using the concept of mutual nearest neighbourhood

A method for determining the mutual nearest neighbours (MNN) and mutual neighbourhood value (mnv) of a sample point, using the conventional nearest neighbours, is suggested. A nonparametric, hierarchical, agglomerative clustering algorithm is developed using the above concepts. The algorithm is simple, deterministic, noniterative, requires low storage and is able to discern spherical and nonspherical clusters. The method is applicable to a wide class of data of arbitrary shape, large size and high dimensionality. The algorithm can discern mutually homogenous clusters. Strong or weak patterns can be discerned by properly choosing the neighbourhood width.

Disaggregative Clustering Using the Concept of Mutual Nearest Neighborhood

A nonparametric, hierarchical, disaggregative clustering algorithm is developed using a novel similarity measure, called the mutual neighborhood value (MNV), which takes into account the conventional nearest neighbor ranks of two samples with respect to each other. The algorithm is simple, noniterative, requires low storage, and needs no specification of the expected number of clusters. The algorithm appears very versatile as it is capable of discerning spherical and nonspherical clusters, linearly nonseparable clusters, clusters with unequal populations, and clusters with lowdensity bridges. Changing of the neighborhood size enables discernment of strong or weak patterns.

Model-theoretic approach to clustering

The paper deals with a model-theoretic approach to clustering. The approach can be used to generate cluster description based on knowledge alone. Such a process of generating descriptions would be extremely useful in clustering partially specified objects. A natural byproduct of the proposed approach is that missing values of attributes of an object can be estimated with ease in a meaningful fashion. An important feature of the approach is that noisy objects can be detected effectively, leading to the formation of natural groups. The proposed algorithm is applied to a library database consisting of a collection of books.

A heuristic clustering algorithm using union of overlapping pattern-cells

Relative geometric arrangements of the sample points, with reference to the structure of the imbedding space, produce clusters. Hence, if each sample point is imagined to acquire a volume of a small M-cube (called pattern-cell), depending on the ranges of its (M) features and number (N) of samples; then overlapping pattern-cells would indicate naturally closer sample-points. A chain or blob of such overlapping cells would mean a cluster and separate clusters would not share a common pattern-cell between them. The conditions and an analytic method to find such an overlap are developed. A simple, intuitive, nonparametric clustering procedure, based on such overlapping pattern-cells is presented. It may be classified as an agglomerative, hierarchical, linkage-type clustering procedure. The algorithm is fast, requires low storage and can identify irregular clusters. Two extensions of the algorithm, to separate overlapping clusters and to estimate the nature of pattern distributions in the sample space, are also indicated.

A critical re-examination and a revised phase diagram of La1-xSrxCoO3

We report the results of a comprehensive study on dc magnetization, ac susceptibility, and the magnetotransport properties of the La1-xSrxCoO3(0 <= x <= 0.5) system. At higher Sr doping (x >= 0.18), the system exhibits Brillouin-like field cooled magnetization (M-FC). However, for x < 0.18, the system exhibits a kink in the M-FC, a peak at the intermediate field in the thermoremnant magnetization and a non-saturating tendency in the M-H plot that all point towards the characteristic of spin glass behavior. More interestingly, dc magnetization studies for x < 0.18 do not suggest the existence of ferromagnetic correlation that can give rise to an irreversible line in the spin glass regime. The ac susceptibility study for x > 0.2 exhibits apparently no frequency dependent peak shift around the ferromagnetic transition region. However, a feeble signature of glassiness is verified by studying the frequency dependent shoulder position in chi `' (T) and the memory effect below the Curie temperature. But, for x < 0.18, the ac susceptibility study exhibits a considerable frequency dependent peak shift, time dependent memory effect, and the characteristic spin relaxation time scale tau(0) similar to 10(-13) s. The reciprocal susceptibility versus temperature plot adheres to Curie-Weiss behavior and does not provide any signature of preformed ferromagnetic clusters well above the Curie temperature. The magnetotransport study reveals a cross over from metallic to semiconducting-like behavior for x <= 0.18. On the semiconducting side, the system exhibits a large value of magnetoresistance (upto 75%) towards low temperature and it is strongly connected to the spin dependent part of the random potential distribution in the spin glass phase. Based on the above observations, we have reconstructed a new magnetic phase diagram and characterized each phase with associated properties.

A comparison between conceptual clustering and conventional clustering

Clustering is a process of partitioning a given set of patterns into meaningful groups. The clustering process can be viewed as consisting of the following three phases: (i) feature selection phase, (ii) classification phase, and (iii) description generation phase. Conventional clustering algorithms implicitly use knowledge about the clustering environment to a large extent in the feature selection phase. This reduces the need for the environmental knowledge in the remaining two phases, permitting the usage of simple numerical measure of similarity in the classification phase. Conceptual clustering algorithms proposed by Michalski and Stepp [IEEE Trans. PAMI, PAMI-5, 396–410 (1983)] and Stepp and Michalski [Artif. Intell., pp. 43–69 (1986)] make use of the knowledge about the clustering environment in the form of a set of predefined concepts to compute the conceptual cohesiveness during the classification phase. Michalski and Stepp [IEEE Trans. PAMI, PAMI-5, 396–410 (1983)] have argued that the results obtained with the conceptual clustering algorithms are superior to conventional methods of numerical classification. However, this claim was not supported by the experimental results obtained by Dale [IEEE Trans. PAMI, PAMI-7, 241–244 (1985)]. In this paper a theoretical framework, based on an intuitively appealing set of axioms, is developed to characterize the equivalence between the conceptual clustering and conventional clustering. In other words, it is shown that any classification obtained using conceptual clustering can also be obtained using conventional clustering and vice versa.

Lenient Execution and Concurrent Execution of Re-entrant Routines: Efficient Implementation in Data Flow Systems

Implementation details of efficient schemes for lenient execution and concurrent execution of re-entrant routines in a data flow model have been discussed in this paper. The proposed schemes require no extra hardware support and utilise the existing hardware resources such as the Matching Unit and Memory Network Interface, effectively to achieve the above mentioned goals.

Perturbative growth of cosmological clustering. I: Formalism

Here we rederive the hierarchy of equations for the evolution of distribution functions of various orders using a convenient parameterization. We use this to obtain equations for two- and three-point correlation functions in powers of a small parameter, viz., the initial density contrast. The correspondence of the lowest order solutions of these equations to the results from the linear theory of density perturbations is shown for an OMEGA = 1 universe. These equations are then used to calculate, to the lowest order, the induced three-point correlation function that arises from Gaussian initial conditions in an OMEGA = 1 universe. We obtain an expression which explicitly exhibits the spatial structure of the induced three-point correlation function. It is seen that the spatial structure of this quantity is independent of the value of OMEGA. We also calculate the triplet momentum. We find that the induced three-point correlation function does not have the ''hierarchical'' form often assumed. We discuss possibilities of using the induced three-point correlation to interpret observational data. The formalism developed here can also be used to test a validity of different schemes to close the

Knowledge-based clustering

In the knowledge-based clustering approaches reported in the literature, explicit know ledge, typically in the form of a set of concepts, is used in computing similarity or conceptual cohesiveness between objects and in grouping them. We propose a knowledge-based clustering approach in which the domain knowledge is also used in the pattern representation phase of clustering. We argue that such a knowledge-based pattern representation scheme reduces the complexity of similarity computation and grouping phases. We present a knowledge-based clustering algorithm for grouping hooks in a library.

Perturbative growth of cosmological clustering .2. The two-point correlation

We use the BBGKY hierarchy equations to calculate, perturbatively, the lowest order nonlinear correction to the two-point correlation and the pair velocity for Gaussian initial conditions in a critical density matter-dominated cosmological model. We compare our results with the results obtained using the hydrodynamic equations that neglect pressure and find that the two match, indicating that there are no effects of multistreaming at this order of perturbation. We analytically study the effect of small scales on the large scales by calculating the nonlinear correction for a Dirac delta function initial two-point correlation. We find that the induced two-point correlation has a x(-6) behavior at large separations. We have considered a class of initial conditions where the initial power spectrum at small k has the form k(n) with 0 < n less than or equal to 3 and have numerically calculated the nonlinear correction to the two-point correlation, its average over a sphere and the pair velocity over a large dynamical range. We find that at small separations the effect of the nonlinear term is to enhance the clustering, whereas at intermediate scales it can act to either increase or decrease the clustering. At large scales we find a simple formula that gives a very good fit for the nonlinear correction in terms of the initial function. This formula explicitly exhibits the influence of small scales on large scales and because of this coupling the perturbative treatment breaks down at large scales much before one would expect it to if the nonlinearity were local in real space. We physically interpret this formula in terms of a simple diffusion process. We have also investigated the case n = 0, and we find that it differs from the other cases in certain respects. We investigate a recently proposed scaling property of gravitational clustering, and we find that the lowest order nonlinear terms cause deviations from the scaling relations that are strictly valid in the linear regime. The approximate validity of these relations in the nonlinear regime in l(T)-body simulations cannot be understood at this order of evolution.

Re-examination of compaction characteristics of fine-grained soils

Standard Proctor compaction test data were generated for 3 soils with liquid limit water contents ranging from 48% to 84%. It has been established that by defining a soil by its liquid limit and coarse fraction, the path of compaction for a specific compactive effort can be predicted via a simple density-water content-liquid limit relationship. (Abstract quotes from original text)

Superconducting YBa2Cu3O7-delta thick films on Ba2RETaO6 (RE=Pr, Nd, Eu, and Dy) substrates

Thick films of YBa2Cu3O7-delta fabricated on polycrystalline Ba2RETaO6 (where RE= Pr, Nd, Eu, and Dy) substrates by dip-coating and partial melting techniques are textured and c-axis oriented, showing predominantly (00l) orientation. All the thick films show a superconducting zero resistance transition of 90 K. SEM studies clearly indicate platelike and needlelike grain growth over a wide area of the thick films. The values of the critical current density for these thick films are similar to 10(4) A/cm(2) at 77 K as determined by the nonresonant R.F. absorption method. Various processing conditions that affect the critical current density of thick films are also discussed.

Re-examination of crack opening model of interface fracture

The complex singularity associated with a crack at the interface between two dissimilar, isotropic and homogeneous materials leads to mathematical artefacts, such as stress oscillations and crack face interpenetrations in the vicinity of the crack tip. To avoid these unrealistic features, Sinclair (Sinclair GB. On the stress singularity at an interface crack. International Journal of Fracture 1980;16(2):111-9) assumed a finite crack opening angle (COA) such that the singularity lambda became real equal to 1/2. This paper extends the COA model by considering real singularities not necessarily equal to 1/2. When COA is 0 degrees: the interface crack singularity is complex with a real part equal to 1/2. On increasing COA, the imaginary part of the singularity decreases and becomes zero at a threshold value of COA; at this point, the singularity is a real, repeated value. A further increase in COA results in a pair of real singularities. Different crack opening configurations and material combinations are studied, and results presented for threshold COAs and associated values of singularity. Stress analyses for these three regimes: (a) complex, (b) real pair and (c) real repeated singularities, are reported. It is seen that additional complexities are present in the last case. Typical results for stress fields are also included for comparing with standard fields. (C) 1999 Elsevier Science Ltd. All rights reserved.