Efficient algorithms for learning kernels from multiple similarity matrices with general convex loss functions

In this paper we consider the problem of learning an n × n kernel matrix from m(1) similarity matrices under general convex loss. Past research have extensively studied the m = 1 case and have derived several algorithms which require sophisticated techniques like ACCP, SOCP, etc. The existing algorithms do not apply if one uses arbitrary losses and often can not handle m > 1 case. We present several provably convergent iterative algorithms, where each iteration requires either an SVM or a Multiple Kernel Learning (MKL) solver for m > 1 case. One of the major contributions of the paper is to extend the well knownMirror Descent(MD) framework to handle Cartesian product of psd matrices. This novel extension leads to an algorithm, called EMKL, which solves the problem in O(m2 log n 2) iterations; in each iteration one solves an MKL involving m kernels and m eigen-decomposition of n × n matrices. By suitably defining a restriction on the objective function, a faster version of EMKL is proposed, called REKL,which avoids the eigen-decomposition. An alternative to both EMKL and REKL is also suggested which requires only an SVMsolver. Experimental results on real world protein data set involving several similarity matrices illustrate the efficacy of the proposed algorithms.

Sequence-Structure Similarity: Do Sequentially Identical Peptide Fragments have Similar Three-Dimensional Structures?

The rapidly growing structure databases enhance the probability of finding identical sequences sharing structural similarity. Structure prediction methods are being used extensively to abridge the gap between known protein sequences and the solved structures which is essential to understand its specific biochemical and cellular functions. In this work, we plan to study the ambiguity between sequence-structure relationships and examine if sequentially identical peptide fragments adopt similar three-dimensional structures. Fragments of varying lengths (five to ten residues) were used to observe the behavior of sequence and its three-dimensional structures. The STAMP program was used to superpose the three-dimensional structures and the two parameters (Sequence Structure Similarity Score (Sc) and Root Mean Square Deviation value) were employed to classify them into three categories: similar, intermediate and dissimilar structures. Furthermore, the same approach was carried out on all the three-dimensional protein structures solved in the two organisms, Mycobacterium tuberculosis and Plasmodium falciparum to validate our results.

Positive Relationships between Association Strength and Phenotypic Similarity Characterize the Assembly of Mixed-Species Bird Flocks Worldwide

Competition theory predicts that local communities should consist of species that are more dissimilar than expected by chance. We find a strikingly different pattern in a multicontinent data set (55 presence-absence matrices from 24 locations) on the composition of mixed-species bird flocks, which are important sub-units of local bird communities the world over. By using null models and randomization tests followed by meta-analysis, we find the association strengths of species in flocks to be strongly related to similarity in body size and foraging behavior and higher for congeneric compared with noncongeneric species pairs. Given the local spatial scales of our individual analyses, differences in the habitat preferences of species are unlikely to have caused these association patterns; the patterns observed are most likely the outcome of species interactions. Extending group-living and social-information-use theory to a heterospecific context, we discuss potential behavioral mechanisms that lead to positive interactions among similar species in flocks, as well as ways in which competition costs are reduced. Our findings highlight the need to consider positive interactions along with competition when seeking to explain community assembly.

Efficient external memory structures for range-aggregate queries

We present external memory data structures for efficiently answering range-aggregate queries. The range-aggregate problem is defined as follows: Given a set of weighted points in R-d, compute the aggregate of the weights of the points that lie inside a d-dimensional orthogonal query rectangle. The aggregates we consider in this paper include COUNT, sum, and MAX. First, we develop a structure for answering two-dimensional range-COUNT queries that uses O(N/B) disk blocks and answers a query in O(log(B) N) I/Os, where N is the number of input points and B is the disk block size. The structure can be extended to obtain a near-linear-size structure for answering range-sum queries using O(log(B) N) I/Os, and a linear-size structure for answering range-MAX queries in O(log(B)(2) N) I/Os. Our structures can be made dynamic and extended to higher dimensions. (C) 2012 Elsevier B.V. All rights reserved.

Similarity relations in visual search predict rapid visual categorization

How do we perform rapid visual categorization?It is widely thought that categorization involves evaluating the similarity of an object to other category items, but the underlying features and similarity relations remain unknown. Here, we hypothesized that categorization performance is based on perceived similarity relations between items within and outside the category. To this end, we measured the categorization performance of human subjects on three diverse visual categories (animals, vehicles, and tools) and across three hierarchical levels (superordinate, basic, and subordinate levels among animals). For the same subjects, we measured their perceived pair-wise similarities between objects using a visual search task. Regardless of category and hierarchical level, we found that the time taken to categorize an object could be predicted using its similarity to members within and outside its category. We were able to account for several classic categorization phenomena, such as (a) the longer times required to reject category membership; (b) the longer times to categorize atypical objects; and (c) differences in performance across tasks and across hierarchical levels. These categorization times were also accounted for by a model that extracts coarse structure from an image. The striking agreement observed between categorization and visual search suggests that these two disparate tasks depend on a shared coarse object representation.

Combination of similarity measures for time series classification using genetic algorithms

Time series classification deals with the problem of classification of data that is multivariate in nature. This means that one or more of the attributes is in the form of a sequence. The notion of similarity or distance, used in time series data, is significant and affects the accuracy, time, and space complexity of the classification algorithm. There exist numerous similarity measures for time series data, but each of them has its own disadvantages. Instead of relying upon a single similarity measure, our aim is to find the near optimal solution to the classification problem by combining different similarity measures. In this work, we use genetic algorithms to combine the similarity measures so as to get the best performance. The weightage given to different similarity measures evolves over a number of generations so as to get the best combination. We test our approach on a number of benchmark time series datasets and present promising results.

Similarity of Matrices Over Local Rings of Length Two

Let R be a (commutative) local principal ideal ring of length two, for example, the ring R = Z/p(2)Z with p prime. In this paper, we develop a theory of normal forms for similarity classes in the matrix rings M-n (R) by interpreting them in terms of extensions of R t]-modules. Using this theory, we describe the similarity classes in M-n (R) for n <= 4, along with their centralizers. Among these, we characterize those classes which are similar to their transposes. Non-self-transpose classes are shown to exist for all n > 3. When R has finite residue field of order q, we enumerate the similarity classes and the cardinalities of their centralizers as polynomials in q. Surprisingly, the polynomials representing the number of similarity classes in M-n (R) turn out to have non-negative integer coefficients.

Application of Similarity Theory to the Characterization of Non-Transferred Laminar Plasma Jet Generation

Laminar-flow non-transferred DC plasma jets were generated by a torch with an inter-electrode insert by which the arc column was limited to a length of about 20 mm. Current–voltage characteristics, thermal efficiency and jet length, a parameter which changes greatly with the generating parameters in contrast with the almost unchangeable jet length of the turbulent plasma, were investigated systematically, by using the similarity theory combined with the corresponding experimental examination. Formulae in non-dimensional forms were derived for predicting the characteristics of the laminar plasma jet generation, within the parameter ranges where no transfer to turbulent flow occurs. Mean arc temperature in the torch channel and mean jet-flow temperature at the torch exit were obtained, and the results indicate that the thermal conductivity feature of the working gas seems to be an important factor affecting thermal efficiency of laminar plasma generation.

Novel Investigation on Principle of Similarity for Microgravity Two-Phase Flow

根据两流体同心环状流线性稳定性分析的结果 ,对微重力气 /液两相流地面模拟实验所应遵循的相似准则进行了探讨 ,得到了一个新的重力无关性准则 ,即Bond数和环形区流体相的毛细数之比的绝对值不大于 1 .此外 ,微重力气 /液两相流模拟实验还必须满足两个条件 ,即流量比和气相表观Weber数应与所模拟的流动中对应数值相等 . In the present paper, the principle of similarity for two phase flows at microgravity is studied based on the results of the linear stability analysis of the two fluid concentric annular flow configuration. A new criterion of gravity independence, namely the absolute value of the ratio between the Bond number and the capillary number of the phase flowing in the annulus is no more than one, is achieved. It is also pointed out that the flowrate ratio and the gas superficial Weber number must have the same ...

Similarity consideration of structural bifurcation buckling

Many structural bifurcation buckling problems exhibit a scaling or power law property. Dimensional analysis is used to analyze the general scaling property. The concept of a new dimensionless number, the response number-Rn, suggested by the present author for the dynamic plastic response and failure of beams, plates and so on, subjected to large dynamic loading, is generalized in this paper to study the elastic, plastic, dynamic elastic as well as dynamic plastic buckling problems of columns, plates as well as shells. Structural bifurcation buckling can be considered when Rn(n) reaches a critical value.

A key role for similarity in vicarious reward.

Humans appear to have an inherent prosocial tendency toward one another in that we often take pleasure in seeing others succeed. This fact is almost certainly exploited by game shows, yet why watching others win elicits a pleasurable vicarious rewarding feeling in the absence of personal economic gain is unclear. One explanation is that game shows use contestants who have similarities to the viewing population, thereby kindling kin-motivated responses (for example, prosocial behavior). Using a game show-inspired paradigm, we show that the interactions between the ventral striatum and anterior cingulate cortex subserve the modulation of vicarious reward by similarity, respectively. Our results support studies showing that similarity acts as a proximate neurobiological mechanism where prosocial behavior extends to unrelated strangers.

Designing Of Partial Similarity Models And Evaluation Method In Polymer Flooding Experiment

Based on the scaling criteria of polymer flooding reservoir obtained in our previous work in which the gravity and capillary forces, compressibility, non-Newtonian behavior, absorption, dispersion, and diffusion are considered, eight partial similarity models are designed. A new numerical approach of sensitivity analysis is suggested to quantify the dominance degree of relaxed dimensionless parameters for partial similarity model. The sensitivity factor quantifying the dominance degree of relaxed dimensionless parameter is defined. By solving the dimensionless governing equations including all dimensionless parameters, the sensitivity factor of each relaxed dimensionless parameter is calculated for each partial similarity model; thus, the dominance degree of the relaxed one is quantitatively determined. Based on the sensitivity analysis, the effect coefficient of partial similarity model is defined as the summation of product of sensitivity factor of relaxed dimensionless parameter and its relative relaxation quantity. The effect coefficient is used as a criterion to evaluate each partial similarity model. Then the partial similarity model with the smallest effect coefficient can be singled out to approximate to the prototype. Results show that the precision of partial similarity model is not only determined by the number of satisfied dimensionless parameters but also the relative relaxation quantity of the relaxed ones.