A review of topic modeling methods

Topic modeling is a popular analytical tool for evaluating data. Numerous methods of topic modeling have been developed which consider many kinds of relationships and restrictions within datasets; however, these methods are not frequently employed. Instead many researchers gravitate to Latent Dirichlet Analysis, which although flexible and adaptive, is not always suited for modeling more complex data relationships. We present different topic modeling approaches capable of dealing with correlation between topics, the changes of topics over time, as well as the ability to handle short texts such as encountered in social media or sparse text data. We also briefly review the algorithms which are used to optimize and infer parameters in topic modeling, which is essential to producing meaningful results regardless of method. We believe this review will encourage more diversity when performing topic modeling and help determine what topic modeling method best suits the user needs.     

Web服务提供方安全模型的分析与设计

讨论了Web服务提供方的安全问题,通过对Web服务调用方式的分析,为Web服务的提供方设计了安全模型.所设计的模型不依赖于服务提供方选用的安全策略,方便策略的更改和安全检查机制的更新.该模型同样适用于Web服务的需求方.     

Automatic playlist generation based on tracking user’s listening habits

Algorithms for automatic playlist generation solve the problem of tedious and time consuming manual selection of musical playlists. These algorithms generate playlists according to the user’s music preferences of the moment. The user describes his preferences either by manually inputting a couple of example songs, or by defining constraints for the choice of music. The approaches to automatic playlist generation up to now were based on examining the metadata attached to the music pieces. Some of them took also the listening history into account. But anyway, a heavy accent has been put on the metadata, while the listening history, if it was used at all, had a minor role. Missings and errors in metadata frequently appear, especially when the music is acquired from the Internet. When the metadata is missing or wrong, the approaches proposed so far cannot work. Besides, entering constraints for the playlist generation can be a difficult activity. In our approach we ignored the metadata and focused on examining the listening habits. We developed two simple algorithms that track the listening habits and form a listener model—a profile of listening habits. The listener model is then used for automatic playlist generation. We developed a simple media player which tracks the listening habits and generates playlists according to the listener model. We tried the solution with a group of users. The experiment was not a successful one, but it threw some new light on the relationship between the listening habits and playlist generation.     

Inference and learning in hybrid probabilistic network

This paper proposed a novel hybrid probabilistic network, which is a good tradeoff between the model complexity and learnability in practice. It relaxes the conditional independence assumptions of Naive Bayes while still permitting efficient inference and learning. Experimental studies on a set of natural domains prove its clear advantages with respect to the generalization ability.     

Methodological Challenges in Union Commitment Studies.

Methodological problems in studies of union commitment were identified and illustrated with data from 4,641 members and 479 stewards in 297 local teachers' unions. Using a 20-item union commitment scale, results confirmed the existence of 3 substantive factors and 1 method factor at the individual level of analysis: loyalty to the union, responsibility to the union, willingness to work for the union, and a factor of negatively worded items. Tests of measurement invariance showed that the scale captured commitment for rank-and-file members but not for union stewards. The authors also found partial measurement invariance between long-time and newer members and full measurement invariance between men and women. Finally, the authors found that violation of the statistical assumption of independence reduced model fit when individual commitment scores were analyzed without attention to the hierarchical nature of the data. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

NUMERICAL ANALYSIS OF THE EFFECTS OF LOCAL HYDRODYNAMICS ON MASS TRANSFER IN HETEROGENEOUS POROUS MEDIA

Many engineering problems require the estimation of mass transfer coefficients in porous materials. In heterogeneous materials or in cases where mass transfer sites are not spatially uniform, empirical equations for mass transfer coefficients vary widely, and the origin of these differences is not well understood. In this article, we use a stochastic algorithm to model mass transfer from single particles in a two-dimensional heterogeneous packed bed. The computed mass transfer coefficients are used to generate a distribution of local Peclet numbers in the bed. Detailed hydrodynamics are then used to interpret variations in the local Peclet number. The results show clear relationships between pore structure, streamline patterns, and mass transfer rates.     

The impact of presentation graphics on students’ experience in the classroom

To investigate the benefits and perceived effectiveness of instructional technology, students enrolled in several courses were compared on student evaluations of instruction, grades and an attitudinal questionnaire. The instructors of the courses taught the same course across two successive semesters, using traditional “chalk-and-talk” methods the first semester, and PowerPoint the next; all other techniques (e.g., exams, lecture material) were held constant. Results suggest that organization and clarity, entertainment and interest, professor likeability, and good professor behaviors were enhanced with PowerPoint although final grades were not.     

Radiation-laminar free convection in a square duct with specular reflection by absorbing-emitting medium

The purpose of this work is to study the effects of specularly reflecting wall under the combined radiative and laminar free convective heat transfer in an infinite square duct. An absorbing and emitting gray medium is enclosed by the opaque and diffusely emitting walls. The walls may reflect diffusely or specularly. Boussinesq approximation is used for the buoyancy term. The radiative heat transfer is evaluated using the direct discrete ordinates method. The parameters under considerations are Rayleigh number, conduction to radiation parameter, optical thickness, wall emissivity and reflection mode. The differences caused by the reflection mode on the stream line, and temperature distribution and wall heat fluxes are studied. Some differences are observed for the categories mentioned above if the order of the conduction to radiation parameter is less than order of 10^-3 for the range of Rayleigh number studied. The differences at the side wall heat flux distributions are observed as long as the medium is optically thin. As the top wall emissivity decreases, the differences between these two modes are increased. As the optical thickness decreases at the fixed wall emissivity, the differences also increase. The difference of the streamlines or the temperature contours is not as distinct as the side wall heat flux distributions. The specular reflection may alter the fluid motion.     

On the monotonicity of (LDL) logic programs with set

LDL is one of the recently proposed logical query languages, which incorporate set, for data and knowledge base systems. Since LDL programs can simulate negation, they are not monotonic in general. On the other hand, there are monotonic LDL programs. This paper addresses the natural question of “When are the generally nonmonotonic LDL programs monotonic?” and investigates related topics such as useful applications for monotonicity. We discuss four kinds of monotonicity, and examine two of them in depth. The first of the two, called “ω-monotonicity”, is shown to be undecidable even when limited to single-stratum programs. The second, called “uniform monotonicity”, is shown to implyω-monotonicity. We characterize the uniform monotonicity of a program (i) by a relationship between its Bancilhon-Khoshafian semantics and its LDL semantics, and (ii) with a useful property called subset completion independence. Characterization (ii) implies that uniformly monotonie programs can be evaluated more efficiently by discarding dominated facts. Finally, we provide some necessary and/or sufficient, syntactic conditions for uniform monotonicity. The conditions pinpoint (a) enumerated set terms, (b) negations of membership and inclusion, and (c) sharing of set terms as the main source for nonuniform monotonicity.