Extending boosting for large scale spoken language understanding

We propose three methods for extending the Boosting family of classifiers motivated by the real-life problems we have encountered. First, we propose a semisupervised learning method for exploiting the unlabeled data in Boosting. We then present a novel classification model adaptation method. The goal of adaptation is optimizing an existing model for a new target application, which is similar to the previous one but may have different classes or class distributions. Finally, we present an efficient and effective cost-sensitive classification method that extends Boosting to allow for weighted classes. We evaluated these methods for call classification in the AT&;T VoiceTone^® spoken language understanding system. Our results indicate that it is possible to obtain the same classification performance by using 30% less labeled data when the unlabeled data is utilized through semisupervised learning. Using model adaptation we can achieve the same classification accuracy using less than half of the labeled data from the new application. Finally, we present significant improvements in the “important” (i.e., higher weighted) classes without a significant loss in overall performance using the proposed cost-sensitive classification method.     

Prediction and quality control of the melt index during production of high-density polyethylene

In polyolefin processes the melt index (MI) is the most important control variable indicating product quality. Because of the difficulty in the on-line measurement of MI, a lot of MI estimation and correlation methods have been proposed. In this work a new dynamic MI estimation scheme is developed based on system identification techniques. The empirical MI estimation equation proposed in the present study is derived from the 1 ^st -order dynamic models. Effectiveness of the present estimation scheme was illustrated by numerical simulations based on plant operation data including grade change operations in high density polyethylene (HDPE) processes. From the comparisons with other estimation methods it was found that the proposed estimation scheme showed better performance in MI predictions. The virtual sensor model developed based on the estimation scheme was combined with the virtual on-line analyzer (VOA) to give a quality control system to be implemented in the actual HDPE plant. From the application of the present control system, significant reduction of transition time and the amount of off-spec during grade changes was achieved     

棉织物手感值的模糊综合评判研究

利用模糊综合评判数学模型对棉织物的手感值进行了计算与分析，这种方法简单、可靠，与主观评估结果有很好的一致性，为纺织生产的质量判断提供了便利。     

Sandia National Laboratories Validation Workshop: Structural dynamics application

This article specifies a virtual structural dynamic subsystem and some systems, as well as mathematical models that approximately simulate them. The purpose is to define a setting for model validations conceived by participants in the Sandia National Laboratories Validation Workshop. Some broad guidelines for the model validations are set, as well as a regulatory requirement to be satisfied by the target system. Participants are directed to present techniques for subsystem model validation, to analytically predict whether or not the regulatory requirement will be met, and to estimate the likelihood of accuracy of the prediction.     

Modelling the FEBEX THM experiment using a state surface approach

Buffer materials being considered as engineered barriers in nuclear fuel waste (NFW) disposal systems possess a pronounced nonlinear behaviour in the unsaturated state. In order to simulate such non-linear responses,the authors adopted an incrementally nonlinear poro-elastic approach where the coefficients of the governing equations are assumed to be functions of suction and the void ratio. These functions are in turn developed from a state-surface equation obtained from suction-controlled oedometric tests. In this paper we show the derivation of the governing equations of the poro-elastic model. A finite element computer code, FRACON, was developed by the authors to numerically solve the above equations. We first use the code to simulate laboratory tests to characterize the swelling properties of a typical bentonite. That same bentonite was used in the FEBEX in-situ heater experiment, conducted at the Grimsel site, Switzerland. The FRACON code was also used to perform blind predictions of the FEBEX heater experiment. It is shown that the model correctly predicts drying of the bentonite near the heaters and re-saturation near the rock interface. The evolution of temperature and the heater thermal output were also reasonably well predicted by the model. The trends in the total stresses developed in the bentonite were correctly predicted; the absolute values however were underestimated probably due to the omission of pore pressure build-up in the rock mass.     

化学事故危害评估

根据化学事故危害的主要影响因素对化学事故进行危害评估,考查了危害纵深、人员杀伤率、危害地域估算、爆炸燃烧事故危害等内容。     

Results of a User Study on 2D Hurricane Visualization

We present the results from a user study looking at the ability of observers to mentally integrate wind direction and magnitude over a vector field. The data set chosen for the study is an MM5 (PSU/NCAR Mesoscale Model) simulation of Hurricane Lili over the Gulf of Mexico as it approaches the southeastern United States. Nine observers participated in the study. This study investigates the effect of layering on the observer's ability to determine the magnitude and direction of a vector field. We found a tendency for observers to underestimate the magnitude of the vectors and a counter‐clockwise bias when determining the average direction of a vector field. We completed an additional study with two observers to try to uncover the source of the counter‐clockwise bias. These results have direct implications to atmospheric scientists, but may also be able to be applied to other fields that use 2D vector fields.     

Lattice structure on some fuzzy algebraic systems

In this paper, we study the lattice structure of some fuzzy algebraic systems such as (G-)fuzzy groups, some fuzzy ordered algebras and fuzzy hyperstructures. We prove that under suitable conditions, these structures form a distributive or modular lattice. This research partially is supported by the “ Fuzzy Systems and its Applications Center of Excelence, Shahid Bahonar University of Kerman, Iran”.     

Education-driven research in CAD

We argue for a new research category, named education-driven research (EDR), which fills the gap between traditional field-specific research that is not concerned with educational objectives and research in education that focuses on fundamental teaching and learning principles and possibly on their customization to broad areas (such as mathematics or physics), but not to specific disciplines (such as CAD). The objective of EDR is to simplify the formulation of the underlying theoretical foundations and of specific tools and solutions in a specialized domain, so as to make them easy to understand and internalize. As such, EDR is a difficult and genuine research activity, which requires a deep understanding of the specific field and can rarely be carried out by generalists with primary expertise in broad education principles. We illustrate the concept of EDR with three examples in CAD: (1) the Split and Tweak subdivisions of a polygon and its use for generating curves, surfaces, and animations; (2) the construction of a topological partition of a plane induced by an arbitrary arrangement of edges; and (3) a romantic definition of the minimal and Hausdorff distances. These examples demonstrate the value of using analogies, of introducing evocative terminology, and of synthesizing the simplest fundamental building blocks. The intuitive understanding provided by EDR enables the students (and even the instructor) to better appreciate the limitations of a particular solution and to explore alternatives. In particular, in these examples, EDR has allowed the author to: (1) reduce the cost of evaluating a cubic B-spline curve; (2) develop a new subdivision curve that is better approximated by its control polygon than either a cubic B-spline or an interpolating 4-point subdivision curve; (3) discover how a circuit inclusion tree may be used for identifying the faces in an arrangement; and (4) rectify a common misconception about the computation of the Hausdorff error between triangle meshes. We invite the scientific community to encourage the development of EDR by publishing its results as genuine research contributions in peer-reviewed professional journals.     

Estimation for the number of components in a mixture model using stepwise split-and-merge EM algorithm

The main difficulty with EM algorithm for mixture model concerns the number of components, say g. This is the question of model selection, and the EM algorithm itself could not estimate g. On the contrary, the algorithm requires g to be specified before the remaining parameters can be estimated. To solve this problem, a new algorithm, which is called stepwise split-and-merge EM (SSMEM) algorithm, is proposed. The SSMEM algorithm alternately splits and merges components, estimating g and other parameters of components simultaneously. Also, two novel criteria are introduced to efficiently select the components for split or merge. Experimental results on simulated and real data demonstrate the effectivity of the proposed algorithm.