Advanced product planning: a comprehensive process for systemic definition of new product requirements

This paper reports results of research into the definition of requirements for new consumer products––specifically, electro-mechanical products. The research dealt with the derivation of design requirements that are demonstrably aligned with stakeholder needs. The paper describes a comprehensive process that can enable product development teams to deal with statements of product requirements, as originally collected through market research activities, in a systematic and traceable manner from the early, fuzzy front end, stages of the design process. The process described has been based on principles of systems engineering. A case study from its application and evaluation drawn from the power sector is described in this paper. The case study demonstrates how the process can significantly improve product quality planning practices through revision of captured product requirements, analysis of stakeholder requirements and derivation of design requirements. The paper discusses benefits and issues from the use of the process by product development teams, and identifies areas for further research. Finally, the conclusions drawn from the reported research are presented.

Building Sparse Multiple-Kernel SVM Classifiers

The support vector machines (SVMs) have been very successful in many machine learning problems. However, they can be slow during testing because of the possibly large number of support vectors obtained. Recently, Wu (2005) proposed a sparse formulation that restricts the SVM to use a small number of expansion vectors. In this paper, we further extend this idea by integrating with techniques from multiple-kernel learning (MKL). The kernel function in this sparse SVM formulation no longer needs to be fixed but can be automatically learned as a linear combination of kernels. Two formulations of such sparse multiple-kernel classifiers are proposed. The first one is based on a convex combination of the given base kernels, while the second one uses a convex combination of the so-called ldquoequivalentrdquo kernels. Empirically, the second formulation is particularly competitive. Experiments on a large number of toy and real-world data sets show that the resultant classifier is compact and accurate, and can also be easily trained by simply alternating linear program and standard SVM solver.     

MatSeek: An Ontology-Based Federated Search Interface for Materials Scientists

The MatSeek system is an ontology-based federated search interface to key materials science databases and analytical tools. By combining Semantic Web and Web 2.0 technologies, MatSeek provides materials scientists with a single Web interface that enables them to search across disparate databases containing crystal-structure data, ionic-conductivity data, and phase stability data; render 3D crystal-structure images; calculate bond lengths and angles; retrieve relevant scholarly references; and identify potential new materials with the structure and properties required to satisfy specific applications. The MatOnto ontology underlying MatSeek enables integration of data across disparate databases, and Web 2.0 technologies enable iterative searching across the databases. The results retrieved from searching the previous database are used as input to the query on the next database. By providing materials scientists with a single, integrated Web interface to the critical materials science databases and analytical tools, MatSeek represents a significant advance toward a full-fledged materials-informatics workbench.     

Maximum Penalized Likelihood Kernel Regression for Fast Adaptation

This paper proposes a nonlinear generalization of the popular maximum-likelihood linear regression (MLLR) adaptation algorithm using kernel methods. The proposed method, called maximum penalized likelihood kernel regression adaptation (MPLKR), applies kernel regression with appropriate regularization to determine the affine model transform in a kernel-induced high-dimensional feature space. Although this is not the first attempt of applying kernel methods to conventional linear adaptation algorithms, unlike most of other kernelized adaptation methods such as kernel eigenvoice or kernel eigen-MLLR, MPLKR has the advantage that it is a convex optimization and its solution is always guaranteed to be globally optimal. In fact, the adapted Gaussian means can be obtained analytically by simply solving a system of linear equations. From the Bayesian perspective, MPLKR can also be considered as the kernel version of maximum a posteriori linear regression (MAPLR) adaptation. Supervised and unsupervised speaker adaptation using MPLKR were evaluated on the Resource Management and Wall Street Journal 5K tasks, respectively, achieving a word error rate reduction of 23.6% and 15.5% respectively over the speaker-independently model.     

Text detection in images using sparse representation with discriminative dictionaries

Text detection is important in the retrieval of texts from digital pictures, video databases and webpages. However, it can be very challenging since the text is often embedded in a complex background. In this paper, we propose a classification-based algorithm for text detection using a sparse representation with discriminative dictionaries. First, the edges are detected by the wavelet transform and scanned into patches by a sliding window. Then, candidate text areas are obtained by applying a simple classification procedure using two learned discriminative dictionaries. Finally, the adaptive run-length smoothing algorithm and projection profile analysis are used to further refine the candidate text areas. The proposed method is evaluated on the Microsoft common test set, the ICDAR 2003 text locating set, and an image set collected from the web. Extensive experiments show that the proposed method can effectively detect texts of various sizes, fonts and colors from images and videos.     

Towards inverse modeling of turbidity currents: The inverse lock-exchange problem

A new approach is introduced for turbidite modeling, leveraging the potential of computational fluid dynamics methods to simulate the flow processes that led to turbidite formation. The practical use of numerical flow simulation for the purpose of turbidite modeling so far is hindered by the need to specify parameters and initial flow conditions that are a priori unknown. The present study proposes a method to determine optimal simulation parameters via an automated optimization process. An iterative procedure matches deposit predictions from successive flow simulations against available localized reference data, as in practice may be obtained from well logs, and aims at convergence towards the best-fit scenario. The final result is a prediction of the entire deposit thickness and local grain size distribution. The optimization strategy is based on a derivative-free, surrogate-based technique. Direct numerical simulations are performed to compute the flow dynamics. A proof of concept is successfully conducted for the simple test case of a two-dimensional lock-exchange turbidity current. The optimization approach is demonstrated to accurately retrieve the initial conditions used in a reference calculation.     

Why Audit a Web Server?: Auditing Web servers: IIS 4.0 — Part 1

A Web server is a wonderful business communication channel — it relays information cheaply, conveniently and, in theory, to almost everyone. Traditionally, in my experience at least, the Web is not an area auditors have taken particularly seriously: until recently, it has not been used to transact business, but simply to publicize the organization. Most companies run their Web applications on a firewall interface or interfaces outside the internal network, so the audit, more concerned with the core business applications, could rely on the firewall to protect against hackers, and treat the Web server as, at worst, a bastion host.     

The evidence framework applied to support vector machines

We show that training of the support vector machine (SVM) can be interpreted as performing the level 1 inference of MacKay's evidence framework (1992). We further on show that levels 2 and 3 of the evidence framework can also be applied to SVMs. This integration allows automatic adjustment of the regularization parameter and the kernel parameter to their near-optimal values. Moreover, it opens up a wealth of Bayesian tools for use with SVMs. Performance of this method is evaluated on both synthetic and real-world data sets.