Visual search tree profiling

Understanding how the search space is explored for a given constraint problem – and how it changes for different models, solvers or search strategies – is crucial for efficient solving. Yet programmers often have to rely on the crude aggregate measures of the search that are provided by solvers, or on visualisation tools that can show the search tree, but do not offer sophisticated ways to navigate and analyse it, particularly for large trees. We present an architecture for profiling a constraint programming search that is based on a lightweight instrumentation of the solver. The architecture combines a visualisation of the search tree with various tools for convenient navigation and analysis of the search. These include identifying repeated subtrees, high-level abstraction and navigation of the tree, and the comparison of two search trees. The resulting system is akin to a traditional program profiler, which helps the user to focus on the parts of the execution where an improvement to their program would have the greatest effect.     

Accuracy estimate and optimization techniques for SimRank computation

The measure of similarity between objects is a very useful tool in many areas of computer science, including information retrieval. SimRank is a simple and intuitive measure of this kind, based on a graph-theoretic model. SimRank is typically computed iteratively, in the spirit of PageRank. However, existing work on SimRank lacks accuracy estimation of iterative computation and has discouraging time complexity. In this paper, we present a technique to estimate the accuracy of computing SimRank iteratively. This technique provides a way to find out the number of iterations required to achieve a desired accuracy when computing SimRank. We also present optimization techniques that improve the computational complexity of the iterative algorithm from O(n ⁴) in the worst case to min(O(nl), O(n ³/ log₂ n)), with n denoting the number of objects, and l denoting the number object-to-object relationships. We also introduce a threshold sieving heuristic and its accuracy estimation that further improves the efficiency of the method. As a practical illustration of our techniques, we computed SimRank scores on a subset of English Wikipedia corpus, consisting of the complete set of articles and category links.     

Supervised Learning of Quantizer Codebooks by Information Loss Minimization

This paper proposes a technique for jointly quantizing continuous features and the posterior distributions of their class labels based on minimizing empirical information loss such that the quantizer index of a given feature vector approximates a sufficient statistic for its class label. Informally, the quantized representation retains as much information as possible for classifying the feature vector correctly. We derive an alternating minimization procedure for simultaneously learning codebooks in the euclidean feature space and in the simplex of posterior class distributions. The resulting quantizer can be used to encode unlabeled points outside the training set and to predict their posterior class distributions, and has an elegant interpretation in terms of lossless source coding. The proposed method is validated on synthetic and real data sets and is applied to two diverse problems: learning discriminative visual vocabularies for bag-of-features image classification and image segmentation.     

An Efficient Scaling Algorithm for the Minimum Weight Bibranching Problem

Let G=(VG,AG) be a digraph and let S ⊔ T be a bipartition of VG. A bibranching is a subset B⊆AG such that for each node s∈S there exists a directed s–T path in B and, vice versa, for each node t∈T there exists a directed S–t path in B.     

Analyzing the efficacy of using digital ink devices in a learning environment

There has been increased interest on the impact of mobile devices such as PDAs and Tablet PCs in introducing new pedagogical approaches and active learning experiences. We propose an intelligent system that efficiently addresses the inherent subjectivity in student perception of note taking and information retrieval. We employ the idea of cross indexing the digital ink notes with matching electronic documents in the repository. Latent Semantic Indexing is used to perform document and page level indexing. Thus for each retrieved document, the user can go over to the relevant pages that match the query. Techniques to handle problems such as polysemy (multiple meanings of a word) in large databases, document folding and no match for query are discussed. We tested our system for its performance, usability and effectiveness in the learning process. The results from the exploratory studies reveal that the proposed system provides a highly enhanced student learning experience, thereby facilitating high test scores.

Towards the development of a virtual environment-based training system for mechanical assembly operations

In this paper, we discuss the development of Virtual Training Studio (VTS), a virtual environment-based training system that allows training supervisors to create training instructions and allows trainees to learn assembly operations in a virtual environment. Our system is mainly focused on the cognitive side of training so that trainees can learn to recognize parts, remember assembly sequences, and correctly orient the parts during assembly operations. Our system enables users to train using the following three training modes: (1) Interactive Simulation, (2) 3D Animation, and (3) Video. Implementing these training modes required us to develop several new system features. This paper presents an overview of the VTS system and describes a few main features of the system. We also report user test results that show how people train using our system. The user test results indicate that the system is able to support a wide variety of training preferences and works well to support training for assembly operations.

---

Abductive Theorem Proving for Analyzing Student Explanations to Guide Feedback in Intelligent Tutoring Systems

The Why2-Atlas tutoring system presents students with qualitative physics questions and encourages them to explain their answers through natural language. Although there are inexpensive techniques for analyzing explanations, we claim that better understanding is necessary for use within tutoring systems. In this paper we motivate and describe how the system creates and uses a deeper proof-based representation of student essays in order to provide students with substantive feedback on their explanations. We describe in detail the abductive reasoner, Tacitus-lite+, that we use within the tutoring system. We also discuss evaluation results for an early version of the Why2-Atlas system and a subsequent evaluation of the theorem-proving module. We conclude with the discussion of work in progress and additional future work for deriving more benefits from a proof-based approach for tutoring applications.     

Identification of the Primary Factors Determining the Specificity of Human VKORC1 Recognition by Thioredoxin-Fold Proteins

Redox (reduction–oxidation) reactions control many important biological processes in all organisms, both prokaryotes and eukaryotes. This reaction is usually accomplished by canonical disulphide-based pathways involving a donor enzyme that reduces the oxidised cysteine residues of a target protein, resulting in the cleavage of its disulphide bonds. Focusing on human vitamin K epoxide reductase (hVKORC1) as a target and on four redoxins (protein disulphide isomerase (PDI), endoplasmic reticulum oxidoreductase (ERp18), thioredoxin-related transmembrane protein 1 (Tmx1) and thioredoxin-related transmembrane protein 4 (Tmx4)) as the most probable reducers of VKORC1, a comparative in-silico analysis that concentrates on the similarity and divergence of redoxins in their sequence, secondary and tertiary structure, dynamics, intraprotein interactions and composition of the surface exposed to the target is provided. Similarly, hVKORC1 is analysed in its native state, where two pairs of cysteine residues are covalently linked, forming two disulphide bridges, as a target for Trx-fold proteins. Such analysis is used to derive the putative recognition/binding sites on each isolated protein, and PDI is suggested as the most probable hVKORC1 partner. By probing the alternative orientation of PDI with respect to hVKORC1, the functionally related noncovalent complex formed by hVKORC1 and PDI was found, which is proposed to be a first precursor to probe thiol–disulphide exchange reactions between PDI and hVKORC1.     

Quantifying adherence of oxide scales on steels exposed to high temperature and pressure steam

Oxide scale exfoliation is a major concern in fossil fuel power generation because it can cause tube blockages and erode valves and steam turbine components downstream. There is still considerable scientific and commercial interest to improve the mechanistic understanding of oxide failures by developing models to predict exfoliation and the extent of tube blockage as a function of operating conditions and component geometries. Tensile testing inside a scanning electron microscope was conducted on ferritic–martensitic and austenitic steel specimens with the steam side (Fe,Cr)-rich oxides grown after exposures for up to 1000 h in steam with ~100 ppb O₂ at 276 bar and 550°C. Multiple oxide layer cracks and delamination events were observed and analyzed in detail during the tests. Results from the testing agreed well with earlier observations that had identified the failure location at the outer–inner oxide layer for all tested materials. Calculated adhesion energies identified the outer–inner oxide interface of alloy 347HFG as the weakest interface.