Feature selection plays a significant role in data mining and machine learning. It is challenging to determine how many features are necessary to form an optimal feature subset. To address this challenge, an innovative visual 2D feature selection framework is introduced, in which the feature discernibility and independence are defined to evaluate its capability for classification and its relevance to other features, respectively. All features are represented in 2D space with discernibility as$x$-axis and independence as$y$-axis. The features located in the upper right corner represent high discernibility and high independence, so comprise the optimal feature subset. This leads to the formation of a family of feature selection algorithms. Three such algorithms are proposed in this paper referred to as FSDIE, FSDIR, and FSDIS (Feature Selection based on the Discernibility and the Independence, respectively, of Exponent, Reciprocal, and anti-Similarity). To speed-up these three algorithms, a clustering based feature preselection first eliminates some unrelated and redundant features. Extensive experiments on UCI datasets, face datasets and gene expression datasets demonstrate that these three 2D feature selection algorithms are superior to the state-of-the-art methods indicating the power of our 2D feature selection framework.

This paper describes the process of developing a visualisation system with the use of Game Engine. A bottom up approach was used in developing the system, which allows process alignment with the Game Engine structure. The approach allows large abstract data processing, which are channelled through the components of visual interfaces and supported with relational data input structure. The developed system provides automation in much of the pipeline in producing dynamic visualisation selections. As case studies, the UK Research Assessment Exercise (RAE2008) and 2010 World Oil Reserves datasets were used to demonstrate the usability of the Game Engine based visualisation system.

Facial fiducial point localization is a crucial step for most facial analysis applications, e.g., face recognition, expression recognition and facial aging simulation. Although state-of-art methods have the ability to provide good salient point location on frontal faces, finding a global solution under large variations caused by off-plane rotations and exaggerated expression changes is still a challenge. In this paper, we present a system with a two-level shape model to facilitate accurate facial fiducial point localization. In the first level, two local component models interact with each other in order to offer novel shape constraints. At the same time, the clamped local shape model provides constrained non-linear shape initialization for better convergence performance of the shape model as a whole. The experimental results confirm that the proposed method is capable of dealing with the face alignment under large shape variations.

Natural image statistics is an important area of research in cognitive sciences and computer vision. Visualization of statistical results can help identify clusters and anomalies as well as analyze deviation, distribution, and correlation. Furthermore, they can provide visual abstractions and symbolism for categorized data. In this paper, we begin our study of visualization of image statistics by considering visual representations of power spectra, which are commonly used to visualize different categories of images. We show that they convey a limited amount of statistical information about image categories and their support for analytical tasks is ineffective. We then introduce several new visual representations, which convey different or more information about image statistics. We apply ANOVA to the image statistics to help select statistically more meaningful measurements in our design process. A task-based user evaluation was carried out to compare the new visual representations with the conventional power spectra plots. Based on the results of the evaluation, we made further improvement of visualizations by introducing composite visual representations of image statistics.

In written and spoken communications, figures of speech (e.g., metaphors and synecdoche) are often used as an aid to help convey abstract or less tangible concepts. However, the benefits of using rhetorical illustrations or embellishments in visualization have so far been inconclusive. In this work, we report an empirical study to evaluate hypotheses that visual embellishments may aid memorization, visual search and concept comprehension. One major departure from related experiments in the literature is that we make use of a dual-task methodology in our experiment. This design offers an abstraction of typical situations where viewers do not have their full attention focused on visualization (e.g., in meetings and lectures). The secondary task introduces "divided attention", and makes the effects of visual embellishments more observable. In addition, it also serves as additional masking in memory-based trials. The results of this study show that visual embellishments can help participants better remember the information depicted in visualization. On the other hand, visual embellishments can have a negative impact on the speed of visual search. The results show a complex pattern as to the benefits of visual embellishments in helping participants grasp key concepts from visualization.

Abstract Over the past decade, computer scientists and psychologists have made great efforts to collect and analyze facial dynamics data that exhibit different expressions and emotions. Such data is commonly captured as videos and are transformed into feature‐based time‐series prior to any analysis. However, the analytical tasks, such as expression classification, have been hindered by the lack of understanding of the complex data space and the associated algorithm space. Conventional graph‐based time‐series visualization is also found inadequate to support such tasks. In this work, we adopt a visual analytics approach by visualizing the correlation between the algorithm space and our goal – classifying facial dynamics. We transform multiple feature‐based time‐series for each expression in measurement space to a multi‐dimensional representation in parameter space. This enables us to utilize parallel coordinates visualization to gain an understanding of the algorithm space, providing a fast and cost‐effective means to support the design of analytical algorithms.

Computerised facial aging estimation, which has the potential for many applications in human-computer interactions, has been investigated by many computer vision researchers in recent years. In this paper, a feature-based discriminant subspace is proposed to extract more discriminating and robust representations for aging estimation. After aligning all the faces by a piece-wise affine transform, orthogonal locality preserving projection (OLPP) is employed to project local binary patterns (LBP) from the faces into an age-discriminant subspace. The feature extracted from this manifold is more distinctive for age estimation compared with the features using in the state-of-the-art methods. Based on the public database FG-NET, the performance of the proposed feature is evaluated by using two different regression techniques, quadratic function and neural-network regression. The proposed feature subspace achieves the best performance based on both types of regression.

Abstract In this paper, we propose a novel approach for automatic generation of visualizations from domain‐specific data available on the web. We describe a general system pipeline that combines ontology mapping and probabilistic reasoning techniques. With this approach, a web page is first mapped to a Domain Ontology, which stores the semantics of a specific subject domain (e.g., music charts). The Domain Ontology is then mapped to one or more Visual Representation Ontologies, each of which captures the semantics of a visualization style (e.g., tree maps). To enable the mapping between these two ontologies, we establish a Semantic Bridging Ontology, which specifies the appropriateness of each semantic bridge. Finally each Visual Representation Ontology is mapped to a visualization using an external visualization toolkit. Using this approach, we have developed a prototype software tool, SemViz, as a realisation of this approach. By interfacing its Visual Representation Ontologies with public domain software such as ILOG Discovery and Prefuse, SemViz is able to generate appropriate visualizations automatically from a large collection of popular web pages for music charts without prior knowledge of these web pages.

Abstract Aging has considerable visual effects on the human face and is difficult to simulate using a universally‐applicable global model. In this paper, we focus on the hypothesis that the patterns of age progression (and regression) are related to the face concerned, as the latter implicitly captures the characteristics of gender, ethnic origin, and age group, as well as possibly the person‐specific development patterns of the individual. We use a data‐driven framework for automatic image‐based facial transformation in conjunction with a database of facial images. We build a novel parameterized model for encoding age‐transformation in addition with the traditional model for face description. We utilize evolutionary computing to learn the relationship between the two models. To support this work, we also developed a new image warping algorithm based on non‐uniform radial basis functions (NURBFs). Evolutionary computing was also used to handle the large parameter space associated with NURBFs. In comparison with several different methods, it consistently provides the best results against the ground truth.

Using only simple stimuli from the local environment to coordinate the activity of the group, social insects such as termites build highly complex structures. This was discovered by Grasse (1959) and is known as stigmergy. We employ such a distributed approach for meshing finite element domains. In two dimensions, a triangular mesh is a discretization of a domain into a set of approximately equilateral triangles. The finite element method can then be used to solve fluid flow problems at the vertices in discrete time. The size of the triangles is related to rate of change of the solution values and hence meshing is a complex task. We use a genetic algorithm to evolve a set of rules which can be used by a colony of termite-like agents to mesh an arbitrary domain. Work on building activity of swarms has shown that stigmergic algorithms can produce more complex structures than sequential algorithms which are dependent on past building activity. We compare our stigmergic algorithm with a similar sequential algorithm. Whilst the sequential algorithm generalizes poorly to new domains, our stigmergic approach generalizes well and produces better quality meshes on the examples considered. In this approach, meshing is in response to conditions in the local environment, hence changing conditions can easily be accommodated, making it particularly useful for applications such as injection moulding where the mesh must 'grow' with the domain.

Abstract This paper investigates several sparse matrix representation schemes and associated algorithms in Haskell for solving linear systems of equations arising from solving realistic computational fluid dynamics problems using a finite element algorithm. This work complements that of Wainwright and Sexton (1992) in that a Choleski direct solver (with an emphasis on its forward/backward substitution steps) is examined. Experimental evidence comparing time and space efficiency of these matrix representation schemes is reported, together with associated forward/backward substitution implementations. Our results are in general agreement with Wainwright and Sexton's.

Abstract Experiences are described of parallelizing a functional finite‐element program (written in Haskell) for the solution of computational fluid‐dynamics problems. A transformation prototyping approach using a system developed at the University of York, which simulates idealised parallel machines on a conventional sequential workstation, is presented. In this paper we demonstrate: (a) the relative simplicity of the functional approach for parallelizing a complex program compared with the conventional procedural approach; (b) the suitability of functional languages for prototyping parallel algorithms to improve an implementation; and (c) the considerable assistance provided by the simulator.

Throughout the paper A will be a transitive set closed under finite subsets and the formulas in various classes mentioned are allowed to contain parameters from A (or from B in §2). By use of a refinement of Moschovakis' notion of the game-quantifier [13], [14], [15] we are able to obtain a game-theoretic description of s-Π11 predicates over countable sets which then leads to a classification of positive Σ1 inductive sets. Similar results are then proved for certain sets of cofinality ω. As a consequence we obtain the compactness results of Green [8], [11], Nyberg [16] and Makkai [12]. The use of games to classify inductive sets was initiated by Moschovakis [13], [14], [15] and has been extended to Q-inductive sets by Aczel [2]. Games were also used in a slightly different setting by Vaught [18] and Makkai [12]. In fact, Vaught's proof of the compactness theorem is very close to our proof in §1 and Makkai's extension to cofinality ω sets uses a result similar to Theorem 3 in §2. We are indebted to the referee for many helpful suggestions, in particular, for bringing to our attention the related works of Vaught and Makkai cited above.