Nathan Eloe

Efficient determination of spatial relations using composition tables and decision trees

In order for Qualitative Spatial Reasoning applications to be both useful and usable, the information feedback loop between the computational engine and the user must be as seamless as possible. Inherently, computational geometry can be quite expensive, and every effort must be made to avoid inefficient or unnecessary calculations. Within the field of Region Connection Calculi, the 9-Intersection model often is used to determine the spatial relation between two regions. Consequently, optimization efforts typically focus on calculations involving the intersections between the interiors, boundaries, and exteriors of the regions, or the use of composition tables to narrow down the possibilities for the relations that can hold between two regions. The few implementations of spatial reasoners that have been attempted have been simply proofs-of-concept and/or have been limited to two dimensions. Herein we present a novel approach that combines the use of composition tables and decision trees to efficiently determine the spatial relation between two objects in 3D considering both connectivity and obscuration. This approach has been fully implemented for the VRCC-3D+ spatial reasoning system, and benchmarks are included to corroborate our claims of efficiency.

Scaffolding Version Control into the Computer Science Curriculum.

Version control systems (VCS) are widely-used in the software industry. They provide a powerful, collaborative framework that allows software engineers to work together effectively. VCS allow users to track changes and merge ongoing work into concurrently evolving software projects. Distributed VCS such as Git, allow a great degree of flexibility, and provide powerful options for managing personal code and evolving collaborative content. Power incurs responsibility, and introducing collaborative coding and version control tools to new developers can create many challenges. Yet these tools, once mastered, are crucial skills for professional developers. In this paper, the authors introduce VCS to computer science students both in a custom environment specifically designed to support new developers and in a commercially-available native environment suitable for more experienced students. Results show that proper introduction of these powerful tools can make early exposure a positive and valued experience.

VeCVL: A Visual Language for Version Control.

Version control systems (VCS), such as Subversion and Git, are pervasive in industry; they are invaluable tools for collaborative development that allow software engineers to track changes, monitor issues, merge work from multiple people, and manage releases. These tools are most effective when they are a part of a developer’s habitual workflow. Unfortunately, the use of these powerful tools is often taught much later in a developer’s educational career than other tools like programming languages or databases. Even an experienced student’s first experience with version control can be unpleasant. In this paper, the authors analyze the workflow of two common Version Control Systems with different version controls (Subversion and Git) to build a common visual language for these systems (Version Control Visual Language, or VeCVL), and show that the same visual language applies to other version control systems. Keywordscomputer science education, education technology, pedagogy, version control, visual language

Dual graph partitioning for Bottom-Up BVH construction

Bounding Volume Hierarchies (BVHs) are essential tools in performing collision detection on three-dimensional information. They reduce the number of expensive calculations required to determine whether or not two geometrical entities collide by using inexpensive calculations to rule out parts of the objects that could not possibly intersect. Quickly producing a high quality BVH is an important aspect of three-dimensional multimedia analysis. As such a powerful optimization, efficient and high quality BVHs are still an active area of research. Herein, the authors present a novel BVH representation that reduces the redundancy in the tree structure by allowing a node to contain an arbitrary number of children, as well as compressing non-unique nodes and combining their children. A new partitioning scheme using a graphical representation of the object is also presented to improve the quality of the generated BVH. HighlightsWe present a way to partition triangulated faces for BVH construction.A representation of an AABB tree that exploits a hashmap is examined.BVH representation and partitioning increase quality of BVH.

A More Efficient Representation of Obscuration for VRCC-3D+ Relations

Abstract—VRCC-3D+ is an implementation of a regionconnection calculus that qualitatively determines the spatialrelation between two 3D objects in terms of connectivity andobscuration. The eight connectivity relations are conceptually thesame as RCC8, but calculated in 3D rather than 2D. The fifteenobscuration relations are calculated using the projection of the 3Dobjects on a particular 2D plane and the distance of the objectsfrom the viewpoint. Herein we present a smaller, more preciseset of VRCC-3D+ obscuration relations that retains the qualitiesof being jointly exhaustive and pairwise disjoint. However, thisnew set of relations overcomes two problems that existed in theprevious set of fifteen relations: (1) lack of a precise mathematicaldefinition for a key predicate, InFront, and (2) lack of an intuitivemapping of converse relations.Index Terms—Computer vision, qualitative spacial reasoning,VRCC-3D, region connection calculus, spatial relations. I. I NTRODUCTION Q UALITATIVE spatial reasoning (QSR) in two dimen-sions is a well-studied field, and includes models suchas the connectivity-based RCC systems [1], [2], [3], andobscuration-based systems such as LOS-14 [4], OCS-14 [5],and OCC [6]. These systems, while expressive, do notaccurately portray the real world wherein objects exist andare perceived in three dimensions, not two. As computingpower increases and the need to analyze three-dimensionaldata (e.g., stereoscopic video, robotic vision, etc.) increases,two-dimensional reasoning systems can be inefficient, or eveninadequate, for sophisticated applications.To ameliorate the shortcomings of two-dimensional QSRsystems, Albath et al. developed RCC-3D [7], whicheventually evolved into VRCC-3D+ [8]. VRCC-3D+ usescomposite relations that express both connectivity andobscuration from a given perspective. The connectivity-basedrelations are the RCC8 relations (DC, EC, EQ, PO, TPP, TPPc,NTPP, NTPPc) defined in three dimensions; these relationshave been an ongoing focus of optimization and refinementin the implementation as a QSR system [9]. The obscurationportion of the composite relations are refinements on the basicconcepts of no obscuration (nObs), partial obscuration (pObs),equal obscuration (eObs), and complete obscuration (cObs).

Gitsubmit and VeCVL: Integrating Version Control in Introductory Computer Science Education.

Version control systems (VCS), such as Subversion and Git, are pervasive in industry; they are invaluable tools for collaborative development that allow software engineers to track changes, monitor issues, merge work from multiple people, and manage releases. These tools are most effective when they are a part of a developer’s habitual workflow. Unfortunately, the use of these powerful tools is often taught much later in a developer’s educational career than other tools like programming languages or databases. Even an experienced student’s first experience with version control can be unpleasant. In this work, an assignment submission system built around the Git version control system is introduced and analyzed for usability and suitability for use in entry level computer science classes. Keywordscomputer science education, education technology, pedagogy, version control, visual language

Interactive Visualization of Robustness Enhancement in Scale-free Networks with Limited Edge Addition (RENEA).

Error tolerance and attack vulnerability of scale-free networks are usually used to evaluate the robustness of these networks. While new forms of attacks are developed everyday to compromise infrastructures, service providers are expected to develop strategies to mitigate the risk of extreme failures. Recently, much work has been devoted to design networks with optimal robustness, whereas little attention has been paid to improve the robustness of existing ones. Herein we present RENEA, a method to improve the robustness of a scale-free network by adding a limited number of edges. While adding an edge to a network is an expensive task, our system, during each iteration, allows the user to select the best option based on the cost, amongst all proposed ones. The edge-addition interactions are performed through a visual user interface while the algorithm is running. RENEA is designed based on the evolution of the network’s largest component during a sequence of targeted attacks. Through experiments on synthetic and real-life data sets, we conclude that applying RENEA on a scale-free network while interacting with the user can significantly improve its attack survivability at the

VisCFSM: Visual, Constraint-Based, Frequent Subgraph Mining.

Graphs long have been valued as a pictorial way of representing relationships between entities. Contemporary applications use graphs to model social networks, protein interactions, chemical structures, and a variety of other systems. In many cases, it is useful to detect patterns within graphs. For example, one could be interested in identifying frequently occurring subgraphs, which is known as the frequent subgraph mining problem. A complete solution to this problem can result in numerous subgraphs and can be time-consuming to compute. An approximate solution is faster, but is subject to static heuristics that are beyond the control of the user. Herein we present VisCFSM, a visual, constraint-based, frequent subgraph mining system which allows the user to dynamically specify a variety of constraints on the subgraphs to be found while the mining algorithm is running. The constraint specification interactions are performed through a visual user interface, thereby facilitating a form of visual algorithm steering. This approach can be integrated with any frequent subgraph mining algorithm. Most importantly, this approach has the potential for the user to better, and more quickly, find the information that is of most interest to him/her in

PhysQSR: Improving Reasoning in Three Dimensions and Time With Image Processing and Physics.

Qualitative Spatial Reasoning (QSR) is an exceptionally powerful tool in the fields of computer cognition and automated computer reasoning. Recent results have shown the potential feasibility of pairing image processing techniques with basic principles of physics that humans inherently understand in order to allow the computer to extrapolate additional information about the environment in which it exists. Initial results showed that, while using the tenets of conservation of mass, conservation of energy, and inertia allowed the computer to gain more information than was initially apparent, noise in the perceived input data resulted in the software erroneously reasoning about the state of the system. Hence improving the image processing techniques used in analyzing the data should ameliorate the errors in reasoning. In this paper, the authors investigate this claim, and present a system that allows a more precise and correct computational view of the environment.