Final Psst! 19: (29 Mar, 1:15 PM, EM 3.06) Speakers: Rajiv Mural, Xinghui Dong, Eshrag Refaee

It's my pleasure to invite you all to the last Psst seminar this academic year:

Final Psst! 19:
(29 Mar, 1:15 PM, EM 3.06)


Rajiv Murali

The causes of system failure are typically rooted in the complex structures of software systems and their world contexts. An approach to problem analysis is presented in which problems are decomposed into sub problems of recognized classes. These classes can be captured by problem frames, which identify domain structures and interfaces in the problem world. In this presentation, we will investigate this approach using the Avionics Control System (BAE Systems), and we address how Problem Frames can be used to identify potential system failures and mitigate it.

Xinghui Dong: Human Perception Based Computational Texture Similarity Evaluation under a Unified Framework


Texture similarity is very commonplace in the fields of texture classification and retrieval.
In our study, we first compare similarity matrices obtained by a series of either classical or state-of-the-art computational feature extraction algorithms with those acquired in the free grouping and pairs of pairs experiments under a unified evaluation framework. Experimental results show that nearly all these approaches cannot agree on human perceptual similarity matrices well enough. The highest agreement rate obtained by these methods on 5 different resolutions and the combination of all 5 resolutions is 68.2%.

Another human-involved experiment is conducted then in order to investigate what factors will influence human perceptual texture similarity. In this part, 82 texture pairs which are most difficult for most computational approaches to agree on perceptual data are selected to be stimuli. 98 words which are divided into 11 groups are presented to participants and can be chosen at will by them to describe the similarity between each given texture pair. Four groups of words are used most frequently, including regular, lined, netlike and bumpy (in sequence of frequencies). It is obviously that regularity, directionality and bumpiness are very important for human to perceive similarity of textures involved in this study. However, all existed computational algorithms investigated here cannot perform comparatively.


Eshrag Refaee : Knowledge Gathering with Visual Techniques

Recognising the vital role that visualisation plays to capture, represent, and analyse complex data in clear structure has provided a strong foundation towards the information visualisation.  Visualisation is a type of the non-verbal communication methods that provide means of retrieving pictures to facilitate wordless communication. Scientifically, the right part of our brains is responsible about the visual type of thinking, and it is also responsible about the emotions and the intuitive thinking. This implies that the visual thinking approach was not recently established, nor at any particular time, but it is built-into our minds to allow human beings interacting with each other and with their environment. The recognition of this remarkable feature has led to a powerful revolution in the visual thinking, which might have great implications in various fields like information gathering, education, and special needs communication languages. Consequently, new concepts have been emerged and seriously considered i.e. rich pictures, visual language, image scheme, concept map, and image streaming.

Addressing the field-related social aspects has become an issue of importance in a variety of domains. The degree of accuracy and completeness of the captured information is significantly important for the field researcher. In this context, questionnaires have been identified as one of the most common methods applied to gather individual responses. popularity of the questionnaires might have a negative impact on the response rate. In addition, the possibility of having responses that might result in inaccurate or inconclusive results is a considerable risk. Such drawbacks have inspired researchers to attempt to evolve more effective research tools. For instance, the use of visualisation could play a major role as a powerful instrument in educational, personal, and commercial uses.

In this research, we aim to employ the features of rich picture(s) (RP) that have been introduced as a visual representation of a complex problem situation. RP is distinguished by its ability to capture detailed cultural and social aspects that represent individual’s view in a more comprehensive pattern than the conventional approaches. We assume that we can emerge the features of questionnaires and RP to develop a visual questionnaire by comparing their key features. We also aim to examine and validate the ability of this approach in capturing as rich and accurate information as possible.

[Cancelled] Psst! 17

This week there will be no seminar Due to speakers not being able to attend. Sorry

Psst! 16:8 March, 1:15 pm, EM 3.06 Wasit Limprasertu, Lin Qi,Christopher Ritter,

room EM 3.06, Thursday 8th March, 1:15 pm

Wasit Limprasert: Tracking multiple human subjects within a camera network

Abstract: We present an approach to tracking multiple human subjects within a camera network. A particle filter framework is used in which we combine foreground-background subtraction with a novel approach to texture learning and likelihood computation based on an ellipsoid model. As there are inevitable problems with multiple subjects due to occlusion and crossing, we include a robust method to suppress distraction between subjects. To achieve real-time performance, we have also developed our code on a graphics processing unit to achieve a 10-fold reduction in processing time with an approximate frame rate of 10 frame per second.

Qi, Lin: Measuring Perceived Gloss of Rough Surfaces

We have studied how perceived gloss varies with the change of both mesoscale and microscale roughness on 3D surface textures. The mesoscale roughness was changed by varying the roll-off factor (β) of 1/fβ fractal noise surfaces. The microscale roughness was changed by varying the microscale roughness parameter α in the microfacet reflection model. An HDR real-world environment map was used to provide complex illumination and a physically-based path tracer was used for rendering the stimuli. Each simulated surface was rotated about its vertical axis to generate an animated stimulus. Eight observers took part in a 2AFC experiment, and the results were tested against conjoint measurement models. We found that the perceived gloss changes non-monotonically with β (an asymmetric bell curve), and monotonically with α. Although both β and α significantly affect perceived gloss, the additive model is inadequate to describe their interactive and nonlinear influence, which is at variance with previous results.

Ritter, Christopher: Can virtual characters learn to care? How to increase an agent's ability to empathise!

My PhD is motivated by the attempt to find a computational model for intelligent and emotional characters that can learn to be more empathic towards other agents (a dynamic adaptation process). I intend to achieve this goal by including models of emotional (self-)awareness - affective empathy - and cognitive empathy into the agent's reasoning. In order to generate a more natural looking model of the simulated learning process, I started by looking into how real people can learn/be taught to be more empathic. I will present a WIP showcase I am creating based on findings in the area of conflict resolution. I will use the showcase to create a model for the learning process as observed in this particular situation. A later implementation of the showcase will be used to evaluate the model. No explicit implementation model is presented yet!

Psst! 14: Manjula Dissanayake: Feature Selection and Classification in Complex Biological Datasets; , Ioanna Filippou : TBC

Psst this week:
room EM 3.03, 1:15 PM, Thursday 23rd Feb 2012

Ioanna Filippou:
TBC
TBC


Dissanayake, Manjula: Feature Selection and Classification in Complex Biological Datasets

Feature selection continues to grow in importance in many areas of science and engineering as large datasets become increasingly common. For example, DNA micro arrays can measure expression levels of many thousands of genes simultaneously.

For effective data mining in such datasets, tools are required that can reliably distinguish the most relevant features. Previous work has shown promise for an evolutionary algorithm/classifier combination (EA/k-NN), which, in successive phases of the same algorithm, serves first as the feature selection mechanism and second as the machine learning method yielding an accurate classifier.

This work has been followed up by investigating the effect of introducing a weighting scheme for k-NN. The weighting scheme used here has been adopted from the adaptive local hyper plane method. The areas investigated included the way to apply adaptive weights to the k-NN method, the way to scale the weighting scheme up to work with large biological datasets and its effectiveness when combined with two phase EA/k-NN algorithm.

Psst! 2 Feb:Alzahrani,Mohammed@better Web Apps;Suttie, Neil@Serious Games; Methven, Tom@Stereo Gloss Perception; Orzechowski, Pawel@Textile Simulators

Thursday 2 Feb, 1:15, Room EM 3.03


Alzahrani, Mohammed: Model Checking Data-Sensitive and Time-Sensitive Web Applications

Modeling of web-based applications is fundamental to capture and understand their behaviour. As a verification technique, model checking can assist in finding design flaws and simplifying the design of a web application, as a result the quality and the security of web application can be improved. We use the model checker SPIN to simulate and verify essential security and navigation properties of web-based data-sensitive and time-sensitive application. We integrate discrete time in Promela (SPIN's input language) model to allow us represent realistic web applications properties and track sequence of actions in a secure model and in the present of a malicious behaviour. We discuss a comparison between the tools SPIN and UPPAAL in modelling and verfiying web applications



Suttie, Neil: Serious Games:Maintaining Engagement in Learning Through Game-play

Serious Games and intelligent pedagogical systems offer the potential for providing effective and engaging learning experiences tailored to the need of the individual student. They can present a promising means of enhancing problem solving, analytical thinking, decision making and other skills not easily learned in the class room setting.

Essential for the advancement of Serious Games is careful balancing of educational and ludic content. However, there is little understanding of how educational outcomes relate to game-play and concern that greater engagement in gameplay elements may even prove diversionary to the learning process. Therefore, It is necessary to develop a clear understanding of the relationship between game-play, player engagement and how these relate to relevant educational strategies. In this talk, I will discuss the challenges, possible solutions and work conducted thus far in linking learning and serious games through their implemented mechanics.


Methven, Tom: Does Stereopsis Affect Gloss Perception?

Over the past 20 years, there have been many studies looking at how highlight disparity affects an observer's perception of glossiness. Most of these studies have used relatively smooth surfaces, and simple lighting models, meaning their results might not account for all the
complex interactions of lighting and stereopsis with real surfaces. In this talk, I present the results of a pilot experiment designed to look into this problem. These results seem to imply that the relationship between gloss perception, highlight disparity and roughness is more complex than previously reported. In addition, I talk about future work designed to solve problems encountered with the pilot experiment.


Orzechowski, Pawel: Touch-screen Textile Simulators.

Consumers like to touch and deform fabrics that they are about to buy, but currently available devices do not provide the touch and feel modality. iShoogle is a textile simulator for iPad that uses movements natural in handling clothes (crunching, pulling, pinching, etc.) to interact with on-screen movies of fabrics being deformed (crunched, pulled, pinched etc.). Experiments were carried out to determine if the added interactivity afforded by iShoogle could help to describe textile qualities and thus aid in the creation of next generation digital swatches.

New Room for Psst! (and changing almost every other week)

Thursday 1.15-3.15


EM 3.03 for 26/1, 2/2, 9/2, 16/2, 23/2


EM 3.06 for 1/3, 8/3, 15/3, 22/3, 29/3

Psst! 26 Jan: Macvean, Andrew@Adaptive exergames; Totoo, Prabhat@Haskel and n-body problems; Weallans, Allan@Distributed drama management

Thursday 26 Jan, 1:15, Room G45 or G46 [notice will be displayed on the door]

Macvean, Andrew: Designing Adaptive Exergames for Adolescent Children: Lessons Learned from A Preliminary Study

With evermore children in the Western world adopting a sedentary lifestyle, there is an increased emphasis towards physical activity interventions. Thanks to progressions in ubiquitous technologies, exergames, games that facilitate and encourage exercise, have emerged as a potential means to motivate children to exercise in a context they find both familiar and enjoyable. While early research has shown the potential of the genre, few games have been designed specifically with children in mind, accommodating their unique demands. In our work, we aim to close this gap by investigating how children react to, and make use of an exergame designed specifically with them in mind. The aim is to understand how different demographics of children (gender, gaming background, exercise background, etc) react to a location-aware exergame, in order to in the future build more accurate adaptive exergames, suitable for the specific requirements of children users. In this talk we present the first step towards this, the results of a preliminary study on our location-aware exergame - iFitQuest. Our results show early insight into how different users react to our game, and provide lessons on how to approach the design task.


Totoo, Prabhat: Parallel Haskell implementations of the n-body problem

The n-body problem is a problem of predicting the motion of a system of N bodies that interact with each other gravitationally. N-body algorithms are used in several areas such as molecular dynamics, astrophysics.
Several methods exist to solve the problem including bruteforce body-to-body comparison and the more sophisticated Barnes-Hut algorithm. We look at multicore implementations of the problem for simulation consisting of huge input. The implementations cover the use of 3 different parallel programming models in Haskell: GpH, Par monad and Eden. The performance of these models are compared in terms of changes required to the sequential algorithms and runtime/speedup.

Weallans, Allan: DISTRIBUTED DRAMA MANAGEMENT: BEYOND DOUBLE APPRAISAL IN EMERGENT DRAMA

Emergent drama relies on the autonomous actions of AI character agents. While this has many advantages over a more directorial approach, it is not guaranteed to favour dramatically interesting actions or a "story-like" structure. Using affective impact as a surrogate for drama, Double Appraisal was a step towards encouraging dramatic events in autonomous agents. This talk will briefly outline the limitations of Double Appraisal as implemented, and discuss how Distributed Drama Management builds on the Double Appraisal work to help to produce more structured and dramatically interesting narrative experiences while retaining the autonomy of the agents and the benefits associated therewith.