Abstract:

In this talk, brief introduction about the mechanisms of our color vision followed by describing the color vision deficiency. Then, the talk notes that importance of the color universal design which is aiming to avoid the confusing colors used in the web pages, hazard maps, etc. Finally, I will show our research project aiming to realize the color universal design by using the computer-simulation-based detection and modification of the confusing colors and functional spectral filter which simulate color vision deficiency simply by putting it in front of your eyes.

Abstract:

Multimedia technology can enhance the efficacy of educational systems to a great extent by facilitating effective domain knowledge representation. However, just the collection of multimedia objects does not guarantee proper learning. An important aspect in the learning process is the proper interaction of the learner with the educational system interface, especially when learning is recognised as a complex activity (or process) combining various factors such as information retrieval, navigation, and memorisation. This talk will demonstrate the Multiple Representation (MR) approach as guiding the process of multimedia objects selection, navigational objects selection, and integration of multimedia objects to suit different learner needs.

Abstract:

This is a first part of a mini series of workshops aimed at helping PhD students for writing their PhD thesis. In the first session, we perform simple quantitative analysis of the previous thesis in our department. The focus in the first session is on thesis collected from individual publications.

Abstract:

This is the second part of a mini series of workshops aimed at helping PhD students for writing their PhD thesis. In the second session, we continue the theme based on the discussion and results of the first session. The process from submission to dissertation is also outlined.

Abstract:

The third part focuses on how to write the introductory part of thesis consisting of individual publication. Questions about the structure, when to start writing, how to proceed with it, how to desing the structure, which is the proper balance of material, and how to write the summary of publications will be discussed. Language and style of writing will also be considered with a brief example.

Abstract:

Globalization is not longer a trend, but a reality. There is no doubt that internationally shipped products and systems need to be adapted to the foreign markets. Tradeoffs have to be made on whether one technical system fits all international requirements or different variations of the system have to be developed for different geographical and cultural regions. This talk intends to give first a small introduction to international and cross-cultural systems development addressing the issues mentioned above. Second, the main part of the presentations will deal with the setup and conduction of international user research by explaining the most important aspects illustrated by a case study, where the use of mobile technology has been researched in Germany and Mexico with qualitative and quantitative methods. The talk is part of Usability Engineering 2006 course, welcome!

Abstract:

The conventional Chest X-ray Radiography is ubiquitous in diagnosis of abnormitly in clinic practice. The interpretation of Chest X-ray images is still challenging due to the image quality, the inherent anatomical sturctures, and in particular due to the differentiation of clinically image understanding by radiologists. This also incurs a severe missing of abnormality missing in practice. However, an increasingly important and popular computerized interpretation for the Chest X-ray radiography arises in recent years, CAD (Computer Aided Detection / Diagnosis), which is utilized as a second reader in clinical diagnosis. In this talk, a briefly introductive review is presented on algorithmic framework of computer aided detection/diagnosis for Chest X-ray radiography. In addition, a discussion wiil be preferably conducted on the application of classfication algorithms and clustering algorithms in the underlying CAD system.

Abstract:

Best lossless compression results of color map images have been obtained by dividing the color maps into layers, and by compressing the binary layers separately by using an optimized context tree model that exploits inter-layer dependencies. In this paper, we extend the previous context tree based method to operate on color values instead of the binary layers. We generate an n-ary context tree by constructing a complete tree up to a predefined depth, and then prune out nodes that do not provide improvement in compression to generate sub-optimal context tree with incomplete structure. Experiments show that the proposed method outperforms existing methods for a large set of different color map images.

Abstract:

Due to last decade's development of rapid DNA-sequencing technique amount of sequence level information on genes and whole genomes of organisms has rapidly increased. This has further facilitated development of various laboratory based genome wide screening technologies that utilize the availability of genomic sequence. Such methods like DNA-microarrays, protein arrays, analysis of mutations and metabolic profiling are currently in standard use in molecular biology and medicine. Common for the technologies is that they all perform simultaneous screening of a very large gene set, possibly whole genome, and based on that produce information on gene activity levels as numerical measurements. From such outcome possibly large sets of genes that behave similarly under different environmental conditions, chemical treatments, or diseases can be derived. The question is then why those genes show such behaviour, and also why not the others? This question should be answered in the light of current knowledge we have on genes, the information that has accumulated in the biological databanks, gene description databases, or literature collections during the genetic research of last few decades.

In the presentation the biological problem of interpreting outcome from genome wide technologies with the help of external knowledge on genes will be formulated into data mining problems. This is followed by the description of some novel methods we have developed that provide advance over prevailing practice in such analysis. Methodological background is in topic analysis, clustering and segmentation. One example of several is a novel method for analysing genes by using their chromosomal locations which have became available for several different species. Here the biological problem is to found areas of chromosome that associate with the set of genes found behaving similarly under some condition. Such condition could be some disease e.g. cancer that is known to make changes in the activity of neighbouring genes of chromosome. Rather than one type of cancer we might be interested to analyse multiple cancer types and find the areas of chromosome where some combination of them is associated.

In our research we have turned the chromosomal analysis into mathematical formulation. In our setup the consecutive genes of chromosome are represented as a sequence of observations. The different properties of studied biological samples like environmental conditions, chemical treatments, or diseases are indicated as nominal variables and their data classes. Therefore each observation arises from multinomial or binomial distributional family and indicates the association of a gene with the biological property. The aim is then to find the regions from sequence where some data class or combination of multiple classes is exceptionally dominant. We see this as a change-point analysis problem. It is concerned with change-points underlying in the sequential data that separate sub-sequences arising from distributions with dissimilar parameter values. The locations of change-points, and in the most applications the number of them, are unknown and should be detected. We have developed a novel bayesian model selection method for evaluating the change-point model. We use divisive hierarchical segmentation algorithm to optimize the model with respect to the developed evaluation measure. We have also constructed a framework for comparison of different segmentation methods which we use to make comparison between our method and different existing model selection methods. For the evaluation and comparison of methods and the biological application we have implemented CATALIST software.

Abstract:

To implement real intelligence or adaptivity, the models for intelligent tutoring systems should be learnt from data. However, the educational data sets are so small that machine learning methods cannot be applied directly. In this paper, we tackle this problem, and give general outlines for creating accurate classifiers for educational data. We describe our experiment, where we were able to predict course success with more than 80% accuracy in the middle of course, given only hundred rows of data.
Full paper as PDF can be found here.

Abstract:

The aim of this study is to find out what kind of elements (visions, strategies, innovations, processes, clients) are needed for the management of a university research group in order for the group to achieve its goal to be a centre of excellence in research. The research has been done by reviewing the literature on the area of management of hot groups in companies and organizations.

The empirical part of the study uses the qualitative research methods. The population of the research is groups that have the center of excellence in research status in Finland. Seven groups were selected randomly and their leaders were interviewed using general interview with interview guide or "schedule". In addition, one professor from US top university was interviewed. The interviews were analyzed by coding the data to 71 classes or themes. The research questions were a) how the centre of excellence is created and b) how does the centre of excellence work.

Based on the research it can be said that the centre of excellence in research is often created by ecological growth model presented in the study. Normally these groups cannot be created by top-down model.

The most important value in the centre of excellence in research is the quality of the research and the aim to discover deep scientific knowledge. The groups feel uncomforted of measuring their number of publications and journals. Also, the academic freedom is understood to be a major value among these research groups. It can be described as a scientific bohemian. The motivating factor in research work is scientific work and the achievements of scientific goals.

A value in the top research group is also open and conversational atmosphere. The research culture is inherited from older generations. It was clearly seen that the founder or previous leader of the group still had a strong effect on the way the research group works.

Abstract:

3D virtual reality can often provide a realistic presentation of information to the user. With the help of description languages for virtual reality (e.g. VRML, X3D) powerful means are available which can be presented on modern personal computer in a web browser featured with a plug-in.

The goal of this research is to provide the same functionality to users with mobile phones. For this purpose functionality of mobile phones has to be examined. In order to realise optimisation for mobile phones a proxy is introduced to perform the data optimisation for the corresponding mobile phone. Improving the realism, methods to display the 3D virtual reality as stereoscopic presentation are examined.

Abstract:

A two stage lossless compression method based on binary tree representation of colors and on context-based arithmetic coding has been recently proposed. We propose two improvements for this method: merge-based color quantization instead of the original splitting strategy, and context tree modeling optimized for each layer separately. The proposed method achieves better compression performance, and a better reproduction quality in the color progression

Abstract:

Tumaini University has been developing the use of information and communication technology (ICT) in education since the mid-90s. Their first systematic ICT development project called Internet Project Strategic Plan (IPSP) focused on obtaining infrastructure for accessing information; instilling ICT skills into staff and students, and installing and servicing computers to support learning activities. With this in mind, we developed the CATI model (Contextualize, Apply, Transfer, Import) to support sustainable ICT development projects on the basis of our evaluation of the IPSP project and our previous experiences in contextualized ICT education. During the past few years we have focused our research efforts on contextualizing ICT education. One of the outcomes of this initiative was a contextualized Introduction to Programming course (2004-2005). In the presentation I will present analyzes of the course design and implementation by using the CATI model.