Institute of Computer Science, J. Liivi 2, 50409, Tartu, Estonia. Phone: (+372) 737 5445, e-mail: ics [ät] ut.ee, web: www.cs.ut.ee
Supervisor(s): Fredrik Milani and Marlon Dumas
Business processes are the operational backbone of modern organizations. Their continuous improvement is crucial to achieve organizational objectives, be it with respect to efficiency, quality, compliance or agility. Accordingly, a common task for business process analysts is to discover and assess business process improvement opportunities. Current approaches to discover business process improvement opportunities are manual and expertdriven. In these approaches, execution data are used to assess opportunities derived from experience and intuition rather than to discover these opportunities in the first place.
This PhD project will build the foundations of a new generation of process improvement methods that do not exclusively rely on guidelines and heuristics, but rather on a systematic exploration of a space of possible changes derived from process execution data.
The PhD project will start with a literature review of business process improvement methods and of process mining methods. This literature review will be used as a basis to design an initial method for identifying improvement opportunities from business process execution data. The initial method will be validated and refined via case studies and interviews with business process improvement experts.
The PhD project will also lead to a method for visualization of business process improvement opportunities.
Supervisor(s): Eno Tõnisson, Marina Lepp
During the learning process it is very important to solve different programming tasks. The process of solving a programming task is generally invisible to the teacher. They only see the end result and maybe a few snapshots along the way. Using logs can give an overview of students’ task solving processes to the teacher. There are still quite few studies on the behavioral patterns of learners and have not worked out suggestions to learners and teachers based on these patterns.
The aim of the doctoral research is to develop and provide suggestions to learners and teachers based on different behaviour patterns of solving programming tasks. The research is planned to be based on analysis of Thonny logs. The Python IDE Thonny is created for learning and teaching programming. It supports educational research by logging user actions for replaying or analysing the programming process.
It is planned to analyse how code changes during the coding processes, and how it is related with behaviour patterns and how learners’ behaviour patterns depend on the task type; find ways how to detect patterns in log files and how these patterns describe differences in task solving; analyse which kind of differences can be found based on students with different background and programming experience; compare different students coding, for example to find out if they have written it themselves; examine how same students solve different tasks and analyse these solutions in perspective of rationality; figure out what types of learners can be distinguished on the basis of the behaviour patterns of task solving; work out which suggestions can be given to teachers, how to use the results in improving teaching methods, how these results can influence learning and learning materials.
Supervisor(s): Raivo Kolde
Standardisation of electronic medical record data holds the promise to solve many practical problems hindering the progress of re-using this data in furthering public health goals. However, the practical implementation of the standardisation can be costly and it is not clear if the benefits outweigh the costs. In this project, the goal is to explore the utility of the one particular format - OMOP, in facilitating research on Estonian electronic health record data. This will be achieved by conducting actual drug utilisation study on the data, where questions, such as prevalence of treatments, common prescription patterns, drug adherence and following of treatment guidelines are studied. By conducting such study together with international partners it is possible to validate potential benefits of the OMOP data model: using common analysis pipelines on multiple data sources and performing analysis in federated manner, where raw data is not shared with the analysts. The study itself provides opportunity for method and software development and the standardisation of the input format ensures the applicability of the software in future research projects. In addition to technical goals, the results of such representative international study will be of interest itself.
Supervisor(s): Piret Luik
The aim of the doctoral project is to find out how programming is learned on MOOCs and what factors affect the dropout rate of programming MOOCs in Estonia. The factors include the characteristics of participants, motivation, their perceptions of IT, learning activities, activity in the learning environment, usage of support systems (troubleshooters, helpdesk, forums), learning analytics. The qualitative analysis of the learners learning diaries and interviews of instructors and participants of four different MOOCs allows a deeper analysis of the differences in the learning patterns of completers and non-completers. The expected results of the doctoral thesis are a scientifically-based overview of the various factors that influence the dropout rate of programming MOOCs in Estonia and understanding how MOOCs allow new forms of teaching and learning. Also we will found out the periods in which the drop-out rate is the greatest. The project will help to answer to questions about which periods students need more support or motivation; how to improve the existing MOOCs so that their percentage completion rate is also higher without the helpdesk. Project results can be applied in addition to the MOOCs in developing programming courses and improving quality.
Supervisor(s): Dietmar Pfahl
System-level testing has become a highly automated practice in software industry. Growing size and configurability of software, which needs to be tested in ever-shorter cycles, has resulted in high demand for optimizing test suites with regards to efficiency and effectiveness. With the help of machine learning, three approaches with prototypical tool support will be developed for test suite optimization. The approaches provide (i) a method to develop a model that predicts in a mutation testing context which mutants won’t be killed, thus informing about missing tests in the test suite, (ii) a method to support semi-automatic test oracle generation, thus completing automatically generated test data, and (iii) a method to detect usage profile differences between test execution logs and end user execution logs, thus informing about gaps in the test suite. The three approaches will be integrated and evaluated in case studies with industry.
Supervisor(s): Huber Flores
Energy-efficiency remains a critical design consideration for mobile and wearable systems, particularly those operating continuous sensing. Energy footprint of these systems has traditionally been measured using hardware power monitors (such as Monsoon power meter) which tend to provide the most accurate and holistic view of instantaneous power use. Unfortunately, applicability of this approach is diminishing due to lack of detachable batteries in modern devices. In this project, we propose an innovative and novel approach for assessing energy footprint of mobile and wearable systems using thermal imaging. In our approach, an off-the-shelf thermal camera is used to monitor thermal radiation of a device while it is operating an application. We envision a solution to estimate energy footprint of application running in devices that can be applied to any device without instrumenting the device with invasive mechanisms.
Supervisor: Dirk Oliver Theis
Quantum computers with about 1000 logical qubits for fault-tolerant quantum computing could be available within 4 years. While small, these quantum computers might pose a threat to encryption, and they might offer opportunities for computation. Over the past 5-10 years quantum algorithms research has accelerated greatly. New algorithmic techniques have been discovered, and the list of problems for which quantum computers asymptotically outperform classical computers in terms for running time is growing quickly. However, comparatively little effort has been made to answer the question what kind of problems can be solved on a small quantum computer of around 1000 logical qubits. The proposed PhD project will be dedicated to this question.
Supervisor(s): Rajesh Sharma
This doctoral project aims to automatically extract, pre-process, and analyse large volumes of social media data to identify structural and behavioural patterns that could answer questions on socio-economic well-being. In addition, we also plan to explore Call Data Records to understand some of the important topics with respect to Estonian society. The existing studies generally have attempted to answer these questions using either theoretical or empirical approaches, but not both simultaneously. Differently, this project will combine game-theoretical approaches to analyse socio-economic well-being, with empirical insights extracted by social network analysis, natural language processing, and machine learning techniques. In order to scope the project, we will analyse at least five specific socio-economic well-being questions that have been previously approached separately using theoretic and empirical methods (but not both), and we will revisit these questions using a hybrid theoretical-empirical approach. From these experiences, we will derive a generalizable method for socio-economic well-being analysis that combines the rigour of game-theoretic approaches with the wide applicability machine learning approaches and social network analysis.
Supervisor(s): Vesal Vojdani
Sound static analysis can automatically prove properties about computer programs. Unlike testing and heuristic bug-detection tools, sound analyzers can guarantee the absence of vulnerabilities. This is true in theory, but in practice, implementing a sound analyzer that can handle the control flow of real-world software systems is just as prone to error as any other software system. The claim that the program under analysis is free from vulnerabilities is only credible if the analyzer can produce a “witnesses” that testifies to the correctness of its verdict.
The problem of producing meaningful witnesses is the central topic of this thesis. A witness codifies assumptions and hints that allow other verification tools to check that the analysis result is correct. During this thesis project, witness generation will be explored in a number of different settings. We start by producing witnesses for concurrent applications and apply this to interactive security analysis for Android. With that in the bag, we turn to the application of witnesses in an emerging application area, the static analysis of neural networks.
Supervisor(s): Dr Kuldar Taveter
The success of any new software product is ultimately defined by emotional factors, which have been largely ignored by requirements engineering. Emotional aspects can be analysed by the novel approach of motivational modelling. This method allows requirements engineers to elicit and represent for a sociotechnical system emotional requirements related to the goals to be achieved by the system. A motivational model is developed in two stages. The first stage is running a do/be/feel elicitation session, where key stakeholders of the project produce the lists of functional, quality and emotional goals and roles needed to achieve these goals. The second stage is converting the four lists into a single page diagram, representing a hierarchy of functional goals with roles and quality and emotional goals attached to them. We propose to include between the first and second stages of motivational modelling an extra stage where the functional goals to be achieved by the sociotechnical system are mapped to the features of the existing apps of the domain. To understand why the users have liked or disliked these features, emotional reactions of the users to one or another feature are then mined in app stores and social media forums by means of the application to be created in the project, based on emotional goals elicited in the do/be/feel session. Mining for emotional user sentiments will be empowered by machine learning. The results help to understand what is missing or wrong in the existing apps, and ultimately generate requirements for the new app to be designed.
Institute of Physics, W. Ostwaldi tn 1, 50411, Tartu, Estonia. Web: www.fi.ut.ee
Supervisor(s): Taivo Jõgiaas, Aile Tamm, Kaupo Kukli
Proposed doctoral thesis is aimed at mechanical properties of atomic layer deposited single and composite structures. The primary objectives comprise determination of hardness and Young’s modulus, that is, elasticity. In structures, consisting of different materials, the difference in thermal behaviour or elemental composition can generate internal stresses, in addition to possible stresses generated during deposition. Accumulation of stresses could, for example, result in failures of electronic or microelectromechanical devices until mechanical destruction. This will be addressed during the research. The studies are aimed at the systematic combination of structurally and chemically distinctive materials in the form of multi-layered, nanolaminate, structures. The latter can allow one to usefully modify the structural stability and mechanical durability of nanostructured thin films.
Institute of Mathematics and Statistics, J. Liivi 2 - 513, 50409, Tartu, Estonia. Phone: 737 5863, 737 5453, e-mail: ltms [ät] ut.ee, web: math.ut.ee
Supervisor(s): Ago-Erik Riet
The doctoral student will study the mathematical and computational aspects of error correction codes for flash memories. He/she will study the possible types of errors and will design codes to mitigate them. He/she will study questions of encoding and decoding when storing and reading in a flash memory. Among other things he/she will study permutation codes that is a method to mitigate errors in NAND flash memory that happen because of charge leakage and overcharging. He/she will also study codes for inter-cell interference. He/she will construct codes and study their properties theoretically and computationally. He/she will use the computational resources of the University of Tartu, for example the computing cluster and specialized mathematical, algebraic and optimization software. The successful candidate will have good knowledge and interest in coding theory, algebra and combinatorics and he/she knows how to program.
Supervisor(s): R. Haller, M. Põldvere
Lipschitz functions are the most natural non-linear analogue to bounded linear operators between Banach spaces. Every Lipschitz function between metric spaces admits a canonical linear extension between the corresponding Lipschitz-free spaces with its Lipschitz constant being preserved. In this way, one can replace a Lipschitz function by a presumably easier linear one, although by dealing with more complicated spaces. This part of non-linear functional analysis is a comparatively new, active, and expanding field of research with many unsolved internal problems, deep connections to related fields, and with applications in discrete mathematics, optimal transport theory, graph theory, and computer science.
The proposed doctoral project aims at studying big slice phenomena in Lipschitz function spaces and Lipschitz-free spaces. In particular, the following problems are planned to be addressed. Which properties for a metric space M guarantee properties CWO, CWO-S, and CWO-B for the corresponding Lipschitz function space? Do these properties pass to the projective tensor product from its factors? What are the dual octahedrality conditions for these properties?
Supervisor(s): J. Langemets (University of Tartu), A. Lissitsin (University of Tartu)
The geometry of the unit ball plays an important role to understand the structure of a Banach space itself. Since the diameter of the unit ball in a Banach space is always 2, then the maximal diameter of any subset (e.g. a slice) of the unit ball can at most be 2. Probably the most studied geometric property is the Radon–Nikodým property, which assures the existence of slices with arbitrarily small diameter. However, many of the classical Banach spaces enjoy the extremely opposite property that every slice has diameter 2. This project aims to develop further this new world of the big slice phenomenon in the following three directions:
Objective 1. Improve the understanding of almost square Banach spaces
Objective 2. Investigate ball-covering properties versus octahedrality and diameter 2 properties
Objective 3. Describe diameter 2 properties and octahedrality via operators and transfer techniques
Tartu Observatoorium, Observatooriumi 1, Tõravere, tel (+372) 737 4510, info [ät] to.ee
Supervisor(s): Anna Aret (UT Tartu observatory)
Stars are the basic building blocks of the cosmos and the source of life. Understanding their structure and evolution is fundamental for an adequate comprehension of the Universe. While a critical role is played by the most luminous and massive stars, stellar evolution theory is still most uncertain for them. Particular difficulties are caused by late stages of stellar evolution, before death in magnificent supernova explosion. Blue supergiant stars are one of such stages, where missing knowledge on their interior properties and processes triggering mass loss leads to great uncertainties in the theoretical models. The current project aims at providing observational constraints on stellar evolution models via study of pulsational properties of blue supergiant stars and their links to mass ejections. The analysis will be built on the extensive set of observational data resulting from our national facilities, international collaboration network, big international facilities and public archives of ground- and space-based observations. In addition to the existing data, extensive astronomical observations will be carried out during doctoral studies, using both optical spectroscopy and photometry.
Institute of Ecology and Earth Sciences, Vanemuise 46, 51014, Tartu, Estonia. Phone: (+372) 737 5835, e-mail: om [ät] ut.ee, fax: 737 5822, web: www.omi.ut.ee
Supervisors: Toomas Tammaru, Mohamed Ghamizi
Monitoring projects on insects, especially Lepidoptera, are increasingly used as a tool in conservation-oriented research. The aim of the present project is to create the necessary basis for a large-scale application of nocturnal Lepidoptera (moth) surveys in northern Africa. Using bait traps, moths will be sampled in mountainous areas of Morocco to study 1) moth diversity in different vegetation types, 2) the effect of human activities on moth biodiversity, 3) the effect of light pollution on moth assemblages, 4) high altitude species especially vulnerable to climate change. Additionally, host plant use of potentially endangered moth species will be experimentally assessed.
Supervisor(s): Tuul Sepp, Mathieu Giraudeau, Jerome Fort, Orsolya Vincze
Exposure to pollution shortens the life-span of humans and laboratory model animals, but little is known about the effects of pollutants on wild organisms, especially on aging rate and life-history. Animals that are ecologically connected to marine pollution are especially vulnerable to pollution effects, due to biomagnification processes. Aging research has been traditionally conducted on short-lived laboratory organisms, but long-lived organisms differ in their senescence patterns from short-lived ones, and senescence effects on life-history can best be assessed in natural conditions. We suggest a project assessing the effects of pollution on aging and life-history, based on two approaches. First, we will use data collected from wild populations of long-lived seabirds, and assess the effects of pollutants and climate change on changes in life-history parameters and telomere length. Second, we will use the comparative approach, based on the largest databases ever compiled on telomere shortening and cancer prevalence to test if aging rate predicts neoplasia development. The project will be conducted in collaboration of researcher from Estonia and France, who will supervise the recruited PhD student on field work, laboratory analyses, and statistical methods applied in comparative research.