Browsing by master's degree program "Tietojenkäsittelytieteen maisteriohjelma"

The importance of software architecture design decisions has been known for almost 20 years. Knowledge vaporisation is a problem in many projects, especially in the current fast-paced culture, where developers often switch from project to another. Documenting software architecture design decisions helps developers understand the software better and make informed decisions in the future. However, documenting architecture design decisions is highly undervalued. It does not create any revenue in itself, and it is often the disliked and therefore neglected part of the job. This literature review explores what methods, tools and practices are being suggested in the scientific literature, as well as, what practitioners are recommending within the grey literature. What makes these methods good or bad is also investigated. The review covers the past five years and 36 analysed papers. The evidence gathered shows that most of the scientific literature concentrates on developing tools to aid the documentation process. Twelve out of nineteen grey literature papers concentrate on Architecture Decision Records (ADR). ADRs are small template files, which as a collection describe the architecture of the entire system. The ADRs appear to be what practitioners have become used to using over the past decade, as they were first introduced in 2011. What is seen as beneficial in a method or tool is low-cost and low-effort, while producing concise, good quality content. What is seen as a drawback is high-cost, high-effort and producing too much or badly organised content. The suitability of a method or tool depends on the project itself and its requirements.

Parlamentaaristen aineistojen digitointi ja rakenteistaminen tutkimuskäyttöön on nouseva tutkimuksenala, jonka tiimoilta esimerkiksi Euroopassa on tällä hetkellä käynnissä useita kansallisia hankkeita. Tämä tutkielma on osa Semanttinen parlamentti -hanketta, jossa Suomen eduskunnan täysistuntojen puheenvuorot saatetaan ensimmäistä kertaa yhtenäiseksi, harmonisoiduksi aineistoksi koneluettavaan muotoon aina eduskunnan alusta vuodesta 1907 nykypäivään. Puheenvuorot ja niihin liittyvät runsaat kuvailutiedot on julkaistu kahtena versiona, parlamentaaristen aineistojen kuvaamiseen käytetyssä Parla-CLARIN XML -formaatissa sekä linkitetyn avoimen datan tietämysverkkona, joka kytkee aineiston osaksi laajempaa kansallista tietoinfrastruktuuria. Yhtenäinen puheenvuoroaineisto tarjoaa ennennäkemättömiä mahdollisuuksia tarkastella suomalaista parlamentarismia yli sadan vuoden ajalta monisyisesti ja automatisoidusti. Aineisto sisältää lähes miljoona erillistä puheenvuoroa ja linkittyy tiiviisti eduskunnan toimijoiden biografisiin tietoihin. Tässä tutkielmassa kuvataan puheenvuorojen esittämistä varten kehitetyt tietomallit ja puheenvuoroaineistojen keräys- ja muunnosprosessi sekä tarkastellaan prosessin ja syntyneen aineiston haasteita ja mahdollisuuksia. Toteutetun aineistojulkaisun hyödyllisyyden arvioimiseksi on Parla-CLARIN-muotoista aineistoa jo hyödynnetty poliittiseen kulttuuriin liittyvässä digitaalisten ihmistieteiden tutkimuksessa. Linkitetyn datan pohjalta on kehitetty semanttinen portaali, Parlamenttisampo, aineistojen julkaisemista ja tutkimista varten verkossa.

Ohjelmistotuotannon ammattilaiset tarvitsevat monenlaisia kompetensseja. Yksi näistä kompetensseista on kyky elinikäiseen oppimiseen, joka on tarpeellinen laajalla ja jatkuvasti muutoksessa olevalla alalla. ICT-aloille muodostuneen osaajatarpeen myötä elinikäisen oppimisen rooli onkin alkanut korostumaan entisestään. Tutkielman tarkoituksena on ollut lisätä ymmärrystä elinikäisen oppimisen roolista ohjelmistotuotannon ammattilaisen näkökulmasta. Tutkielmassa on pyritty tunnistamaan, mitä oppimisen muotoja hyödynnetään sekä millaisiin tarkoituksiin niitä käytetään, mitkä elinikäisen oppimisen kompetenssin osatekijät ovat tärkeitä sekä mitä haasteita elinikäiseen oppimiseen liittyy. Tutkimuksen aineisto on kerätty puolistrukturoiduilla haastatteluilla ohjelmistotuotannon ammattilaisten kanssa. Näiden haastattelujen tuloksia on verrattu tutkielmaa varten suoritetun kirjallisuuskatsauksen tuloksiin. Oppimisen muodoista informaalia oppimista hyödynnetään eniten ja erityisesti pienempiin oppimistarpeisiin. Nonformaalia ja formaalia oppimista taas hyödynnetään isompiin tarpeisiin, mutta harvemmin. Motivaatio, tiedonhaku ja metaoppiminen korostuvat keskeisinä elinikäisen oppimisen kompetenssin osatekijöinä. Ajanpuute ja itsensä motivoiminen mielletään yleisimmiksi haasteiksi elinikäistä oppimista koskien. Myös tiedonlähteisiin liittyvät puutteet sekä puutteellinen ymmärrys metaoppimisesta mielletään vaikeuttavan elinikäistä oppimista. Tutkielman havainnot tukevat elinikäisen oppimisen kompetenssin keskeistä roolia ohjelmistotuotannon ammattilaisilla. Kehitettävää löytyy kuitenkin vielä ohjelmistotuotannon ammattilaisten valmiuksista elinikäiseen oppimiseen, esimerkiksi metaoppimista koskien. Havainnot perustuvat kuitenkin lyhyemmän aikaa ohjelmistotuotannon ammattilaisina työskennelleiden kokemuksiin, joten lisää tutkimusta tarvitaan etenkin pitempään työskennelleiltä ohjelmistotuotannon ammattilaisilta.

The maximum constraint satisfaction problem (MaxCSP) is a combinatorial optimization problem in which the set of feasible solutions is expressed using decision variables and constraints on how the variables can be assigned. It can be used to represent a wide range of other combinatorial optimization problems. The maximum satisfiability problem (MaxSAT) is a restricted variant of the maximum constraint satisfaction problem with the additional restrictions that all variables must be Boolean variables, and all constraints must be logical Boolean formulas. Because of this, expressing problems using MaxSAT can be unintuitive. The known solving methods for the MaxSAT problem are more efficient than the known solving methods for MaxCSP. Therefore, it is desirable to express problems using MaxSAT. However, every MaxCSP instance that only has finite-domain variables can be encoded into an equivalent MaxSAT instance. Encoding a MaxCSP instance to a MaxSAT instance allows users to combine the strengths of both approaches by expressing problems using the more intuitive MaxCSPs but solving them using the more efficient MaxSAT solving methods. In this thesis, we overview three common MaxCSP to MaxSAT encodings, the sparse, log, and order encodings, that differ in how they encode an integer variable into a set of Boolean variables. We use correlation clustering as a practical example for comparing the encodings. We first represent correlation clustering problems using MaxCSPs, and then encode them into MaxSATs instances. State-of-the-art MaxSAT solvers are then used to solve the MaxSAT instances. We compare the encodings by measuring the time it takes to encode a MaxCSP instance into a MaxSAT instance and the time it takes to solve the MaxSAT instance. The scope of our experiments is too small to draw general conclusions but in our experiments, the log encoding was the best overall choice.

In independent component analysis the data is decomposed into its statistically independent components. In recent years, statistical models have been developed that solve a non-linear version of the independent component analysis. This thesis focuses on the estimation methods of a particular non-linear independent component analysis model called iVAE. It is shown on simulated data that the generative adversarial networks can significantly improve the iVAE model estimation compared with the previously used default iVAE estimation method. The improved model estimation might enable new applications for the iVAE model.

This thesis conducts a systematic literature review on ethical issues of large language models (LLM). These models are a very prudent topic, as both their presence and demand have skyrocketed since the release of ChatGPT - a free to use generative language model. The literature review of 116 studies, both conceptual and empirical, identifies 39 recurring ethical issues. The issues range from methodological to fundamental ones, for example Environmental impacts" and "Biased training data or outputs". These identified issues are analyzed based on the Ethics guidelines for trustworthy AI (Artificial Intelligence), released by the European Commission’s High-Level Expert Group on AI. The guidelines detail requirements that all trustworthy and ethical AI applications should adhere to, e.g., Human agency, Transparency, Accountability. All identified issues are mapped to these requirements, and the conclusion is that LLMs have significant challenges relating to each one. The findings indicate that the use LLMs comes with significant issues, both demonstrated and theorized. While some methods for mitigating these issues are identified, many still remain unanswered. One of these unanswered issues is the most identified one - inherent biases in LLMs. Since there is no universal understanding on biases, there is no way to make LLMs seem unbiased to everyone. This thesis collates the current talking points and issues identified with LLMs. It provides a comprehensive, but not exhaustive, list of these issues and shows that there is much discussion on the topic. The conclusion is that more discussion is required, but more vitally, even more (regulatory) action is needed along with it.

As everyday life becomes digital, more and more daily things are done online. In particular, the increased use of mobile devices has accelerated this development. People are increasingly leaving information online about themselves that can be used to identify a person. On 25 May 2018, the European Union’s General Data Protection Regulation, the GDPR, was repealed in the European Union, repealing the previous European Union Data Protection Directive. The GDPR sets out how personal information should be stored and who can process it. The thesis examined how the introduction of GDPR has affected the customer data storage solutions and IT processes of Finnish SMEs during 2018-2020. The companies were examined in three phases: before, during and after the introduction of the GDPR. The study looked at the number of data breaches in the EU and the penalties imposed for them, and compared the situation in Finland. In addition, Finnish SMEs were interviewed for the dissertation. The interview was conducted as a questionnaire interview with 15 companies. The thesis found that Finland did not stand out in any way among other EU countries in GDPR violations. The answers received as a result of the survey revealed that there has been a clear variation in the interpretation of the content of the GDPR in Finland, which has affected the measures taken by companies. Based on the survey, the measures have also been influenced by the organization and organizational culture. However, the reliability of the results is affected by the small sample size.

Air pollution is considered to be one of the biggest environmental risks to health, causing symptoms from headache to lung diseases, cardiovascular diseases and cancer. To improve awareness of pollutants, air quality needs to be measured more densely. Low-cost air quality sensors offer one solution to increase the number of air quality monitors. However, they suffer from low accuracy of measurements compared to professional-grade monitoring stations. This thesis applies machine learning techniques to calibrate the values of a low-cost air quality sensor against a reference monitoring station. The calibrated values are then compared to a reference station’s values to compute error after calibration. In the past, the evaluation phase has been carried out very lightly. A novel method of selecting data is presented in this thesis to ensure diverse conditions in training and evaluation data, that would yield a more realistic impression about the capabilities of a calibration model. To better understand the level of performance, selected calibration models were trained with data corresponding to different levels of air pollution and meteorological conditions. Regarding pollution level, using homogeneous training and evaluation data, the error of a calibration model was found to be even 85% lower than when using diverse training and evaluation pollution environment. Also, using diverse meteorological training data instead of more homogeneous data was shown to reduce the size of the error and provide stability on the behavior of calibration models.

Simultaneous Localization and Mapping (SLAM) research is gaining a lot of traction as the available computational power and the demand for autonomous vehicles increases. A SLAM system solves the problem of localizing itself during movement (Visual Odometry) and, at the same time, creating a 3D map of its surroundings. Both tasks can be solved on the basis of expensive and spacious hardware like LiDaRs and IMUs, but in this subarea of visual SLAM research aims at replacing those costly sensors by, ultimately, inexpensive monocular cameras. In this work I applied the current state-of-the-art in end-to-end deep learning-based SLAM to a novel dataset comprising of images recorded from cameras mounted to an indoor crane from the Konecranes CXT family. One major aspect that is unique about our proposed dataset is the camera angle that resembles a classical bird’s-eye view towards the ground. This orientation change coming alongside with a novel scene structure has a large impact on the subtask of mapping the environment, which is in this work done through monocular depth prediction. Furthermore, I will assess which properties of the given industrial environments have the biggest impact on the system’s performance to identify possible future research opportunities for improvement. The main performance impairments I examined, that are characteristic for most types of industrial premise, are non-lambertian surfaces, occlusion and texture-sparse areas alongside the ground and walls

Software testing is an important process when ensuring a program's quality. However, testing has not traditionally been a very substantial part of computer science education. Some attempts to integrate it into the curriculum has been made but best practices still prove to be an open question. This thesis discusses multiple attempts of teaching software testing during the years. It also introduces CrowdSorcerer, a system for gathering programming assignments with tests from students. It has been used in introductory programming courses in University of Helsinki. To study if the students benefit from creating assignments with CrowdSorcerer, we analysed the number of assignments and tests they created and if they correlate with their performance in a testing-related question in the course exam. We also gathered feedback from the students on their experiences from using CrowdSorcerer. Looking at the results, it seems that more research on how to teach testing would be beneficial. Improving CrowdSorcerer would also be a good idea.