Browsing by Title

Multilingual Latent Dirichlet Allocation (MLDA) is an extension of Latent Dirichlet Allocation (LDA) in a multilingual setting, which aims to discover aligned latent topic structures of a parallel corpus. Although the two popular training algorithms of LDA, collapsed Gibbs sampling and variational inference, can be naturally adopted to MLDA, the two algorithms both become time-inefficient with MLDA due to its special structure. To address this problem, we propose an approximate training framework of MLDA, which works with both collapsed Gibbs sampling and variational inference. Through the experiments, we show that the proposed training framework is able to reduce the training time of MLDA considerably, especially when there are many languages. We also summarize the scenarios where the approximate framework gives comparable model accuracy to that of the standard framework. Finally, we discuss several possible explorations as a future plan.

This thesis studies methods for inferring time series influences on one another. Specifically transfer entropy is motivated, derived and explained through informa- tion theory and Markov chains. Issues and solutions for the calculation of transfer entropy in the context of discrete, coarse grained and continuous time series are covered. The equality of Granger causality and transfer entropy in the Gaussian case is explored. Software package TISEAN for nonlinear time series analysis is explored and its capabilities are explained. In this thesis nonlinear methods are applied to measurements from SMEAR II station which is located in Hyytiälä Forestry Field Station in Hyytiälä, Finland. The problems caused by nonstationarity of the data are discussed and a solution to these problems is attempted. The ecosystem exchange is found to be nonstationary and containing two dif- ferent strong periodic components implying that the underlying dynamical system changes its characteristics. This prevented obtaining a phase space embedding from the measured data within the scope of this thesis. Further work is needed to find out whether transfer entropy can give new in- sights into analyzing the ecosystem and its influences. Rudimentary ground work has been laid with this thesis and possible ways of mitigating the nonstationarity in the future are given. All the code for reproducing the analysis and plots are included in the appendix to help future work.

Credit rating is one of the core tools for risk management within financial firms. Ratings are usually provided by specialized agencies which perform an overall study and diagnosis on a given firm’s financial health. Dealing with unrated entities is a common problem, as several risk models rely on the ratings’ completeness, and agencies can not realistically rate every existing company. To solve this, credit rating prediction has been widely studied in academia. However, research in this topic tends to separate models amongst the different rating agencies due to the difference in both rating scales and composition. This work uses transfer learning, via label adaptation, to increase the number of samples for feature selection, and appends these adapted labels as an additional feature to improve the predictive power and stability of previously proposed methods. Accuracy on exact label prediction was improved from 0.30, in traditional models, up to 0.33 in the transfer learning setting. Furthermore, when measuring accuracy with a tolerance of 3 grade notches, accuracy increased almost 0.10, from 0.87 to 0.96. Overall, transfer learning displayed better out-of-sample generalization.

In bioinformatics, new genomes are sequenced at an increasing rate. To utilize this data in various bioinformatics problems, it must be annotated first. Genome annotation is a computational problem that has traditionally been approached by using statistical methods such as the Hidden Markov model (HMM). However, implementing these methods is often time-consuming and requires domain knowledge. Neural network-based approaches have also been developed for the task, but they typically require a large amount of pre-labeled data. Genomes and natural language share many properties, not least the fact that they both consist of letters. Genomes also have their own grammar, semantics, and context-based meanings, just like phrases in the natural language. These similarities give motivation to the use of Natural language processing (NLP) techniques in genome annotation. In recent years, pre-trained Transformer neural networks have been widely used in NLP. This thesis shows that due to the linguistic properties of genomic data, Transformer network architecture is also suitable for gene predicting. The model used in the experiments, DNABERT, is pre-trained using the full human genome. Using task-specific labeled data sets, the model is then trained to classify DNA sequences into genes and non-genes. The main fine-tuning dataset is the genome of the Escherichia coli bacterium, but preliminary experiments are also performed on human chromosome data. The fine-tuned models are evaluated for accuracy, F1-score and Matthews correlation coefficient (MCC). A customized estimation method is developed, in which the predictions are compared to ground-truth labels at the nucleotide level. Based on that, the best models achieve a 90.15% accuracy and an MCC value of 0.4683 using the Escherichia coli dataset. The model correctly classifies even the minority label, and the execution times are measured in minutes rather than hours. These suggest that the NLP-based Transformer network is a powerful tool for learning the characteristics of gene and non-gene sequences.

Several extensions of first-order logic are studied in descriptive complexity theory. These extensions include transitive closure logic and deterministic transitive closure logic, which extend first-order logic with transitive closure operators. It is known that deterministic transitive closure logic captures the complexity class of the languages that are decidable by some deterministic Turing machine using a logarithmic amount of memory space. An analogous result holds for transitive closure logic and nondeterministic Turing machines. This thesis concerns the k-ary fragments of these two logics. In each k-ary fragment, the arities of transitive closure operators appearing in formulas are restricted to a nonzero natural number k. The expressivity of these fragments can be studied in terms of multihead finite automata. The type of automaton that we consider in this thesis is a two-way multihead automaton with nested pebbles. We look at the expressive power of multihead automata and the k-ary fragments of transitive closure logics in the class of finite structures called word models. We show that deterministic twoway k-head automata with nested pebbles have the same expressive power as first-order logic with k-ary deterministic transitive closure. For a corresponding result in the case of nondeterministic automata, we restrict to the positive fragment of k-ary transitive closure logic. The two theorems and their proofs are based on the article ’Automata with nested pebbles capture first-order logic with transitive closure’ by Joost Engelfriet and Hendrik Jan Hoogeboom. In the article, the results are proved in the case of trees. Since word models can be viewed as a special type of trees, the theorems considered in this thesis are a special case of a more general result.

Kompleksiluvut jakautuvat algebrallisiin lukuihin ja transkendenttilukuihin. Algebralliset luvut ovat näistä lukujoukoista tutumpi, mutta itse asiassa lähes kaikki kompleksiluvut ovat transkendenttilukuja. Transkendenttilukujen löytäminen ja niiden todistaminen transkendenttisiksi on hyvin vaikeaa, mistä syystä transkendenttilukuja tunnetaan vain harvoja erikoistapauksia, tunnetuimpina pii ja Neperin luku. Transkendenttilukuihin liittyvistä teknisistä vaikeuksista johtuen myös transkendenttilukujen historia on suhteellisen lyhyt. Tämän tutkielman tarkoituksena on antaa lyhyt yleiskuva aiheesta. Ensin tutkielmassa esitellään lyhyesti lukukäsitteen ja transkendenttilukujen historiaa. Tämän jälkeen määritellään ensin transkendenttiluvut, esitetään Liouvillen ja Cantorin olemassaolotodistukset transkendenttiluvuille sekä ensimmäinen esimerkki transkendenttiluvusta, Liouvillen luku. Tämän jälkeen esitetään todistus Lindemannin-Weierstrassin lauseelle, joiden avulla on ollut mahdollista todistaa useita lukuja transkendenttisiksi. Näiden lukujen joukossa ovat muiden muassa pii ja Neperin luku, joiden transkendenttisuustodistukset esitetään lauseen korollaareina. Lopuksi esitellään ilman todistusta Gelfondin-Schneiderin lause sekä esitellään hyvin lyhyesti joitakin transkendenttilukujen teoriaa eteenpäin vieneitä tuloksia sekä transkendenttilukuihin liittyviä avoimia kysymyksiä.

People are internationally increasingly mobile which is why moving should be rather seen as a lifestyle than something exceptional and exotic as tourism research has traditionally viewed traveling. The new mobility paradigm in the tourism field in fact suggests that tourism should be seen as a part of everyday life. When people move more they are also in contact with more places. The new transnational research, that has predominantly been applied in migration studies, is interested in how mobile people form relationships with the different places that they have spent time in, in other words how they form a sense of place. Traditional migration studies has rather focused on how migrants settle in the new place, whereas the new transnational research is interested in how people can feel a belonging to many places at the same time, thus forming a so called multiple sense of self and perhaps feel at home in many places. Exchange students represent one very mobile group of people. The exchange students desire an international experience whereas nations, cities and educational institutions see the students as an important asset in their aims of internationalization. Branding Helsinki an attractive student city was the topic of a work shop meeting in Autumn 2012 held between The University of Helsinki and the City of Helsinki. The significance of exchange-students for this brand was also highlighted. In this study exchange students are understood as educational tourists, and they were studied from the perspective of the new mobility paradigm as well as the new transnational research. The perspective is fresh because the general experiences of exchange students have not been studied a lot, let alone their sense of place. The aim of this study was to examine the exchange students at The University of Helsinki in the semester of 2012-2013 if and what kind of sense of place they formed. Interesting was to examine whether they formed a strong sense of place because according to transnational theories acquiring a multiple sense of place is possible for mobile groups of people. The sense of place was defined through the concepts of emotional and functional attachment to place. The concepts were examined by analyzing how the students settled in Helsinki, if they felt at home, if they felt local and how they felt that the exchange affected their future mobility. Another aim of the study was to examine which specific aspects of Helsinki affected the sense of place of the students and to suggest how these findings could be of help in suggesting how to brand Helsinki an internationally attractive student city. As a method 25 semi structured interviews with exchange students between January and March of 2013 were performed. The method was suitable for this fresh topic because it allowed for the voice of the students to be heard. The interviews were transcribed and analyzed using content analysis. The results show that the students were emotionally and functionally very attached to Helsinki. They settled in well, felt at home and identified many positive tangible and intangible aspects of the city. This argues for the fact that they had acquired a strong sense of place. The students experienced the city in the company of other people, but the individual experiences were important too. The students did however not feel like locals in the city. The study nevertheless proved that in order to acquire a strong sense of place, one does not have to feel local. The background theories demonstrate that also temporary residents and visitors can acquire a strong sense of place. The findings imply that acquiring a multiple sense of place (in their home country and Helsinki) is indeed possible. The fact that the students felt that they were living a mobile lifestyle now and in the future also speaks for a multiple sense of place. The results show that the exchange students appreciate the friendly Finns, the climate, the greenness of the city and the working society in Helsinki. Therefore this study suggests that these aspects should be used to brand Helsinki an internationally attractive student city in order to accomplish the internationalization aims of both the City of Helsinki and The University of Helsinki.

The Internet Of Things (IoT) has been a buzzword for the past decade. While IoT focuses on getting devices connected to the global Internet and other existing networks with new novel communication medias and protocols Web Of Things (WoT), a sidekick of the IoT, is recycling the current web development patterns and tools to integrate everyday objects to the existing World Wide Web. In WoT the idea is to realize the IoT concept using the current web standards and practices. In order to do so WoT uses the REST architectural style implemented on top of the HTTP already in use all over the Internet. The WoT concept can be taken a step further to allow the communications over other protocol stacks than the traditional TCP. By implementing the HTTP client and server components so that they don't directly operate the underlying transport protocol but instead use an abstraction layer between, we are able to switch the whole protocol stack without making any changes to the actual software logic that is built on top of the components. A device designed to be controlled over Ethernet can be for example be managed over any number of compliant media such as Bluetooth, Wi-Fi, and serial port communications.

With the ongoing global biodiversity loss, approaches to measuring and monitoring biodiversity are necessary for effective conservation planning, especially in tropical forests. Remote sensing is a very potential tool for biodiversity mapping, and high spatial resolution imaging spectroscopy allows for direct estimation of tree species diversity based on spectral reflectance. The objective of this study is to test an approach for estimating tree species alpha diversity in a tropical montane forest in the Taita Hills, Kenya. Tree species diversity is estimated based on spectral variation of high spatial resolution imaging spectroscopy data. The approach is an unsupervised classification, or clustering, applied to objects that represent tree crowns. Airborne imaging spectroscopy data and species data from 31 field plots were collected from the study area. After preprocessing of the spectroscopic imagery, a minimum noise fraction (MNF) transformation with a subsequent selection of 13 bands was applied to the data to reduce its noise and dimensionality. The imagery was then segmented to obtain objects that represent tree crowns. A clustering algorithm was applied to the segments, with the aim of grouping spectrally similar tree crowns. Experiments were made to find the optimal range for the number of clusters. Tree species richness and two diversity indices were calculated from the field data and from the clustering results. The clusters were assumed to represent species in the calculations. Correlation analysis and linear regression analysis were used to study the relationship between diversity measures from the field data and from the clustering results. It was found that the approach succeeded well in revealing tree species diversity patterns with all three diversity measures. Despite some factors that added error to the relationship between field-derived and clustering-derived diversity measures, high correlations were observed. Especially tree species richness could be modelled well using the approach (standard error: 3 species). The size of the considered trees was found to be an important determinant of the relationships. Finally, a tree species richness map was created for the study area. With further development, the presented approach has potential for other interesting applications, such as estimation of beta diversity, and tree species identification by linking the reflectance properties of individual crowns to their corresponding species.

Triglyseridimatriisit sisältävät käsitteenä kaikki lipidimatriisit, kuten eläinrasvat ja kasviöljyt. Triglyseridien lisäksi nämä lipidimatriisit voivat sisältää myös muita heteroatomeja sisältäviä yhdisteitä kuten esimerkiksi typpi- tai happiyhdisteitä. Rasvoja ja öljyjä voidaan käyttää elintarviketeollisuuden lisäksi myös polttoaineteollisuudessa esimerkiksi rasvahappojen metyyliestereistä koostuvan biodieselin tai hiilivedyistä koostuvan uusiutuvan dieselin tuotantoon. Tässä kirjallisuuskatsauksessa esitellään erilaisia uusiutuvien polttoaineiden valmistukseen yleisesti käytettyjä triglyseridimatriiseja sekä niiden sisältämiä yhdisteluokkia, ja perehdytään näiden analysointiin käytettäviin kvalitatiivisiin ja kvantitatiivisiin eristys- ja analyysimenetelmiin. Työssä esitellään useita eläinrasva- ja kasviöljymatriiseja sekä niissä yleisimmin esiintyviä heteroatomeja sisältäviä yhdisteryhmiä. Yhdisteiden eristykseen käytettävistä menetelmistä esitellään neste-nesteuutto, kiinteäfaasiuutto ja saippuointi, ja vastaavasti analyysimenetelmistä käsitellään nestekromatografia (LC), kaasukromatografia (GC) ja massaspektrometria (MS). Triglyseridimatriisien sisältämien heteroatomeja sisältävien epäpuhtauksien ja niiden pitoisuuksien analysointi on tärkeää, sillä näillä epäpuhtauksilla on ympäristöpäästöjen muodostumisen lisäksi useita negatiivia vaikutuksia esimerkiksi polttoaineen tuotantoon, säilyvyyteen ja toimintakykyyn polttomoottoreissa. Tutkielman kokeellisessa osuudessa validoitiin "Rasvahappoamidien, sterolien ja muiden raskaiden yhdisteiden analysointi öljyistä ja rasvoista kaasukromatografisesti" -menetelmä amideille. Validointi koostui kuudesta osuudesta: lineaarisuudesta, määritysrajan ja toteamisrajan määrityksestä, toistettavuudesta, uusittavuudesta, vaihtelevuudesta sekä takaisinsaantokokeesta. Validointi onnistui pääpiirteittäin hyvin, tulokset olivat luotettavia ja vastasivat odotuksia. Validoidun menetelmän rinnalle kehitettiin myös uuttoon perustuvaa menetelmää saippuoitumattomalle osuudelle rasvoista. Menetelmää testattiin eri uuttoliuottimilla ja menetelmän toistettavuutta kokeiltiin. Menetelmällä saatiin näytteistä eristettyä validoituun menetelmään verrattuna huomattavan paljon enemmän erilaisia yhdisteitä. Menetelmän kehitys jäi kuitenkin kesken, sillä esimerkiksi reaktio-olosuhteiden optimointi ja saantojen kvantitointiin vaadittavan malliyhdisteen löytäminen ovat vielä työn alla.