Deep convolutional neural networks for semantic video object segmentation

In this thesis, we propose to infer pixel-level labelling in video by utilising only object category information, exploiting the intrinsic structure of video data. Our motivation is the observation that image-level labels are much more easily to be acquired than pixel-level labels, and it is natural to find a link between the image level recognition and pixel level classification in video data, which would transfer learned recognition models from one domain to the other one. To this end, this thesis proposes two domain adaptation approaches to adapt the deep convolutional neural network (CNN) image recognition model trained from labelled image data to the target domain exploiting both semantic evidence learned from CNN, and the intrinsic structures of unlabelled video data. Our proposed approaches explicitly model and compensate for the domain adaptation from the source domain to the target domain which in turn underpins a robust semantic object segmentation method for natural videos. We demonstrate the superior performance of our methods by presenting extensive evaluations on challenging datasets comparing with the state-of-the-art methods.

Calibration of time domain reflectometry (DTR) for soil moisture measurements in cultivated peat soils

Tiivistelmä: TDR-mittausten kalibrointi viljeltyjen turvemaiden kosteuden mittaamiseen

Plant adaptation to temperature and photoperiod

Selostus: Kasvien sopeutuminen lämpötilaan ja päivän pituuteen

Information System ProcessInnovation Adoption, Adaptation, Learning, and Unlearning: A Longitudinal Case Study

This thesis examines the history and evolution of information system process innovation (ISPI) processes (adoption, adaptation, and unlearning) within the information system development (ISD) work in an internal information system (IS) department and in two IS software house organisations in Finland over a 43-year time-period. The study offers insights into influential actors and their dependencies in deciding over ISPIs. The research usesa qualitative research approach, and the research methodology involves the description of the ISPI processes, how the actors searched for ISPIs, and how the relationships between the actors changed over time. The existing theories were evaluated using the conceptual models of the ISPI processes based on the innovationliterature in the IS area. The main focus of the study was to observe changes in the main ISPI processes over time. The main contribution of the thesis is a new theory. The term theory should be understood as 1) a new conceptual framework of the ISPI processes, 2) new ISPI concepts and categories, and the relationships between the ISPI concepts inside the ISPI processes. The study gives a comprehensive and systematic study on the history and evolution of the ISPI processes; reveals the factors that affected ISPI adoption; studies ISPI knowledge acquisition, information transfer, and adaptation mechanisms; and reveals the mechanismsaffecting ISPI unlearning; changes in the ISPI processes; and diverse actors involved in the processes. The results show that both the internal IS department and the two IS software houses sought opportunities to improve their technical skills and career paths and this created an innovative culture. When new technology generations come to the market the platform systems need to be renewed, and therefore the organisations invest in ISPIs in cycles. The extent of internal learning and experiments was higher than the external knowledge acquisition. Until the outsourcing event (1984) the decision-making was centralised and the internalIS department was very influential over ISPIs. After outsourcing, decision-making became distributed between the two IS software houses, the IS client, and itsinternal IT department. The IS client wanted to assure that information systemswould serve the business of the company and thus wanted to co-operate closely with the software organisations.

Mobile terminal adaptation in portal for better usability

Elektroniset finanssipalvelut, erityisesti Internetin kautta käytettynä, on kasvava alue. Elektronisten finanssipalveluiden tarjoajan tulee pystyä tarjoamaan laaja käytettävyys kaikkien kanavien kautta. Laajan käytettävyyden avulla asiakas voi valita haluamansa kanavan haluamanaan aikana. Palveluntarjoajalla tulee olla joustava arkkitehtuuri pystyäkseen tukemaan asiakkaiden muuttuvia vaatimuksia. Joustavalla arkkitehtuurilla päätelaitteeseen mukautuminen on mahdollista ja näin palveluntarjoaja pystyy tarjoamaan tuen monille eri päätelaitteille ja teknologioille helposti ja nopeasti. Diplomityö keskittyy tutkimaan mahdollisuutta monen kanavan tukeen ja päätelaitteeseen mukautumista Nordean tulevassa finanssiportaaliratkaisussa. Tämä pitäisi olla mahdollista uuden arkkitehtuurin kanssa, jonka TietoEnator on toteuttanut yhteistyössä Nordean kanssa. Sivujen rakenteen uudelleenjärjestelyillä saatiin hyviä tuloksia. Nykyisestä arkkitehtuurissa löydettiin myös puutteita ja jäljelle jäi avoimia kysymyksiä, jotka kirjattiin ylös. On selvästi nähtävissä, että tehokas päätelaitteeseen mukautuminen ja tuki monelle kanavalle tuo hyötyjä sekä pankille että asiakkaalle.

IPSec in 3G packet switched domain

Diplomityö käsittelee IPSec-protokollan (IP Security Protocol) implementointia UMTS:n (Universal Mobile Telecommunications System) pakettikytkentäiseen verkkoon. Runkoverkkoa käytetään mobiilikäyttäjän datan siirtämiseen sekä verkkoelementtien väliseen ohjausinformaation välitykseen. Koska UMTS:n runkoverkot ovat IP-pakettikytkentäisiä verkkoja, IPSec-protokollaa voidaan käyttää lähetettyjen IP-datasähkeiden suojaamiseen. IPSec- ja IKE-protokollien (Internet Key Exchange) käyttö on koettu monimutkaiseksi kiinteissä verkoissa. Tämän saman ongelman edessä tulevat olemaan myös operaattorit, kun he alkavat rakentaa UMTS-verkkojaan. On kuitenkin muistettava se, että tulevaisuudessa lähes kaikki data mukaanlukien ääni ja video on tarkoitus siirtää IP-protokollan avulla. IP-teknologiaan perustuva tiedonsiirron kasvu lisää IPSec-protokollan merkitystä ei ainoastaan runkoverkossa mutta myös radioliityntäverkoissa sekä SS7-merkinantoverkoissa (Signaling System No. 7). Diplomityö on tehty osaksi diplomi-insinöörin tutkintoa Lappeenrannan teknillisessä yliopistossa. Työ on tehty Nokia Networksin palveluksessa Helsingissä, vuosien 2002 ja 2003 välisenä aikana.

A Dependency Parsing Approach to Biomedical Text Mining

Biomedical research is currently facing a new type of challenge: an excess of information, both in terms of raw data from experiments and in the number of scientific publications describing their results. Mirroring the focus on data mining techniques to address the issues of structured data, there has recently been great interest in the development and application of text mining techniques to make more effective use of the knowledge contained in biomedical scientific publications, accessible only in the form of natural human language. This thesis describes research done in the broader scope of projects aiming to develop methods, tools and techniques for text mining tasks in general and for the biomedical domain in particular. The work described here involves more specifically the goal of extracting information from statements concerning relations of biomedical entities, such as protein-protein interactions. The approach taken is one using full parsing—syntactic analysis of the entire structure of sentences—and machine learning, aiming to develop reliable methods that can further be generalized to apply also to other domains. The five papers at the core of this thesis describe research on a number of distinct but related topics in text mining. In the first of these studies, we assessed the applicability of two popular general English parsers to biomedical text mining and, finding their performance limited, identified several specific challenges to accurate parsing of domain text. In a follow-up study focusing on parsing issues related to specialized domain terminology, we evaluated three lexical adaptation methods. We found that the accurate resolution of unknown words can considerably improve parsing performance and introduced a domain-adapted parser that reduced the error rate of theoriginal by 10% while also roughly halving parsing time. To establish the relative merits of parsers that differ in the applied formalisms and the representation given to their syntactic analyses, we have also developed evaluation methodology, considering different approaches to establishing comparable dependency-based evaluation results. We introduced a methodology for creating highly accurate conversions between different parse representations, demonstrating the feasibility of unification of idiverse syntactic schemes under a shared, application-oriented representation. In addition to allowing formalism-neutral evaluation, we argue that such unification can also increase the value of parsers for domain text mining. As a further step in this direction, we analysed the characteristics of publicly available biomedical corpora annotated for protein-protein interactions and created tools for converting them into a shared form, thus contributing also to the unification of text mining resources. The introduced unified corpora allowed us to perform a task-oriented comparative evaluation of biomedical text mining corpora. This evaluation established clear limits on the comparability of results for text mining methods evaluated on different resources, prompting further efforts toward standardization. To support this and other research, we have also designed and annotated BioInfer, the first domain corpus of its size combining annotation of syntax and biomedical entities with a detailed annotation of their relationships. The corpus represents a major design and development effort of the research group, with manual annotation that identifies over 6000 entities, 2500 relationships and 28,000 syntactic dependencies in 1100 sentences. In addition to combining these key annotations for a single set of sentences, BioInfer was also the first domain resource to introduce a representation of entity relations that is supported by ontologies and able to capture complex, structured relationships. Part I of this thesis presents a summary of this research in the broader context of a text mining system, and Part II contains reprints of the five included publications.

Adaptive optimization schemes for mobile VoIP applications Battery life and call quality aspects

In this thesis programmatic, application-layer means for better energy-efficiency in the VoIP application domain are studied. The work presented concentrates on optimizations which are suitable for VoIP-implementations utilizing SIP and IEEE 802.11 technologies. Energy-saving optimizations can have an impact on perceived call quality, and thus energy-saving means are studied together with those factors affecting perceived call quality. In this thesis a general view on a topic is given. Based on theory, adaptive optimization schemes for dynamic controlling of application's operation are proposed. A runtime quality model, capable of being integrated into optimization schemes, is developed for VoIP call quality estimation. Based on proposed optimization schemes, some power consumption measurements are done to find out achievable advantages. Measurement results show that a reduction in power consumption is possible to achieve with the help of adaptive optimization schemes.

Type II aromatic polyketide biosynthetic tailoring enzymes: diversity and adaptation in Streptomyces secondary metabolism.

Members of the bacterial genus Streptomyces are well known for their ability to produce an exceptionally wide selection of diverse secondary metabolites. These include natural bioactive chemical compounds which have potential applications in medicine, agriculture and other fields of commerce. The outstanding biosynthetic capacity derives from the characteristic genetic flexibility of Streptomyces secondary metabolism pathways: i) Clustering of the biosynthetic genes in chromosome regions redundant for vital primary functions, and ii) the presence of numerous genetic elements within these regions which facilitate DNA rearrangement and transfer between non-progeny species. Decades of intensive genetic research on the organization and function of the biosynthetic routes has led to a variety of molecular biology applications, which can be used to expand the diversity of compounds synthesized. These include techniques which, for example, allow modification and artificial construction of novel pathways, and enable gene-level detection of silent secondary metabolite clusters. Over the years the research has expanded to cover molecular-level analysis of the enzymes responsible for the individual catalytic reactions. In vitro studies of the enzymes provide a detailed insight into their catalytic functions, mechanisms, substrate specificities, interactions and stereochemical determinants. These are factors that are essential for the thorough understanding and rational design of novel biosynthetic routes. The current study is a part of a more extensive research project (Antibiotic Biosynthetic Enzymes; www.sci.utu.fi/projects/biokemia/abe), which focuses on the post-PKS tailoring enzymes involved in various type II aromatic polyketide biosynthetic pathways in Streptomyces bacteria. The initiative here was to investigate specific catalytic steps in anthracycline and angucycline biosynthesis through in vitro biochemical enzyme characterization and structural enzymology. The objectives were to elucidate detailed mechanisms and enzyme-level interactions which cannot be resolved by in vivo genetic studies alone. The first part of the experimental work concerns the homologous polyketide cyclases SnoaL and AknH. These catalyze the closure of the last carbon ring of the tetracyclic carbon frame common to all anthracycline-type compounds. The second part of the study primarily deals with tailoring enzymes PgaE (and its homolog CabE) and PgaM, which are responsible for a cascade of sequential modification reactions in angucycline biosynthesis. The results complemented earlier in vivo findings and confirmed the enzyme functions in vitro. Importantly, we were able to identify the amino acid -level determinants that influence AknH and SnoaL stereoselectivity and to determine the complex biosynthetic steps of the angucycline oxygenation cascade of PgaE and PgaM. In addition, the findings revealed interesting cases of enzyme-level adaptation, as some of the catalytic mechanisms did not coincide with those described for characterised homologs or enzymes of known function. Specifically, SnoaL and AknH were shown to employ a novel acid-base mechanism for aldol condenzation, whereas the hydroxylation reaction catalysed by PgaM involved unexpected oxygen chemistry. Owing to a gene-level fusion of two ancestral reading frames, PgaM was also shown to adopt an unusual quaternary sturucture, a non-covalent fusion complex of two alternative forms of the protein. Furthermore, the work highlighted some common themes encountered in polyketide biosynthetic pathways such as enzyme substrate specificity and intermediate reactivity. These are discussed in the final chapters of the work.

Utilization of a Time Domain Simulator in the Technical and Economic Analysis of a Wind Turbine Electric Drive Train

The amount of installed wind power has been growing exponentially during the past ten years. As wind turbines have become a significant source of electrical energy, the interactions between the turbines and the electric power network need to be studied more thoroughly than before. Especially, the behavior of the turbines in fault situations is of prime importance; simply disconnecting all wind turbines from the network during a voltage drop is no longer acceptable, since this would contribute to a total network collapse. These requirements have been a contributor to the increased role of simulations in the study and design of the electric drive train of a wind turbine. When planning a wind power investment, the selection of the site and the turbine are crucial for the economic feasibility of the installation. Economic feasibility, on the other hand, is the factor that determines whether or not investment in wind power will continue, contributing to green electricity production and reduction of emissions. In the selection of the installation site and the turbine (siting and site matching), the properties of the electric drive train of the planned turbine have so far been generally not been taken into account. Additionally, although the loss minimization of some of the individual components of the drive train has been studied, the drive train as a whole has received less attention. Furthermore, as a wind turbine will typically operate at a power level lower than the nominal most of the time, efficiency analysis in the nominal operating point is not sufficient. This doctoral dissertation attempts to combine the two aforementioned areas of interest by studying the applicability of time domain simulations in the analysis of the economicfeasibility of a wind turbine. The utilization of a general-purpose time domain simulator, otherwise applied to the study of network interactions and control systems, in the economic analysis of the wind energy conversion system is studied. The main benefits of the simulation-based method over traditional methods based on analytic calculation of losses include the ability to reuse and recombine existing models, the ability to analyze interactions between the components and subsystems in the electric drive train (something which is impossible when considering different subsystems as independent blocks, as is commonly done in theanalytical calculation of efficiencies), the ability to analyze in a rather straightforward manner the effect of selections other than physical components, for example control algorithms, and the ability to verify assumptions of the effects of a particular design change on the efficiency of the whole system. Based on the work, it can be concluded that differences between two configurations can be seen in the economic performance with only minor modifications to the simulation models used in the network interaction and control method study. This eliminates the need ofdeveloping analytic expressions for losses and enables the study of the system as a whole instead of modeling it as series connection of independent blocks with no lossinterdependencies. Three example cases (site matching, component selection, control principle selection) are provided to illustrate the usage of the approach and analyze its performance.

Genesis and Emplacement of Carbonatites and Lamprophyres in the Svecofennian Domain

A small carbonatite dyke swarm has been identified at Naantali, southwest Finland. Several swarms of shoshonitic lamprophyres are also known along the Archean-Proterozoic boundary in eastern Finland and northwest Russia. These intrusions, along with the carbonatite intrusion at Halpanen, eastern Finland, represent a stage of widespread low-volume mantle-sourced alkaline magmatism in the Svecofennian Domain. Using trace element and isotope geochemistry coupled with precise geochronology from these rocks, a model is presented for the Proterozoic metasomatic evolution of the Fennoscandian subcontinental lithospheric mantle. At ~2.2-2.06 Ga, increased biological production in shallow seas linked to continental rifting, resulted in increased burial rates of organic carbon. Subduction between ~1.93-1.88 Ga returned organic carbon-enriched sediments of mixed Archean and Proterozoic provenance to the mantle. Dehydration reactions supplied water to the mantle wedge, driving arc volcanism, while mica, amphibole and carbonate were brought deeper into the mantle with the subducting slab. The cold subducted slab was heated conductively from the surrounding warm mantle, while pressures continued to gradually increase as a result of crustal thickening. The sediments began to melt in a two stage process, first producing a hydrous alkaline silicate melt, which infiltrated the mantle wedge and crystallised as metasomatic veins. At higher temperatures, carbonatite melt was produced, which preferentially infiltrated the pre-existing metasomatic vein network. At the onset of post-collisional extension, deep fault structures formed, providing conduits for mantle melts to reach the upper crust. Low-volume partial melting of the enriched mantle at depths of at least 110 km led to the formation of first carbonatitic magma and subsequently lamprophyric magma. Carbonatite was emplaced in the upper crust at Naantali at 1795.7 ± 6.8 Ma; lamprophyres along the Archean-Proterozoic boundary were emplaced between 1790.1 ± 3.3 Ma and 1781 ± 20 Ma.