Kernel-Based Ranking. Methods for Learning and Performance Estimation

Machine learning provides tools for automated construction of predictive models in data intensive areas of engineering and science. The family of regularized kernel methods have in the recent years become one of the mainstream approaches to machine learning, due to a number of advantages the methods share. The approach provides theoretically well-founded solutions to the problems of under- and overfitting, allows learning from structured data, and has been empirically demonstrated to yield high predictive performance on a wide range of application domains. Historically, the problems of classification and regression have gained the majority of attention in the field. In this thesis we focus on another type of learning problem, that of learning to rank. In learning to rank, the aim is from a set of past observations to learn a ranking function that can order new objects according to how well they match some underlying criterion of goodness. As an important special case of the setting, we can recover the bipartite ranking problem, corresponding to maximizing the area under the ROC curve (AUC) in binary classification. Ranking applications appear in a large variety of settings, examples encountered in this thesis include document retrieval in web search, recommender systems, information extraction and automated parsing of natural language. We consider the pairwise approach to learning to rank, where ranking models are learned by minimizing the expected probability of ranking any two randomly drawn test examples incorrectly. The development of computationally efficient kernel methods, based on this approach, has in the past proven to be challenging. Moreover, it is not clear what techniques for estimating the predictive performance of learned models are the most reliable in the ranking setting, and how the techniques can be implemented efficiently. The contributions of this thesis are as follows. First, we develop RankRLS, a computationally efficient kernel method for learning to rank, that is based on minimizing a regularized pairwise least-squares loss. In addition to training methods, we introduce a variety of algorithms for tasks such as model selection, multi-output learning, and cross-validation, based on computational shortcuts from matrix algebra. Second, we improve the fastest known training method for the linear version of the RankSVM algorithm, which is one of the most well established methods for learning to rank. Third, we study the combination of the empirical kernel map and reduced set approximation, which allows the large-scale training of kernel machines using linear solvers, and propose computationally efficient solutions to cross-validation when using the approach. Next, we explore the problem of reliable cross-validation when using AUC as a performance criterion, through an extensive simulation study. We demonstrate that the proposed leave-pair-out cross-validation approach leads to more reliable performance estimation than commonly used alternative approaches. Finally, we present a case study on applying machine learning to information extraction from biomedical literature, which combines several of the approaches considered in the thesis. The thesis is divided into two parts. Part I provides the background for the research work and summarizes the most central results, Part II consists of the five original research articles that are the main contribution of this thesis.

Osaaminen tietämyksen tukena kehitettäessä aikuisopiston työprosesseja

Osaaminen voi muodostua ongelmaksi yritysten kilpailukyvylle, jos siihen ei kiinnitetä huomiota jo strategiasuunnittelusta lähtien. Vaikka asiakkaiden muuttuneet odotukset kyettäisiinkin kohdentamaan ennen kilpailijoita, saattaa olla, että siihen ei pystytä vastaamaan, jos ei ehditä oppimaan uutta tai uudella tavalla. Diplomityön tavoitteena on laatia Etelä-Karjalan aikuisopistollehenkilöstön osaamisen kehittämiskuvaus siitä, kuinka strategian määrittämisestälähtien voidaan henkilöstön osaamista parantaa ja pyrkiä luomaan kilpailuetua markkinoilla. Henkilöstön osaamisen kehittäminen tulee olla suunnitelmallista, tarvittaessa yksilön, ryhmän ja organisaation edut huomioivaa, riittävän yksinkertaista ja konkreettista, jotta suunnitelma voidaan toteuttaa, seurata ja edelleenkehittää. Työn teoriaosassa on kuvattu vision ja strategian merkitystä osaamisen kehittämiseen. Lisäksi on tarkasteltu yksilön oppimista, oppimisen prosessia ja sen kehittymistä organisaation kyvykkyydeksi. Osaamisen infrastruktuuria on lähestytty organisaatiokulttuurin, sitouttamisen ja kehittämisjärjestelmän näkökulmasta. Empiirisessä osuudessa on tuotu esiin aikuisopiston henkilöstön osaamisen kehittämisen tavoitteet, nykyiset käytännöt, kehittämisen vaihtoehdot sekä jatkotoimenpiteet. Osaaminen on aikuisopiston henkilöstön ammattitaidon perusta. Osaamistarpeen määrittämisen tulee keskittyä aikuisopiston ydinosaamisen kehittämiseen. Henkilöstö tulisi nähdä motivoinnin ja sitouttamisen kautta inhimillistä tietopääomaa kasvattavana tekijänä, johon voidaan sujuvasti liittää aineeton pääoma (data, informaatio jne.) sekä strateginen reservi, kuten kilpailuetua tuottava innovointitoiminta.

Kosketukseton hitsin pintageometrian määrittäminen

Yleisesti tiedetään hitsin pintageometrian vaikuttavan rakenteen väsymislujuuteen. Nopean, edullisen ja luotettavan pintageometrian mittausmenetelmän kehittäminen on askel kohti tarkempaa ja varmempaa rakenteen väsymislujuuden tarkastelua. Tässä työssä on tutkittu hitsejä, joiden pinnan geometria on mitattu norjalaisen SINTEF -yrityksen kehittämällä rakenteellisen valon menetelmällä. Osana työtä kehitettiin MatLab -pohjainen ohjelma, jolla jälkikäsitellään mittauksesta saadut x-y-z -mittapisteet. Mittausdatan jälkikäsittelyssä saadaan mittauksesta määritettyähitsin reunan pyöristys, liittymäkulma, a-mitta, reunahaava ja kateettisuhde. Kehitettyä menetelmää käyttämällä mitattiin lähes 300 voimaakantamatontaristiliitoksen hitsiä. Mittaustuloksia verrattiin vastaavista kappaleista tehtyihin hiemittauksiin. Manuaalisen hieestä tehdyn mittauksen havaittiin olevan tarkempi ja pystyttiin havaitsemaan paikallisempia muotoja. Rakenteellisen valon mittauksissa tapahtunut heijastelu saatiin pienenemään käsittelemällä mitattava pinta mattavalkoisella maalilla. Rakenteellisen valon mittatarkkuudeksi saatiin noin 0,2 mm. Pohjautuen mitattuun hitsin reunan pyöristykseen ja liittymäkulmaan voidaan yksinkertaista kaavaa käyttämällä laskea hitsin jännityskonsentraatio ja näin saada alkuarvaus väsymislujuudelle. Myös muiden tekijöiden tiedetään vaikuttavan hitsin väsymislujuuteen, joten pyöristyksen ja liittymäkulman avulla tehdyt arviot eivät ole absoluuttisen oikeita. Tämä havaittiin väsytyskokeilla, joista yhdessä väsymisvaurio ei syntynyt suurimmankaan jännityskonsentraation alueella.

Analyysi Suomen työeläkevakuutusyhtiöiden sijoitustoiminnasta ja taloudellisesta kehityksestä

Tutkielman tavoitteena on analysoida neljän työeläkeyhtiön, Ilmarisen, Varman, Eläke-Tapiolan ja Eläke-Fennian, viime vuosien taloudellista kehitystä. Tutkimusongelmana on selvittää työeläkevakuutusyhtiöiden tämän hetkinen taloudellinen tilanne sekä sijoitustoiminnan menestys. Tutkielma on toteutettu nomoteettisella tutkimusotteella. Tutkielman empiirisessä osassa tarkastellaan yhtiöiden taloudellista kehitystä yhtiöiden tilinpäätösinformaation perusteella. Tutkielman teoriaosa luo pohjan empiirisen osan tutkimukselle. Suomen eläkejärjestelmä kohtaa parhaillaan suuria muutospaineita. Tulevaisuudessa työeläkevakuutusyhtiöillä on entistä suurempi vaikutus koko Suomen talouteen. Tutkimuksen tulokset osoittavat, että heikoimmin menestynyttä Eläke-Fenniaa lukuun ottamatta yhtiöitä ei voi suoraan asettaa paremmuusjärjestykseen vaan sijoittajan tulee miettiä millaista yhtiötä hän preferoi. Varma menestyi vertailussa lähes kaikilla osa-alueilla hyvin suhteessa muihin yhtiöihin. Ilmarinen on yhtiöistä vakavaraisin, mutta yhtiön kustannustehokkuus onvertailun heikoin. Ilmarisen otti yhtiöistä eniten sijoitusriskiä. Varovainen sijoittaja preferoi Eläke-Tapiolaa. Yhtiö oli vertailluista yhtiöistä kustannustehokkain.

Converter-flux-based current control of voltage source PWM rectifiers - analysis and implementation

Pulsewidth-modulated (PWM) rectifier technology is increasingly used in industrial applications like variable-speed motor drives, since it offers several desired features such as sinusoidal input currents, controllable power factor, bidirectional power flow and high quality DC output voltage. To achieve these features,however, an effective control system with fast and accurate current and DC voltage responses is required. From various control strategies proposed to meet these control objectives, in most cases the commonly known principle of the synchronous-frame current vector control along with some space-vector PWM scheme have been applied. Recently, however, new control approaches analogous to the well-established direct torque control (DTC) method for electrical machines have also emerged to implement a high-performance PWM rectifier. In this thesis the concepts of classical synchronous-frame current control and DTC-based PWM rectifier control are combined and a new converter-flux-based current control (CFCC) scheme is introduced. To achieve sufficient dynamic performance and to ensure a stable operation, the proposed control system is thoroughly analysed and simple rules for the controller design are suggested. Special attention is paid to the estimationof the converter flux, which is the key element of converter-flux-based control. Discrete-time implementation is also discussed. Line-voltage-sensorless reactive reactive power control methods for the L- and LCL-type line filters are presented. For the L-filter an open-loop control law for the d-axis current referenceis proposed. In the case of the LCL-filter the combined open-loop control and feedback control is proposed. The influence of the erroneous filter parameter estimates on the accuracy of the developed control schemes is also discussed. A newzero vector selection rule for suppressing the zero-sequence current in parallel-connected PWM rectifiers is proposed. With this method a truly standalone and independent control of the converter units is allowed and traditional transformer isolation and synchronised-control-based solutions are avoided. The implementation requires only one additional current sensor. The proposed schemes are evaluated by the simulations and laboratory experiments. A satisfactory performance and good agreement between the theory and practice are demonstrated.

Patents in the Information and Communications Technology Sector - Development Trends, Problem Areas and Pressures for Change

The patent system was created for the purpose of promoting innovation by granting the inventors a legally defined right to exclude others in return for public disclosure. Today, patents are being applied and granted in greater numbers than ever, particularly in new areas such as biotechnology and information andcommunications technology (ICT), in which research and development (R&D) investments are also high. At the same time, the patent system has been heavily criticized. It has been claimed that it discourages rather than encourages the introduction of new products and processes, particularly in areas that develop quickly, lack one-product-one-patent correlation, and in which theemergence of patent thickets is characteristic. A further concern, which is particularly acute in the U.S., is the granting of so-called 'bad patents', i.e. patents that do not factually fulfil the patentability criteria. From the perspective of technology-intensive companies, patents could,irrespective of the above, be described as the most significant intellectual property right (IPR), having the potential of being used to protect products and processes from imitation, to limit competitors' freedom-to-operate, to provide such freedom to the company in question, and to exchange ideas with others. In fact, patents define the boundaries of ownership in relation to certain technologies. They may be sold or licensed on their ownor they may be components of all sorts of technology acquisition and licensing arrangements. Moreover, with the possibility of patenting business-method inventions in the U.S., patents are becoming increasingly important for companies basing their businesses on services. The value of patents is dependent on the value of the invention it claims, and how it is commercialized. Thus, most of them are worth very little, and most inventions are not worth patenting: it may be possible to protect them in other ways, and the costs of protection may exceed the benefits. Moreover, instead of making all inventions proprietary and seeking to appropriate as highreturns on investments as possible through patent enforcement, it is sometimes better to allow some of them to be disseminated freely in order to maximize market penetration. In fact, the ideology of openness is well established in the software sector, which has been the breeding ground for the open-source movement, for instance. Furthermore, industries, such as ICT, that benefit from network effects do not shun the idea of setting open standards or opening up their proprietary interfaces to allow everyone todesign products and services that are interoperable with theirs. The problem is that even though patents do not, strictly speaking, prevent access to protected technologies, they have the potential of doing so, and conflicts of interest are not rare. The primary aim of this dissertation is to increase understanding of the dynamics and controversies of the U.S. and European patent systems, with the focus on the ICT sector. The study consists of three parts. The first part introduces the research topic and the overall results of the dissertation. The second part comprises a publication in which academic, political, legal and business developments that concern software and business-method patents are investigated, and contentiousareas are identified. The third part examines the problems with patents and open standards both of which carry significant economic weight inthe ICT sector. Here, the focus is on so-called submarine patents, i.e. patentsthat remain unnoticed during the standardization process and then emerge after the standard has been set. The factors that contribute to the problems are documented and the practical and juridical options for alleviating them are assessed. In total, the dissertation provides a good overview of the challenges and pressures for change the patent system is facing,and of how these challenges are reflected in standard setting.

Anaerobic treatment of deinking pulp plant effluents

Anaerobic treatment as a first biological stage in wastewater treatment is nowadays a well-established technology in recycled paper processing mills using closed water circuits. Today further developed high-rate processes and especially high-tower reactors are also able to handle lower organic loads and become therefore feasible for deinking pulp plant effluents. The interest in the anaerobic method is based on a positive energy balance in form of biogas production and low biomass yield from the process. The anaerobic treatment method was researched and its suitability for the deinking pulp plant effluents was tested experimentally at Stora Enso Maxau mill. In the theory, the deinking pulp process is introduced and the effluents from the deinking process are characterized. The anaerobic treatment is brought up in depth in terms of its use for the deinking effluents, and different kind of reactor types are presented. In addition, other wastewater treatment methods are shortly introduced with the focus on tertiary treatment. Static biodegradability tests were carried out for the wastewaters both anaerobically and aerobically. Based on the results, the deinking effluents can be degraded anaerobically, and inhibition to the methanogenic bacteria was not noticed. In the aerobic static test a good performance of the existing wastewater treatment plant at Maxau mill was proved. Later on pilot trials with sequential anaerobic-aerobic treatment were carried out for the deinking effluents. The anaerobic reactor used was a so called internal circulation reactor. The results confirmed that the combination of the anaerobic treatment and the aerobic activated sludge process is a suitable method for deinking wastewaters with a COD reduction as good as with a two stage aerobic method. When combined with the outstanding quality of the produced biogas and the cost savings acquired from the lower sludge production, the anaerobic treatment was found to be an especially favorable treatment method.

STACK OVERWRITING ATTACKS AND DEFENCES IN UNIX ENVIRONMENT

Työn tarkoituksena on tutkia pinon ylikirjoitukseen perustuvien hyökkäysten toimintaa ja osoittaa kokeellisesti nykyisten suojaustekniikoiden olevan riittämättömiä. Tutkimus suoritetaan testaamalla miten valitut tietoturvatuotteet toimivat eri testitilanteissa. Testatut tuotteet ovat Openwall, PaX, Libsafe 2.0 ja Immunix 6.2. Testaus suoritetaan pääasiassa RedHat 7.0 ympäristössä testiohjelman avulla. Testeissä mitataan sekä tuotteiden kyky havaita hyökkäyksiä että niiden nopeusvaikutukset. Myös erityyppisten hyökkäysten ja niitä vastaan kehitettyjen metodien toimintaperiaatteet esitellään seikkaperäisesti ja havainnollistetaan yksinkertaistetuilla esimerkeillä. Esitellyt tekniikat sisältävät puskurin ylivuodot, laittomat muotoiluparametrit, loppumerkittömät merkkijonot ja taulukoiden ylivuodot. Testit osoittavat, etteivät valitut tuotteet estä kaikkia hyökkäyksiä, joten lopuksi perehdytään myös vahinkojen minimointiin onnistuneiden hyökkäysten varalta.

Positioning Wireless Mobile Terminal Devices

Positioning techniques enable the positioning of mobile transportation vehicles. Location information can be used in transport planning as well as in vehicles. Information can also be transferred between vehicles and transport planning. For example, pickup and delivery information and route instructions can be exchanged wirelessly. In this thesis, techniques suitable for positioning vehicles are studied, as well as techniques enabling wireless connectivity of the vehicles to the transport planning system. A prototype based on one positioning and connectivity technique is implemented. As a result, a system capable of positioning vehicles is created. The vehicles can be located almost in real-time. The accuracy in positioning is sufficient to visualise the locations on a good quality 1:20000 city map.

Kovenantit yritysrahoituksessa

Tutkielman päätavoitteena on analysoida kovenanttien eli luottosopimusten eritysehtojen soveltuvuutta yritysrahoitukseen rahoittajan ja yrityksen näkökulmasta ja luoda viitekehys kovenanttien käytettävyydelle yritysrahoituksessa yhdistelemällä siihen liittyvien keskeisten käsitteiden merkityksiä. Tutkielman empiirinen osa on rakennettu pääosin haastatteluihin pohjautuen. Tutkimusmetodologia on toiminta-analyyttinen ja haastattelut suoritettiin teemahaastatteluina. Tutkimuksessa todetaan, että kovenantit sopivat isoille- ja keskisuurille yrityksille ja tietyin varauksin pienille yrityksille. Niiden käytölle ei näyttäisi olevan voimassa olevista säännöksistä johtuvia olennaisia esteitä. Sopimuksissa on huomioitava voimassa olevat säännökset ja sopijapuolten luonteenpiirteet. Keskeisiä tekijöitä ja perusteita kovenanttien asettamisessa todettiin olevan ne, että yrityksen toiminta on vakiintunut ja yhteisymmärrys ja arvio liiketoiminnan kehityksestä on molemminpuolista. Ne monipuolistavat yritysten rahoitusmahdollisuuksia ja parantavat rahoituksen saatavuutta. Tuotteistaminen ei ole mielekästä, koska jokainen laina räätälöidään tapauskohtaisesti.

Brain Dopamine and Serotonin Receptors in the Perception of Pain. Positron Emission Tomography Studies in Healthy Subjects

The role of dopamine and serotonin in spinal pain regulation is well established. However, little is known concerning the role of brain dopamine and serotonin in the perception of pain in humans. The aim of this study was to assess the potential role of brain dopamine and serotonin in determining experimental pain sensitivity in humans using positron emission tomography (PET) and psychophysical methods. A total of 39 healthy subjects participated in the study, and PET imaging was performed to assess brain dopamine D2/D3 and serotonin 5-HT_1A receptor availability. In a separate session, sensitivity to pain and touch was assessed with traditional psychophysical methods, allowing the evaluation of potential associations between D2/D3 and 5-HT_1A binding and psychophysical responses. The subjects’ responses were also analyzed according to Signal Detection Theory, which enables separate assessment of the subject’s discriminative capacity (sensory factor) and response criterion (non-sensory factor). The study found that the D2/D3 receptor binding in the right putamen was inversely correlated with pain threshold and response criterion. 5-HT_1A binding in cingulate cortex, inferior temporal gyrus and medial prefrontal cortex was inversely correlated with discriminative capacity for touch. Additionally, the response criterion for pain and intensity rating of suprathreshold pain were inversely correlated with 5-HT_1A binding in multiple brain areas. The results suggest that brain D2/D3 receptors and 5-HT_1A receptors modulate sensitivity to pain and that the pain modulatory effects may, at least partly, be attributed to influences on the response criterion. 5-HT_1A receptors are also involved in the regulation of touch by having an effect on discriminative capacity.

Epidemiology of Cytochrome P450-mediated Drug-Drug Interactions

Drug-drug interactions (DDIs) comprise an important cause of adverse drug reactions leading to excess hospitalizations. Drug metabolism is catalyzed by 75% by cytochrome P450 (CYP) enzymes and thus they are often involved in pharmacokinetic DDIs. In general, DDIs are studied in randomized controlled clinical trials in selected study populations. The overall aim of the present studies was to perform observational pharmacoepidemiological surveys on CYP-mediated DDIs in diseases important at the population level. The prevalence of co-administrations of four prodrugs (losartan, codeine, tramadol, and clopidogrel), three sulphonylureas (glibenclamide, glimepiride, and glipizide), or two statins (lovastatin and simvastatin) with well established agents altering CYP activity, as well as of statins with fibrates, was studied in Finland utilizing data from a university hospital medication database (inpatients) and the National Prescription Register of the Social Insurance Institution of Finland, Kela (outpatients). Clinical consequences of potential DDIs were estimated by reviewing laboratory data, and information from hospital care and cause-of-death registers. Concomitant use of study substrates with interacting medication was detected in up to one fifth of patients in both hospital and community settings. Potential CYP3A4 interactions in statin users did not manifest in clear adverse laboratory values but pharmacodynamic DDIs between statins and fibrates predisposed patients to muscular toxicity. Sulphonylurea DDIs with CYP2C9 inhibitors increased the risk of hypoglycaemia. CYP3A4 inhibitor use with clopidogrel was not associated with significant changes in mortality but non-fatal thrombosis and haemorrhage complications were seen less often in this group. Concomitant administration of atorvastatin with clopidogrel moderately attenuated the antithrombotic effect by clopidogrel. The overall mortality was increased in CYP3A4 inducer and clopidogrel co-users. Atorvastatin used concomitantly with prodrug clopidogrel seems to be beneficial in terms of total and LDL cholesterol concentrations, and overall mortality compared with clopidogrel use without interacting medication. In conclusion, CYP-mediated DDIs are a common and often unrecognized consequence of irrational drug prescribing.

The Integrin Tail: A Tale of Cell Motility and Division

Integrin transmembrane receptor functions are regulated by adaptor molecules binding to their alpha and beta subunit intracellular domains, or tails, thus affecting integrin traffic and adhesion during e.g. cell motility. Interestingly, many cellular proteins function in both cell motility and cell division, thus raising the possibility that integrins might be involved in regulating the cell cycle. A thorough understanding of cell division is essential in cell biology and in human malignancies. It is well established that failures to complete cell cycle can give rise to genetically unstable cells with tumorigenic properties. Transformed cells promote the disruption of intercellular adhesions such as tight junctions, and this correlates with the onset of cell motility, invasion and unfavorable prognosis in cancer. In this study, we analyzed integrin regulation, mediated by adaptor binding to the  subunit tail, during cell motility and cell division. We revealed a novel molecular mechanism by which Rab21, through association with the integrin alpha subunits, drives integrin endosomal traffic during mitotic phases. In addition, we found indications for this finding in vivo, as RAB21 gene deletions were mapped in ovarian and prostate cancer samples. Importantly, the multinucleated phenotype of cultured ovarian cancer cells could be reverted by Rab21 overexpression. In this thesis work, we also show how the tight junction protein ZO-1 unexpectedly interacts with the 5 integrin cytoplasmic domain in the lamellipodia to promote cell motility and at the cleavage furrow to support separation of the daughter cells. The alpha5-ZO-1 complex formation was dependent on PKC which regulates ZO-1 phosphorylation and its subcellular localization. In addition, by an in situ detection method, we showed that a subset of metastatic human lung cancers expressed the alpha5beta-ZO-1 complex. Taken together, we were able to identify new molecular pathways that regulate integrin functions in an alpha tail-mediated fashion. These findings firmly suggest that genetic alterations in integrin traffic may lead to progression of tumorigenesis as a result of failed cell division. Also, the interplay of integrins and ZO-1 in forming spatially regulated adhesive structures broadens our view of crosstalk between pathways and distinct adhesive structures that can be involved in cancer cell biology.

Design of Pulsed Electroacoustic Measurement System for Space Charge Characterisation

Pulsed electroacoustic (PEA) method is a commonly used non-destructive technique for investigating space charges. It has been developed since early 1980s. These days there is continuing interest for better understanding of the influence of space charge on the reliability of solid electrical insulation under high electric field. The PEA method is widely used for space charge profiling for its robust and relatively inexpensive features. The PEA technique relies on a voltage impulse used to temporarily disturb the space charge equilibrium in a dielectric. The acoustic wave is generated by charge movement in the sample and detected by means of a piezoelectric film. The spatial distribution of the space charge is contained within the detected signal. The principle of such a system is already well established, and several kinds of setups have been constructed for different measurement needs. This thesis presents the design of a PEA measurement system as a systems engineering project. The operating principle and some recent developments are summarised. The steps of electrical and mechanical design of the instrument are discussed. A common procedure for measuring space charges is explained and applied to verify the functionality of the system. The measurement system is provided as an additional basic research tool for the Corporate Research Centre of ABB (China) Ltd. It can be used to characterise flat samples with thickness of 0.2–0.5 mm under DC stress. The spatial resolution of the measurement is 20 μm.

The effectiveness of financial innovation support tools. Comparative study of Latvia and Finland

In the Innovation Union Scoreboard of 2011, Latvia ranked last amongst the EU countries in innovation performance. Even though there is sufficient scientific and technological basis, the results remain modest or low in most of the indicators concerning innovations. Several aspects influence the performance a national innovation system. In Latvia, the low effectiveness is often attributed to lack of financial support tools. As a comparison, Finland was chosen because of its well-established and documented innovation system. The aim of this study is to research the efficiency and effectiveness of the current financial innovation support tool system in Latvia from the point of view of an innovating company. It also attempts to analyze the support tool system of Latvia and compare to the relevant parts of the Finnish system. The study found that it is problematic for innovative companies in Latvia to receive the necessary funding especially for start-ups and SMEs due to the low number of grant programs, funds and lacking offer from banks, venture capital and business angels. To improve the situation, the Latvian government should restructure the funding mechanisms putting a bigger emphasis on innovative start-ups and SMEs. That would lay a foundation for future growth and boost research and scientific activities in Latvia.