Browsing by Issue Date

There are two primary types of quantum computers: quantum annealers and circuit model computers. Quantum annealers are specifically designed to tackle particular problems, as opposed to circuit model computers, which can be viewed as universal quantum computers. Substantial efforts are underway to develop quantum-based algorithms for various classical computational problems. The objective of this thesis is to implement algorithms for solving graph problems using quantum annealer computers and analyse these implementations. The aim is to contribute to the ongoing development of algorithms tailored for this type of machine. Three distinct types of graph problems were selected: all pairs shortest path, graph isomorphism, and community detection. These problems were chosen to represent varying levels of computational complexity. The algorithms were tested using the D-Wave quantum annealer Advantage system 4.1, equipped with 5760 qubits. D-Wave provides a cloud platform called Leap and a Python library, Ocean tools, through which quantum algorithms can be designed and run using local simulators or real quantum computers in the cloud. Formulating graph problems to be solved on quantum annealers was relatively straightforward, as significant literature already contains implementations of these problems. However, running these algorithms on existing quantum annealer machines proved to be challenging. Even though quantum annealers currently boast thousands of qubits, algorithms performed satisfactorily only on small graphs. The bottleneck was not the number of qubits but rather the limitations imposed by topology and noise. D-Wave also provides hybrid solvers that utilise both the Quantum Processing Unit (QPU) and CPU to solve algorithms, which proved to be much more reliable than using a pure quantum solver.

The Venusian atmosphere has everything to be an exciting natural sulfur laboratory. In addition to relatively high concentrations of sulfur dioxide, suitable conditions in the atmosphere make both thermo- and photochemical reactions possible, allowing for complex chemical reactions and the formation of new sulfur containing compounds. These compounds could explain or contribute to the enigmatic 320-400 nm absorption feature in the atmosphere. One of the proposed absorbers is polysulfur compounds. While some experimentally obtained UV-VIS spectra have been published, studying the different polysulfur species individually is extremely difficult due to the reactive nature of sulfur. In this thesis UV-VIS spectra for polysulfur species S2 to S8 were simulated using the nuclear ensemble approach to determine if they fit the absorption profile. In total, 38 polysulfur species were considered. All were optimized at the wB97X-D/aug-cc-pV(T+d)Z level of theory, with the S2, S3, and S4 structures also being optimized at the CCSD(T)/aug-cc-pV(T+d)Z level of theory. For 13 structures UV-VIS spectra were simulated using a nuclear ensemble of 2000 geometries, with vertical excitations calculated at the EOM-CCSD/def2-TZVPD or the wB97X-D/def2-TZVPD levels of theory. The simulated UV-VIS spectra for the smaller species were in quite good agreement with experimental ones. Two different molecules were identified with substantial absorption cross sections in the range of the unknown absorber: The open chain isomer of S3 (3.78×10^-17 cm^2 at 370 nm), and the trigonal isomer of S4 (4.76×10^-17 cm^2 at 360 nm). However, the mixing ratios of these species in the Venusian atmosphere are also needed to make a more conclusive statement. Other polysulfur compounds have insignificant absorption cross sections in the 320-400 nm range and can therefore be excluded. The calculated absorption cross sections can be used to calculate photolysis rates, which can be straight away added to atmospheric models of Venus. In addition, this work will help future space missions to Venus, for example by focusing their search for the unknown absorber.

The traditional method for identifying sulfate soils has been the incubation method, which typically takes 9-19 weeks. However, in collaboration, the Finnish Environment Institute (SYKE), Geological Survey of Finland (GTK), and Åbo Akademi developed a faster hydrogen peroxide oxidation method for identifying sulfate soils and assessing acidity potential. This method allows for sulfate soil identification and acidity potential estimation in just a few hours. The hydrogen peroxide oxidation method was used to identify sulfate soils in the Helsinki region and to evaluate the method. The study areas included the Sunnuntaipalsta-field area in Malmi, the area associated with the relocation of Gasgrid’s gas pipeline in Pihlajamäki, and the Hermanninranta- Kyläsaari area. Sulfate concentrations determined by the oxidation method were compared with concentrations obtained through water extraction at the Helsinki geophysical, environmental and mineralogical laboratories (Hellabs) of the University of Helsinki's Department of Geology and Geophysics, and acid extraction at ALS Finland Ltd. In Malmi, the method worked well and reliably, indicating naturally acidified soil with relatively low sulfur concentrations. Deeper layers revealed potential acidic sulfate soil materials. In Pihlajamäki, the method was effective, identifying clear potential acidic sulfate soils even with samples consisting of clay fillings. Challenges arose in the Hermanninranta-Kyläsaari area due to contaminated fill soils with high pH values and various hydrocarbons. The lower layers of the samples were rich in organic matter (LOI > 10%), causing the hydrogen peroxide oxidation method to overestimate sulfate concentrations, resulting in deviations with both acid and water extraction results. Based on the results, the hydrogen peroxide oxidation method performs most reliably when loss on ignition (LOI) is < 10% and the pH change (ΔpH) after oxidation is less than 5 units. The method could be a valuable addition to soil investigations conducted by the City of Helsinki's construction services public enterprise, Stara, in their Street and ground laboratory. The method is effective and enables the rapid identification of potential acidic sulfate soils.

Sums of log-normally distributed random variables arise in numerous settings in the fields of finance and insurance mathematics, typically to model the value of a portfolio of assets over time. In particular, the use of the log-normal distribution in the popular Black-Scholes model allows future asset prices to exhibit heavy tails whilst still possessing finite moments, making the log-normal distribution an attractive assumption. Despite this, the distribution function of the sum of log-normal random variables cannot be expressed analytically, and has therefore been studied extensively through Monte Carlo methods and asymptotic techniques. The asymptotic behavior of log-normal sums is of especial interest to risk managers who wish to assess how a particular asset or portfolio behaves under market stress. This motivates the study of the asymptotic behavior of the left tail of a log-normal sum, particularly when the components are dependent. In this thesis, we characterize the asymptotic behavior of the left and right tail of a sum of dependent log-normal random variables under the assumption of a Gaussian copula. In the left tail, we derive exact asymptotic expressions for both the distribution function and the density of a log-normal sum. The asymptotic behavior turns out to be closely related to Markowitz mean-variance portfolio theory, which is used to derive the subset of components that contribute to the tail asymptotics of the sum. The asymptotic formulas are then used to derive expressions for expectations conditioned on log-normal sums. These formulas have direct applications in insurance and finance, particularly for the purposes of stress testing. However, we call into question the practical validity of the assumptions required for our asymptotic results, which limits their real-world applicability.

Machine Learning (ML) has experienced significant growth, fuelled by the surge in big data. Organizations leverage ML techniques to take advantage of the data. So far, the focus has predominantly been on increasing the value by developing ML algorithms. Another option would be to optimize resource consumption to reach cost optimality. This thesis contributes to cost optimality by identifying and testing frameworks that enable organizations to make informed decisions on cost-effective cloud infrastructure while designing and developing ML workflows. The two frameworks we introduce to model Cost Optimality are: "Cost Optimal Query Processing in the Cloud" for data pipelines and "PALEO" for ML model training pipelines. The latter focuses on estimating the training time needed to train a Neural Net, while the first one is more generic in assessing cost-optimal cloud setup for query processing. Through the literature review, we show that it is critical to consider both the data and ML training aspects when designing a cost-optimal ML workflow. Our results indicate that the frameworks provide accurate estimates about cost-optimal hardware configuration in the cloud for ML workflow. There are deviations when we dive into the details: our chosen version of the Cost Optimal Model does not consider the impact of larger memory. Also, the frameworks do not provide accurate execution time estimates: PALEO estimates our accelerated EC2 instance to execute the training workload with half of the time it took. However, the purpose of the study was not to provide accurate execution or cost estimates, but we aimed to see if the frameworks estimate the cost-optimal cloud infrastructure setup among the five EC2 instances that we chose to execute our three different workloads.

Asukkaiden tyytyväisyyttä omaan asuinalueeseensa pidetään yhtenä yleisimmistä asuinalueiden laatua tarkastelevista mittareista. Asumistyytyväisyys on myös liitetty oleelliseksi osaksi ihmisen yleistä elämänlaatua ja asuinympäristöä voidaan pitää ihmisen arkielämän tärkeimpänä elinympäristön osa-alueena. Tässä tutkielmassa tutkimuskohteena olivat lähiöissä asuvien henkilöiden asumistyytyväisyys. Lähiöt ovat oleellinen osa suomalaista kaupunkikuvaa ja Suomessa lähiöiden, ja niissä asuvien ihmisten määrä on merkittävä. Kun puhutaan tyytyväisyydestä, niin puhutaan ihmisten subjektiivisesta kokemuksesta. Tutkittaessa aihetta, joka on ihmisten subjektiivisista kokemuksista riippuvainen, luo se tutkimukselle omat haasteensa. On tiedossa, että ihmisillä on suuria eroja siinä, kuinka he pystyvät ilmaisemaan epämukavuutta sekä tyytyväisyyttä. Kun siis tehdään tutkimusta, jossa tyytyväisyyden mittarina käytetään henkilön subjektiivista arviota, tulokset perustuvat siihen olettamukseen, että ihmisten omat arviot elämästään ja tyytyväisyyden tasostaan ovat luotettavia mittareita. Tämän tutkielman tavoitteena on selvittää, millä tekijöillä on yhteys suomalaisissa lähiöissä asuvien henkilöiden asumistyytyväisyyteen. Tutkimuksen aineistona toimii Helsingin yliopiston ja Suomen Akatemian rahoittaman PREFARE-hankkeen (2012–2015) aikana hankittu aineisto. Aineistoon on kyselydatan lisäksi yhdistetty vastaajien sijainnin perusteella kontekstuaalista rekisteritietoa, jotka saatiin ruututietokannasta. Aineisto on siis kaksitasoinen, yksilötaso, jossa vastaajia oli 7728 ja lähiötaso, jossa alueita oli 71. Aineiston analyysimenetelmänä toimi monitasoinen lineaarinen regressiomalli, jonka avulla pystyttiin analysoimaan monitasoista aineistoa. Analyysissä aineistosta luotiin asumistyytyväisyyttä kuvaava tyytyväisyysmuuttuja ja tutkielmassa tarkasteltiin aluetasonmuuttujien, yksilötason taustamuuttujien ja asuinympäristön kokemiseen liittyvien muuttujien yhteyksiä tähän tyytyväisyysmuuttujaan. Tutkielman oleellisin tulos oli maineen merkitys asumistyytyväisyyteen. Aluetason muuttujien yhteydet tyytyväisyysmuuttujaan kulkivat hyvin vahvasti mainemuuttujan kautta ja maineella oli myös tarkasteltavista muuttujista vahvin yhteys tyytyväisyyteen. Maineen lisäksi kaksi muuta kokemusmuuttujaa, turvallisuus ja naapurisuhteet olivat hyvin vahvassa yhteydessä asumistyytyväisyyden kanssa. Yksilötason taustamuuttujista vahvimmat yhteydet tyytyväisyysmuuttujaan olivat tulotasolla, iällä ja asunnon hallintamuodolla.

This work examines how neural networks can be used to qualitatively analyze systems of differential equations depicting population dynamics. We present a novel numerical method derived from physics informed learning, capable of extracting equilibria and bifurcations from population dynamics models. The potential of the framework is showcased three different example problems, a logistic model with outside inference, the Rosenzweig-MacArthur model and one model from a recent population dynamics paper. The key idea behind the method is having a neural network learn the dynamics of a free parameter ODE system, and then using the derivatives of the neural network to find equilibria and bifurcations. We, a bit clunkily, refer to these networks as physics informed neural networks with free parameters and variable initial conditions. In addition to these examples, we also survey how and where these neural networks could be further utilized in the context of population dynamics. To answer the how, we document our experiences choosing good hyperparameters for these networks, even venturing into previously unexplored territory. For the where, we suggest potentially useful neural network frameworks to answer questions from an external survey concerning contemporary open questions in population dynamics. The research of the work is preceded by a short dive on qualitative population dynamics, where we ponder what are the problems we want to solve and what are the tools we have available for that. Special attention is paid to parameter sensitivity analysis of ordinary differential equation systems through bifurcation theory. We also provide a beginner friendly introduction to deep learning, so that the research can be understood even by someone not previously familiar with the field. The work was written, and all included contents were selected, with the goal of establishing a basis for future research.

The rapid growth of artificial intelligence (AI) and machine learning (ML) solutions has created a need to develop, deploy and maintain AI/ML those to production reliably and efficiently. MLOps (Machine Learning Operations) framework is a collection of tools and practices that aims to address this challenge. Within the MLOps framework, a concept called the feature store is introduced, serving as a central repository responsible for storing, managing, and facilitating the sharing and reuse of extracted features derived from raw data. This study gives first an overview of the MLOps framework and delves deeper into feature engineering and feature data management, and explores the challenges related to these processes. Further, feature stores are presented, what they exactly are and what benefits do they introduce to organizations and companies developing ML solutions. The study also reviews some of the currently popular feature store tools. The primary goal of this study is to provide recommendations for organizations to leverage feature stores as a solution to the challenges they encounter in managing feature data currently. Through an analysis of the current state-of-the-art and a comprehensive study of organizations' practices and challenges, this research presents key insights into the benefits of feature stores in the context of MLOps. Overall, the thesis highlights the potential of feature stores as a valuable tool for organizations seeking to optimize their ML practices and achieve a competitive advantage in today's data-driven landscape. The research aims to explore and gather practitioners' experiences and opinions on the aforementioned topics through interviews conducted with experts from Finnish organizations.

Response Surface Models (RSM) are cheap, reduced complexity, and, usually, statistical models that are fit to the response of more complex models to approximate their outputs with higher computational efficiency. In atmospheric science, there has been a continuous push to reduce the amount of training data required to fit an RSM. With this reduction in costly data gathering, RSMs can be used more ad hoc and quickly adapted to new applications. However, with the decrease in diverse training data, the risk increases that the RSM is eventually used on inputs on which it cannot make a prediction. If there is no indication from the model that its outputs can no longer be trusted, trust in an entire RSM decreases. We present a framework for building prudent RSMs that always output predictions with confidence and uncertainty estimates. We show how confidence and uncertainty can be propagated through downstream analysis such that even predictions on inputs outside the training domain or in areas of high variance can be integrated. Specifically, we introduce the Icarus RSM architecture, which combines an out-of-distribution detector, a prediction model, and an uncertainty quantifier. Icarus-produced predictions and their uncertainties are conditioned on the confidence that the inputs come from the same distribution that the RSM was trained on. We put particular focus on exploring out-of-distribution detection, for which we conduct a broad literature review, design an intuitive evaluation procedure with three easily-visualisable toy examples, and suggest two methodological improvements. We also explore and evaluate popular prediction models and uncertainty quantifiers. We use the one-dimensional atmospheric chemistry transport model SOSAA as an example of a complex model for this thesis. We produce a dataset of model inputs and outputs from simulations of the atmospheric conditions along air parcel trajectories that arrived at the SMEAR II measurement station in Hyytiälä, Finland, in May 2018. We evaluate several prediction models and uncertainty quantification methods on this dataset and construct a proof-of-concept SOSAA RSM using the Icarus RSM architecture. The SOSAA RSM is built on pairwise-difference regression using random forests and an auto-associative out-of-distribution detector with a confidence scorer, which is trained with both the original training inputs and new synthetic out-of-distribution samples. We also design a graphical user interface to configure the SOSAA model and trial the SOSAA RSM. We provide recommendations for out-of-distribution detection, prediction models, and uncertainty quantification based on our exploration of these three systems. We also stress-test the proof-of-concept SOSAA RSM implementation to reveal its limitations for predicting model perturbation outputs and show directions for valuable future research. Finally, our experiments affirm the importance of reporting predictions alongside well-calibrated confidence scores and uncertainty levels so that the predictions can be used with confidence and certainty in scientific research applications.

Node classification is an important problem on networks in many different contexts. Optimizing the graph embedding has great potential to help improve the classification accuracy. The purpose of this thesis is to explore how graph embeddings can be exploited in the node classification task in the context of citation networks. More specifically, this thesis looks into the impact of different kinds of embeddings on the node classification, comparing their performance. Using three different similarity functions and different dimensions for the embedding vector ranging from 1 to 800, we examined the impact of graph embeddings on accuracy in node classification using three benchmark datasets: Cora, Citeseer, and PubMed. Our experimental results indicate that there are some common tendencies in the way dimensionality impacts the graph embedding quality regardless of the graph. We also established that some network-specific hyperparameter tuning clearly affects classification accuracy.