No description available.

One common shortcoming of modern computer vision is the inability of most models to generalize to new classes—one/few shot image recognition. We propose a new problem formulation for this task and present a network architecture and training methodology to solve this task. Further, we provide insights into how careful focus on how not just the data, but the way data presented to the model can have significant impact on performance. Using these method, we achieve high accuracy in few-shot image recognition tasks.

In this thesis we developed the Green's Function for a tapered equivalent cylinder model of dendritic electrical propagation. We then use the Green's Function to develop a Carleman linear embedding scheme which is used to estimate the effects of a nonlinear ion channel hot-spot on the tapered cylinder solution. Mathematica© was used to implement the Carleman embedding scheme.

The probit function is the inverse of the cumulative distribution function associated with the standard normal distribution. It is of great utility in statistical modelling. The Carleman embedding technique has been shown to be effective in solving first order and, less efficiently, second order nonlinear differential equations. In this thesis, we show that solutions to the second order nonlinear differential equation for the probit function can be approximated efficiently using a variant of the Carleman embedding technique.

The Carleman Embedding is a method that allows us to embed a finite dimensional system of nonlinear differential equations into a system of infinite dimensional linear differential equations. This technique works well when dealing with first-order nonlinear differential equations. However, for higher order nonlinear ordinary differential equations, it is difficult to use the Carleman Embedding method. This project will examine the Carleman Embedding and a variation of the method which is very convenient in applying to second order systems of nonlinear equations.

High-dimensional data sets can be difficult to visualize and analyze, while data in low-dimensional space tend to be more accessible. In order to aid visualization of the underlying structure of a dataset, the dimension of the dataset is reduced. The simplest approach to accomplish this task of dimensionality reduction is by a random projection of the data. Even though this approach allows some degree of visualization of the underlying structure, it is possible to lose more interesting underlying structure within the data. In order to address this concern, various supervised and unsupervised linear dimensionality reduction algorithms have been designed, such as Principal Component Analysis and Linear Discriminant Analysis. These methods can be powerful, but often miss important non-linear structure in the data. In this thesis, manifold learning approaches to dimensionality reduction are developed. These approaches combine both linear and non-linear methods of dimension reduction.

Redox reactions are crucial to biological processes that protect organisms against oxidative stress. Metalloenzymes, such as cytochrome c peroxidases which reduce excess hydrogen peroxide into water in the periplasm of multiple bacterial organisms, play a key role in detoxification mechanisms. While accurate computational tools can be used to simulate ground state redox potentials in biomolecules, adapting such approaches to properly describe redox reactions in transition metal complexes, particularly in hemes in heterogeneous protein environments, remains a significant challenge. Here we present the results of polarizable hybrid QM/MM studies of the reduction potentials of two heme sites in the cytochrome c peroxidase of Nitrosomonas europaea. The simulated redox potential of the catalytic site Low Potential (LP) is in good agreement with the experiment, while for the High Potential (HP) heme the computational estimate significantly overestimate the experimental value. We have found that environment polarization shifts the computed value of the redox potential of the catalytic LP heme by 1.3 V, while it does not affect that of the non-catalytic High Potential (HP) heme. We demonstrate that it is necessary to account for mutual polarization of heme site and the protein environment when describing redox processes, particularly those that involve more charged heme sites. We have explored the role of various factors such as heme geometries, axial ligands, propionate side chains, and electrostatic field of the protein in tuning the redox potentials of hemes in NeCcP. The fluctuations in computed vertical ionization and electron attachment energies are predominantly affected by fluctuations in the electrostatic field of the environment but not by fluctuations in heme geometries. We attribute the difference in computed LP and HP heme reduction potentials of 0.05 V and 1.15 V, respectively, to different axial ligands and electrostatic interactions of the hemes with the protein environment. / 2023-06-30T00:00:00Z

Throughout the history, humans continue to generate an ever-growing volume of documents about a wide range of topics. We now rely on computer programs to automatically process these vast collections of documents in various applications. Many applications require a quantitative measure of the document similarity. Traditional methods first learn a vector representation for each document using a large corpus, and then compute the distance between two document vectors as the document similarity.In contrast to this corpus-based approach, we propose a straightforward model that directly discovers the topics of a document by clustering its words, without the need of a corpus. We define a vector representation called normalized bag-of-topic-embeddings (nBTE) to encapsulate these discovered topics and compute the soft cosine similarity between two nBTE vectors as the document similarity. In addition, we propose a logistic word importance function that assigns words different importance weights based on their relative discriminating power.Our model is efficient in terms of the average time complexity. The nBTE representation is also interpretable as it allows for topic discovery of the document. On three labeled public data sets, our model achieved comparable k-nearest neighbor classification accuracy with five stateof-art baseline models. Furthermore, from these three data sets, we derived four multi-topic data sets where each label refers to a set of topics. Our model consistently outperforms the state-of-art baseline models by a large margin on these four challenging multi-topic data sets. These works together provide answers to the research question of this thesis:Can we construct an interpretable document represen-tation by clustering the words in a document, and effectively and efficiently estimate the document similarity? / Under hela historien fortsätter människor att skapa en växande mängd dokument om ett brett spektrum av publikationer. Vi förlitar oss nu på dataprogram för att automatiskt bearbeta dessa stora samlingar av dokument i olika applikationer. Många applikationer kräver en kvantitativmått av dokumentets likhet. Traditionella metoder först lära en vektorrepresentation för varje dokument med hjälp av en stor corpus och beräkna sedan avståndet mellan two document vektorer som dokumentets likhet.Till skillnad från detta corpusbaserade tillvägagångssätt, föreslår vi en rak modell som direkt upptäcker ämnena i ett dokument genom att klustra sina ord , utan behov av en corpus. Vi definierar en vektorrepresentation som kallas normalized bag-of-topic-embeddings (nBTE) för att inkapsla de upptäckta ämnena och beräkna den mjuka cosinuslikheten mellan två nBTE-vektorer som dokumentets likhet. Dessutom föreslår vi en logistisk ordbetydelsefunktion som tilldelar ord olika viktvikter baserat på relativ diskriminerande kraft.Vår modell är effektiv när det gäller den genomsnittliga tidskomplexiteten. nBTE-representationen är också tolkbar som möjliggör ämnesidentifiering av dokumentet. På tremärkta offentliga dataset uppnådde vår modell jämförbar närmaste grannklassningsnoggrannhet med fem toppmoderna modeller. Vidare härledde vi från de tre dataseten fyra multi-ämnesdatasatser där varje etikett hänvisar till en uppsättning ämnen. Vår modell överensstämmer överens med de högteknologiska baslinjemodellerna med en stor marginal av fyra utmanande multi-ämnesdatasatser. Dessa arbetsstöd ger svar på forskningsproblemet av tisthesis:Kan vi konstruera en tolkbar dokumentrepresentation genom att klustra orden i ett dokument och effektivt och effektivt uppskatta dokumentets likhet?

For any user-adaptive system the most important task is to provide the users with what they want and need without them asking for it explicitly. This process can be called personalisation and is done by tailoring the service or product for individual users or user groups. In this thesis, we explore the possibilities to build a model that clusters users based on the user’s photo library. This was to create a better personalised experience within a service called Degoo. The model used to perform the clustering is called Deep Embedding Clustering and was evaluated on several internal indices alongside an automated categorization model to get an indication of what type of images the clusters had. The user clustering was later evaluated based on split-tests running within the Degoo service. The results shows that four out of five clusters had some general indication of types such as vacation photos, clothes, text, and people. The evaluation of the clustering impact on the split-tests shows that we could see patterns that indicated optimal attribute values for certain user clusters. / Det ultimata målet för alla användaranpassade system är att ge användarna det som de behöver utan att de begär det explicit. Denna process kan kallas användaranpassning och görs genom att skräddarsy tjänsten eller produkten för enskilda användare eller användargrupper. I denna avhandling undersöker vi möjligheterna att bygga en modell som grupperar användare baserat på användarnas fotodata. Motivationen bakom detta var att skapa en bättre personlig upplevelse inom en tjänst som heter Degoo. Modellen som används för att utföra grupperingen heter Deep Embedding Clustering och utvärderades på flera interna index tillsammans med en automatiserad kategoriseringsmodell för att få en indikation av vilken typ av bilder grupperna hade. Användargrupperingen utvärderades senare baserat på flera split-test som körs inom Degoo tjänsten. Resultaten visar att fyra av fem grupper hade en allmän indikation på typer som semesterbilder, kläder, text och människor. Utvärderingen av grupperingseffekten på split-testerna visar att vi kunde se mönster som indikerar optimala attributvärden för vissa grupper.

Modern retailers have been collecting more and more data over the past decades. The increased sizes of collected data have led to higher demand for data analytics expertise tools, which the Umeå-founded company Infobaleen provides. A recurring challenge when developing such tools is the data itself. Difficulties in finding relevant open data sets have led to a rise in the popularity of using synthetic data. By using artificially generated data, developers gain more control over the input when testing and presenting their work. However, most methods that exist today either depend on real-world data as input or produce results that look synthetic and are difficult to extend. In this thesis, I introduce a method specifically designed to generate synthetic transactional data stochastically. I first examined real-world data provided by Infobaleen to determine suitable statistical distributions to use in my algorithm empirically. I then modelled individual decision-making using points in an embedding space, where the distance between the points serves as a basis for individually unique probability weights. This solution creates data distributed similarly to real-world data and enables retroactive data enrichment using the same embeddings. The result is a data set that looks genuine to the human eye but is entirely synthetic. Infobaleen already generates data with this model when presenting its product to new potential customers or partners.