Abstract This study aimed to investigate the applicability of three different information theoretic similarity measures in image matching, mutual information (MI), cross-cumulative residual entropy (CCRE) and sum of conditional variances (SCV). An experiment was conducted to assess the impact on the performances of the similarity measures when dealing with multimodality, in this case in the context of infrared and visible light. This was achieved by running simulations of four different scenarios using images taken in infrared and visible light, and additionally with variations in amount of details to create different experimental setups. Namely experimental setup A: unimodal data sets with more and less details and experimental setup B: multimodal datasets with more and less details. The result showed that the concept of multimodality gives a statistically significant effect on the performances of all similarity measures. Observations were made that the similarity measures performances also, when trying to match images with different amount of details, differed from each other. This provided a basis for judgement on what measure to use as to give as clear and sound results as possible depending on the variation of detail amount in the data. With this study, it was concluded that the similarity measure CCRE gave the most clear and sound results in the context of multimodality concerning infrared and visible light for both cases of more or less details. Even though the other similarity measures performed well in some cases, CCRE would be to recommend as observed by this study. Keywords : Image matching, image registration, information theoretic similarity measures, multimodal imaging, similarity measures, MI, CCRE, SCV, infrared, visible light. / Denna studie syftade till att undersöka tillämpligheten av tre olika informationsteoretiska likhetsmått vid matchning av bilder, mutual information (MI), cross cumulative residual entropy (CCRE) och sum of conditional variances (SCV). Ett experiment genomfördes för att bedöma hur de olika likhetsmåtten påverkades i kontexten av multimodalitet, i detta fall i samband med infrarött och synligt ljus. Detta uppnåddes genom att köra simuleringar av fyra olika scenarier med hjälp av bilder tagna i infrarött och synligt ljus, och dessutom med variationer i mängden detaljer för att skapa olika experimentella uppsättningar. Nämligen experimentuppsättning A: unimodala datamängder med mer / mindre detaljer och experimentuppsättning B: multimodala datamängder med mer / mindre detaljer.   Resultatet visade att multimodalitet har en statistiskt signifikant påverkan på alla likhetsmått. Observationer gjordes att likhetsmåttens beteenden, när man försöker matcha bilder med olika mängd detaljer, skilde sig från varandra. Detta gav en grund för bedömning av vilken av dessa likhetsmått som då kunde användas för att ge de mest tydliga och stabila resultaten som möjligt beroende på variationen av mängden detaljer i datat. Med denna studie drogs slutsatsen att likhetsmåttet CCRE gav mest de tydliga och stabila resultaten i samband med multimodalitet gällande infrarött och synligt ljus för båda fallen av mer eller mindre detaljer. Även om de andra likhetsmåtten också gav goda resultat i vissa fall, skulle CCRE vara att rekommendera, som observerat i denna studie.

Recommender systems are a relatively new technology that is commonly used by e-commerce websites and streaming services among others, to predict user opinion about products. This report studies two specific recommender algorithms, namely FunkSVD, a matrix factorization algorithm and Item-based collaborative filtering, which utilizes item similarity. This study aims to compare the prediction accuracy of the algorithms when ran on a small and a large dataset. By performing cross-validation on the algorithms, this paper seeks to obtain data that supposedly may clarify ambiguities regarding the accuracy of the algorithms. The tests yielded results which indicated that the FunkSVD algorithm may be more accurate than the Item-based collaborative filtering algorithm, but further research is required to come to a concrete conclusion.

Technical indicators are used by experts in stock trading. The purpose of this report is to investigate whether or not some indicators perform better when applied to stocks of specific market sectors. The investigation was conducted by implementing one algorithm for each of three different technical indicators, Relative Strength Index, Moving Average Convergence-Divergence, and Larry Williams %R. Each algorithm considered one trading strategy. Three market sectors defined by the GICS were included in the tests, Consumer Staples, Utilities, Information Technology. For each of these sectors at least one stock from each industry were tested. Results suggest that the performance of the Relative Strength Index indicator may be related to the sector of the stock to which it is applied, while %R showed no such indication, and MACD showed only a slight performance deviation between sectors. Further and more in-depth studies are required to confirm the results and conclusions drawn in this report.

This thesis evaluates and compares two different clustering strategies for clustering users in Sweden’s political Twittersphere: clustering based on tweet content and clustering based on social interactions data. Users were detected by filtering a stream of tweets filtered on a list of politically charged keywords. The top 10 % of the detected users with the most followers were selected and their social interactions data as well 2,000 of their latest tweets were downloaded. The gathered data was used to construct one similarity matrix for each of the strategies studied. Spectral clustering of the matrices was performed to form two separate sets of clusters, one based on tweet content and one based on social interactions. After analyzing the two cluster sets manually, we find that the content based clustering is biased towards topic based clusters while clustering based on social interactions is more effective in finding clusters centered around ideology and political partisanship. / Denna uppsats utvärderar och jämför två olika strategier för att klustra användare i Sveriges politiska twittersfär: klustring baserat på tweet-innehåll och klustring baserat på sociala interaktioner. Användare upptäcktes genom att filtrera en ström av tweets med hjälp av en lista med politiskt laddade nyckelord. De 10 % av användarna med högst antal följare valdes ut och information om deras sociala interaktioner samt deras 2 000 senaste tweets laddades ner. Denna data användes för att konstruera en likhetsmatris för varje studerad klustringsstrategi. Spektral klustring av matriserna utfördes för att bilda två uppsättningar kluster, en för varje strategi. Efter manuell analys av klustren drogs slutsatsen att innehållsbaserad klustring tenderar att ge genrebaserade kluster medan klustring baserat på sociala interaktioner tenderar att ge kluster som i högre utsträckning cirkulerar kring ideologisk inriktning och politisk partitillhörighet.

This essay is aimed for using as a template for future creations of recommender system. The main purpose of the study is to provide additional information that can be beneficial when creating recommendersystems and it’s an uprising topic in the field of data mining were it’simportant to analyze large collections of data and calculate patterns. Inthe beginning these systems were only useful for simpler tasks, but hasevolved with the help of this contest to a much more complex systemand this study will mainly focus on demonstrating leading methods forcreating such recommender systems. Firstly the methods used are moredetailed explained and some main concepts are brought forward, endingwith a description of the datasets that were released. The results fromthe winning team BellKor’s Pragmatic Chaos will demonstrate the difference between each year for corresponding method and will result in aconclusion that by a analyze of the testing on the data, some of the finalpredictors could be found by using this technique. This will further reducing the quantity of combinations needed for testing and if there is atime pressure or financial issue, this is a strong argument for using thisanalysis as a template for future creations of recommender systems.

This thesis intends to investigate the usefulness of various aspects of product data for user behavior prediction in the online shopping market. Specifically, a data set from BestBuy was used, containing information regarding what product a user clicked on given their search query. Decision trees are machine learning algorithms used for making predictions. The decision tree algorithm ID3 was used because of its simplicity and interpretability. It uses information gain to measure how different attributes help the tree split the set into smaller subsets. The approach was to use one decision tree for each product in the data set, and analyze the distribution of the attributes' maximum information gains in the root splits across the various trees. For each of these splits, all possible pivot values (a pivot value being the value split on) were attempted, and the pivot values were also recorded to analyze which pivot values that resulted in the most gain. The results show that how well the query string matches the product title and description are the two most important aspects, followed by the product's novelty. The number of days since the last two reviews were written before the query proved a decent way to identify trends. The paper also presents how the attributes were used by analyzing the pivot value distributions, with the conclusion that many attributes were used in similar ways for most products, suggesting it might be possible to create a universal tree applicable for all products. Regarding the usefulness of decision trees, it was found that they are not very efficient for highly volatile databases, such as those found in the online shopping market. The notion of a universal tree, however, suggests that future work might investigate whether their efficiency could be improved using this, more flexible, approach.

Making good predictions for stock prices is an important task for the financial industry. The way these predictions are carried out is often by using artificial intelligence that can learn from the data using machine learning algorithms. This study compares three different approaches in this area, using a regression tree and two artificial neural networks with two different learning algorithms. The learning algorithms used was Levenberg-Marquardt and Bayesian regularization. These three approaches was evaluated using the average misprediction and worst misprediction they made in the selected interval from two different indexes, OMXS30 and S&P-500. Out of these three approaches the artificial neural networks outperformed the regression tree and the Bayesian regularization algorithms performed the best out of the two learning algorithms. The conclusions did support the usage of artificial neural networks but was not able to fully establish that the Bayesian regularization algorithm would be the best performing in the general case.

While the thought of having computers recognize objects in images have been around for a long time it is only in the last 20 years that this has become a reality.One of the first successful recognition algorithms was called SIFT and to this day it is one of the most used. However in recent years new algorithms have beenpublished claiming to outperform SIFT. It is the goal of this report to investigate if SIFT still is the top performer 17 years after its publicationor if the newest generation of algorithms are superior. By creating a new data-set of over 170 test images with categories such as scale, rotation, illumination and general detectiona thorough test has been run comparing four algorithms, SIFT, KAZE, AKAZE and ORB. The result of this study contradicts the claims from the creators of KAZE and show thatSIFT has higher score on all tests. It also showed that AKAZE is at least as accurate as KAZE while being significantly faster. Another result was that whileSIFT, KAZE and AKAZE were relatively evenly matched when comparing single invariances that changed when performing tests that contained multiple variables. Whentesting detection in cluttered environments SIFT proved vastly superior to the other algorithms. This led to the conclusion that if the goal is the best possibledetection in every-day situations SIFT is still the best algorithm.

Natural language is a powerful tool of communication between humans. But can it be expanded to entail communication between human and computer? Or is it too complex and ambiguous for computers to understand the meaning behind the words? This paper will explore the possibility of a computer game interface that can perfectly understand the meaning behind the input given in natural language. This is done through an implementation of such a interface on a chess game. The result gathered show that it is not too far fetched to expect future user interfaces to handle natural language with perfect accuracy in certain contexts.

Programming languages such as C/C++ allow for great flexibility in how code can be written. This leads to pro- grammers developing their own “code style” that can be used to identify them among a group of other programmers, in a setting such as a programming competition. Recent research has shown that some of the identifying stylistic features present in source code survive the compilation pro- cess, and that authorship classification can be performed on the compiled executables alone. This was originally per- formed by Rosenblum et al. in their 2011 paper on the subject. This thesis takes the approach of Rosenblum et al. and in- vestigates how the author classification process is a ected by changes in the compilation process of the training dataset, specifically di erent levels of optimisation (-O1 to -O3) and static linkage. We find that full optimisation yields a 10% drop in accuracy in datasets with 413 and 20 authors re- spectively. Static linkage results in a significant drop in accuracy in datasets with 20 and 10 authors, respectively. In both cases, the classifiers still perform significantly bet- ter than random chance and as such these methods cannot guarantee anonymity to the programmer. It is not clear how these results translate to other datasets, although there is reason to believe they would be reproducible using other classifiers found in the literature.