Social media has become an integral part of the Internet. There have been users across the world sharing content like images, texts, videos, and so on. There is a huge amount of data being generated and it has become a challenge to the social media platforms to group the content for further usage like recommending a video. Especially, grouping videos based on similarity requires extracting features. This thesis investigates potential approaches to extract features that can help in determining the similarity between videos. Features of given videos are extracted using Object Detection and Action Recognition. Bag-of-features representation is used to build the vocabulary of all the features and transform data that can be useful in clustering videos. Probabilistic model-based clustering, Multinomial Mixture model is used to determine the underlying clusters within the data by maximizing the expected log-likelihood and estimating the parameters of data as well as probabilities of clusters. Analysis of clusters is done to understand the genre based on dominant actions and objects. Bayesian Information Criterion(BIC) and Akaike Information Criterion(AIC) are used to determine the optimal number of clusters within the given videos. AIC/BIC scores achieved minimum scores at 32 clusters which are chosen to be the optimal number of clusters. The data is labeled with the genres and Logistic regression is performed to check the cluster performance on test data and has achieved 96% accuracy

This project seeks to demonstrate routing optimization that would allow pedestrian and micromobility user groups to select and prioritize different route features according to their preferences. Through the creation of a routing demonstrator that considers both static and dynamic parameters in the form of pavement quality, elevation climb, travel time, and air quality, along with user-specified weights for their prioritization of each of these parameters, a number of routes were created and mapped to qualitatively compare against routes representing only a shortest path. / <p>Examensarbetet är utfört vid Institutionen för teknik och naturvetenskap (ITN) vid Tekniska fakulteten, Linköpings universitet</p>

Submitted to the faculty of the Bioinformatics Graduate Program in partial fulfillment of the requirements for the degree Master of Science in the School of Informatics, Indiana University December 2005 / The last decade has witnessed numerous proteomic studies which have predicted and successfully confirmed the existence of extended structurally flexible regions in protein molecules. Parallel to these advancements, the last five years of structural bioinformatics has also experienced an explosion of results on molecular recognition and its importance in protein-protein interactions. This work provides an extension to past and ongoing research efforts by looking specifically at the “flexibility and disorder†found in protein sequences involved in molecular recognition processes and known as, Molecular Recognition Elements or Molecular Recognition Features (MoREs or MoRFs, as we call them). MoRFs are relatively short in length (10 – 70 residues length); loosely structured protein regions within longer sequences that are largely disordered in nature. Interestingly, upon binding to other proteins, these MoRFs are able to undergo disorder-to-order transition. Thus, in our interpretation, MoRFs could serve as potential binding sites, and that this binding to another protein lends a functional advantage to the whole protein complex by enabling interaction with their physiological partner. There are at least three basic types of MoRFs: those that form α-helical structures upon binding, those that form β-strands (in which the peptide forms a β-sheet with additional β-strands provided by the protein partner), and those that form irregular structures when bound. Our proposed names for these structures are α-MoRF (also known as α-MoRE, alpha helical molecular recognition feature/element), β-MoRF (beta sheet molecular recognition feature/element), and I-MoRF (Irregular molecular recognition feature/element), respectively. The results presented in this work suggest that functionally significant residual structure can exist in MoRF regions prior to the actual binding event. We also demonstrate profound conformational preferences within MoRF regions for α-helices. We believe that the results from this study would subsequently improve our understanding of protein-protein interactions especially those related to the molecular recognition, and may pave way for future work on the development of protein binding site predictions. We hope that via the conclusions of this work, we would have demonstrated that within only a few of years of its conception, intrinsic protein disorder has gained wide-scale importance in the field of protein-protein interactions and can be strongly associated with molecular recognition.

Forged videos of swapped faces, so-called deepfakes, have gained a  lot  of  attention in recent years. Methods for automated detection of this type of manipulation are also seeing rapid progress in their development. The purpose of this thesis work is to evaluate the possibility and effectiveness of using deep embeddings from facial recognition networks as base for detection of such deepfakes. In addition, the thesis aims to answer whether or not the identity embeddings contain information that can be used for detection while analyzed over time and if it is suitable to include information about the person's head pose in this analysis. To answer these questions, three classifiers are created with the intent to answer one question each. Their performances are compared with each other and it is shown that identity embeddings are suitable as a basis for deepfake detection. Temporal analysis of the embeddings also seem effective, at least for deepfake methods that only work on a frame-by-frame basis. Including information about head poses in the videos is shown to not improve a classifier like this.

Webbaserade kontrollerade experiment håller alltmer på att bli normen för mjukvaruföretagsom vill bli data-drivna. Ett av de vanligaste experimenten idag är A/Btester.Att få resultat från A/B-tester är enkelt, att få trovärdiga resultat är svårare.Även om forskningsområdet gällande A/B-testning är aktivt så finns det ännu ingenöverenskommen definition av nödvändiga förutsättningar för trovärdig A/B-testning.Denna uppsats presenterar en uppsättning förutsättningar och kopplar samman dessamed egenskaper för A/B-testningsplattformar. Resultaten visar vilka egenskaper enplattform ska inneha för att uppnå trovärdiga resultat. Uppsatsen presenterar ocksåen utvärdering av två vanliga A/B-testningsplattformar. Under denna utvärdering upptäcktesegenskaper som kan ha motsatt effekt på trovärdigheten av resultaten vilketockså diskuteras i uppsatsen. / Web-based controlled experiments are increasingly becoming the norm for softwaredevelopment companies who want to become data-driven. One of the most commonexperiments today is A/B-tests. While it is easy to get results from A/B-tests, gettingtrustworthy results is harder. Although the research field of A/B-testing is active,there is no agreed definition of the necessary conditions for running trustworthy A/Btest.This thesis presents a set of trustworthy conditions and connect these with A/Btestingplatform features. The results show what features to look for in platforms inorder to achieve trustworthy results. The thesis also presents an evaluation of twocommon A/B-testing platforms available today. During the platform evaluation otherfeatures were found that could have the opposite impact on the trustworthiness of theresults and this is also discussed in this thesis.

Att en spelare lever sig in i ett spel kan ses som en av de viktigaste delarna av ett bra spel och spelare vill ständigt ha spel där dom känner mer och mer inlevelse. Tidigare forskning visar på att inlevelse i digitala spel inte är ett enkelt område och för att kunna forska på det så krävs det att man delar upp det i mindre, mer hanterbara, delområden som kan undersökas både som enskilda områden och i relation till andra. Denna uppsats bryter ut tre delområden som alla bidrar till inlevelse i spel för att utforska, testa och utvärdera. De valda delområdena används för att skapa en artefakt i form av ett spel där delområdena är implementerade och kan testas. De resultat som presenteras i detta arbete visar på att olika delområden inom inlevelse påverkar inlevelsen i ett spel olika mycket. Vidare visar även denna uppsats på hur vissa av dessa delområden relaterar till varandra och hur de tillsammans påverkar inlevelsen i ett spel som helhet. / Immersion can be considered as an essential part in digital games and developers are constantly challenged when trying to create immersive game experiences to an ever growing demand. However, as previous work suggests, immersion is not an easy concept to grasp and the area must be divided into smaller sub-areas. The sub-areas can then be investigated both individually and in relation to one another. This thesis breaks out three sub-areas (immersive features), that contribute to the overall feeling of immersion, to explore and test. The immersive features are used to create an artifact in the form of a game where all features can be tested. The data presented in this thesis shows that the three features have different amounts of impact on immersion. Furthermore, this thesis shows how the selected features relate to each other and how they together affect the overall game.

Students who speak local varieties (i.e., dialects) of English that differ from the codified variety promoted in school are at a disadvantage. Research illustrates that differences in sound systems, grammar, vocabulary, and usage can negatively affect literacy development and achievement in math and science, and lead to misunderstandings and changes in teacher attitudes toward students. Moreover, the use of inappropriate assessment tools may result in unnecessary pathologization and inappropriate pedagogical approaches. Since many Indigenous children may speak local varieties, it is reasonable to assume that the same issues that hinder school success for speakers of other varieties affect many Indigenous students in Canada in similar ways. However, to date, research concerning Indigenous Englishes in Canada is scant. Similarly, virtually no empirical evidence has been gathered on use in Canadian schools. By extension, the trajectory of use of features as children progress through grades remains unknown. The goal of this research was to begin to address the crucial necessity of learning more about Indigenous English varieties, in order that appropriate language assessment and pedagogical practices can be implemented. The research, conducted in a remote community in Northern British Columbia, Canada, concentrates on differences in grammar used by a group of First Nations school-aged children. I analyzed oral narrative language samples of Kindergarteners, and oral and written narrative language samples of students in Kindergarten to Grade 5, over a three-year period. Results reveal the presence of at least 23 distinct grammatical features, many of which may have been influenced by the structure of the ancestral language. At school entry, students used grammatical features at high rates, regardless of whether or not they later required speech-language pathology or special education services. As children progressed through the grades, the rate at which they produced features appeared to follow a curvilinear trajectory, declining until grades 3 and 4 and then gradually rising again in middle school. A preference for using shorter sentences with less use of subordination and embedding of clauses also appears to be a feature of this variety. Most of the features the children used in their speech, they also used in their writing. Children had the most difficulty switching to standard English forms of verb tense, and so verb tense may require more direct instruction. While my results may not be directly generalizable to other First Nations communities, it is anticipated that educators will use them as a guide in their practice and instruction, so they can cease confusing features of a local variety with errors requiring “correction”, avoid unnecessary pathologization, and adjust expectations regarding the rate at which children can be expected to acquire the codified standard language model. It is also hoped that this study will contribute to the preservation and celebration of the unique ways of speaking English that have evolved in northern communities. / Graduate / 2021-09-16

The work presented here is an important component of an on going project of developing an automated mass classification system for breast cancer screening and diagnosis for Digital Mammogram applications. Specifically, in this work the task of automatically separating mass tissue from normal breast tissue given a region of interest in a digitized mammogram is investigated. This is the crucial stage in developing a robust automated classification system because the classification depends on the accurate assessment of the tumor-normal tissue border as well as information gathered from the tumor area. In this work the Expectation Maximization (EM) method is developed and applied to high resolution digitized screen-film mammograms with the aim of segmenting normal tissue from mass tissue. Both the raw data and summary data generated by Laws' texture analysis are investigated. Since the ultimate goal is robust classification, the merits of the tissue segmentation are assessed by its impact on the overall classification performance. Based on the 300 image dataset consisting of 97 malignant and 203 benign cases, a 63% sensitivity and 89% specificity was achieved. Although, the segmentation requires further investigation, the development and related computer coding of the EM algorithm was successful. The method was developed to take in account the input feature correlation. This development allows other researchers at this facility to investigate various input features without having the intricate understanding of the EM approach.

It has been estimated that approximately fifty percent of the United States’ population lives behind levees. Because these earth structures are typically long, subjected to seasonal changes and spatial variability, it is logical to analyze them in an uncertainty-based approach. This research is focused on assessing the potential of internal erosion related failure due to underseepage with the general objective of assessing the failure potential of individual geomorphic features while considering length effects. The project team was granted $204,000 from the National Science Foundation and $10,000 from the United States Society on Dams which resulted in research collaboration within graduate students and University of Delaware faculty as well as several presentations in prestigious conferences (in the U.S and Europe) and publication of journal papers. Findings from this research should be easily transferrable to other linear earth structures (such as dams, construction excavations, detention ponds, road embankments, etc.), and should significantly enhance reliability analysis across a wide array of structure types and geologic settings allowing a broad impact on future research, as well as geotechnical engineering practice.

An AdaBoost algorithm for construction of strong classifier from several weak hypotesis will be presented in this work. Theoretical background of the algorithm and the method of construction of strong classifiers will be explained. WaldBoost extension to the algorithm will be described. The thesis deals with image features that are often used as element of weak classifiers. Brief introduction to pattern recognition in context of computer vision will be outlined in the begining of the work. Also some widely used methods of classifier training will be presented. An object detection library based on AdaBoost classifiers was developed as part of the work. The library was used in implementation of software that in praktice demonstrates object detection in videosquences. Last part of the work describes tool for training of AdaBoost classifiers.