Biometrics are widely used in identification systems, such as face, fingerprint, iris, etc. Polar code is the only code that can be strictly proved to achieve channel capacity, and it has been proved to be optimal for channel and source coding. In this degree project, our goal is to apply polar codes algorithms to biometric identification systems, and to design a biometric identification system with high identification accuracy, low system complexity, and good privacy preservation. This degree project has carried out specific and in-depth research in four aspects, following results are achieved: First, idea of polar codes is learnt, for example channel combination, channel splitting, successive cancellation decoding. The successive cancellation and successive cancellation list algorithm are also applied to encoding, which further realizes polar codes for source coding. Second, using autoencoder to process biometrics. Autoencoder is introduced to compress fingerprints into binary sequences of length 1024, it has 5 encoding layers and 12 decoding layers, achieved reconstruction error is 0.03. The distribution is close to Gaussian distribution, and compressed codes are quantized into binary sequences. Properties of sequences are similar with random sequences in terms of entropy, correlation, variance. Third, the identification system under Wyner-Ziv problem is studied with fingerprints. In enrollment phase, encoding algorithms are designed to compress biometrics, and in identification phase, decoding algorithms are designed to estimate the original sequence based on decoded results and noisy sequence. Maximum mutual information method is used to identify users. Results show that with smaller number of users, longer code length, smaller noise, then recognition error rate is lower. Fourth, human faces are used in the generated secret key system. After fully considering the trade off to achieve optimal results, in enrollment phase both public data and secure data are generated, in identification phase user’s index and secret key are estimated. A hierarchical structure is further studied. First, CNN is used to classify the age of faces, and then the generated secret key system is used for identification after narrowing the range. The system complexity is reduced by 80% and the identification accuracy is not reduced. / Biometriska kännetecken används i stor utsträckning i identifieringssystem, kännetecken såsom ansikte, fingeravtryck, iris, etc. Polär kod är den enda koden som strikt bevisats uppnå kanalkapacitet och den har visat sig vara optimal för kanal- och källkodning. Målet med detta examensarbete är att tillämpa polära kodalgoritmer på biometriska identifieringssystem, och att designa ett biometriskt identifieringssystem med hög identifieringsnoggrannhet, låg systemkomplexitet och bra integritetsskydd. Under examensarbetet har det genomförts specifik och djupgående forskning i fyra aspekter, följande resultat har uppnåtts: För det första introduceras idén om polära koder, till exempel kanalkombination, kanaluppdelning, successiv annulleringsavkodning. Algoritmerna för successiv annullering och successiv annulleringslista tillämpas även på kodning,vilket ytterligare realiserar polära koders användning för källkodning. För det andra används autoencoder för att bearbeta biometriska uppgifter. Autoencoder introduceras för att komprimera fingeravtryck till binära sekvenser med längden 1024, den har 5 kodningslager och 12 avkodningslager, det uppnådda rekonstruktionsfelet är 0,03. Fördelningen liknar en normaldistribution och komprimerade koder kvantiseras till binära sekvenser. Egenskaperna för sekvenserna liknar slumpmässiga sekvenser vad gäller entropi, korrelation, varians. För det tredje studeras identifieringssystemet under Wyner-Ziv-problemet med fingeravtryck. I inskrivningsfasen är kodningsalgoritmer utformade för att komprimera biometriska kännetecken, och i identifieringsfasen är avkodningsalgoritmer utformade för att estimera den ursprungliga sekvensen baserat på avkodade resultat och brusiga sekvenser. Maximal ömsesidig informationsmetod används för att identifiera användare. Resultaten visar att med ett mindre antal användare, längre kodlängd och mindre brus så är identifieringsfelfrekvensen lägre. För det fjärde används mänskliga ansikten i det genererade hemliga nyckelsystemet. Efter att noggrant ha övervägt kompromisser fullt ut för att uppnå det optimala resultatet genereras både offentlig data och säker data under registreringsfasen, i identifieringsfasen uppskattas användarens index och säkerhetsnyckel. En hierarkisk struktur studeras vidare. Först används CNN för att klassificera ålder baserat på ansikten och sedan används det genererade hemliga nyckelsystemet för identifiering efter att intervallet har begränsats. Systemkomplexiteten reduceras med 80% men identifieringsnoggrannheten reduceras inte.

[pt] Nos últimos anos, o aprendizado auto-supervisionado demonstrou desempenho estado da arte em áreas como visão computacional e processamento de linguagem natural. No entanto, ajustar esses modelos para tarefas específicas de classificação, especialmente com dados rotulados, permanece sendo um desafio. Esta dissertação apresenta uma abordagem para ajuste fino de modelos auto-supervisionados usando Redes Neurais Siamesas, aproveitando a função de perda semi-hard triplet loss. Nosso método visa refinar as representações do espaço latente dos modelos auto-supervisionados para melhorar seu desempenho em tarefas posteriores de classificação. O framework proposto emprega Masked Autoencoders para pré-treinamento em um conjunto abrangente de dados de radiografias, seguido de ajuste fino com redes siamesas para separação eficaz de características e melhor classificação. A abordagem é avaliada no conjunto de dados COVIDx 9 para detecção de COVID-19 a partir de radiografias frontais de peito, alcançando uma nova precisão recorde de 98,5 por cento, superando as técnicas tradicionais de ajuste fino e o modelo COVID-Net CRX 3. Os resultados demonstram a eficácia de nosso método em aumentar a utilidade de modelos auto-supervisionados para tarefas complexas de imagem médica. Trabalhos futuros explorarão a escalabilidade dessa abordagem para outros domínios e a integração de funções de perda de espaço de embedding mais sofisticadas. / [en] In recent years, self-supervised learning has demonstrated state-of-theart performance in domains such as computer vision and natural language processing. However, fine-tuning these models for specific classification tasks, particularly with labeled data, remains challenging. This thesis introduces a novel approach to fine-tuning self-supervised models using Siamese Neural Networks, specifically leveraging a semi-hard triplet loss function. Our method aims to refine the latent space representations of self-supervised models to improve their performance on downstream classification tasks. The proposed framework employs Masked Autoencoders for pre-training on a comprehensive radiograph dataset, followed by fine-tuning with Siamese networks for effective feature separation and improved classification. The approach is evaluated on the COVIDx dataset for COVID-19 detection from frontal chest radiographs, achieving a new record accuracy of 98.5 percent, surpassing traditional fine-tuning techniques and COVID-Net CRX 3. The results demonstrate the effectiveness of our method in enhancing the utility of self-supervised models for complex medical imaging tasks. Future work will explore the scalability of this approach to other domains and the integration of more sophisticated embedding-space loss functions.

A common task in data science at internet companies is to develop metrics that capture aspects of the user experience. In this thesis, we are interested in systems of measurement variables without direct causal relations such that covariance is explained by unobserved latent common causes. A framework for modeling the data generating process is given by Neuro-Causal Factor Analysis (NCFA). The graphical model consists of a directed graph with edges pointing from the latent common causes to the measurement variables; its functional relations are approximated with a constrained Variational Auto-Encoder (VAE). We refine the estimation of the graphical model by developing an MCMC algorithm over Bayesian networks from which we read marginal independence relations between the measurement variables. Unlike standard independence testing, the method is guaranteed to yield an identifiable graphical model. Our algorithm is competitive with the benchmark, and it admits additional flexibility via hyperparameters that are natural to the approach. Tuning these parameters yields superior performance over the benchmark. We train the improved NCFA model on Spotify user behavior data. It is competitive with the standard VAE on data reconstruction with the benefit of causal interpretability and model identifiability. We use the learned latent space representation to characterize clusters of Spotify users. Additionally, we train an NCFA model on data from a randomized control trial and observe treatment effects in the latent space. / En vanlig uppgift för en data scientist på ett internetbolag är att utveckla metriker som reflekterar olika aspekter av användarupplevelsen. I denna uppsats är vi intresserade av system av mätvariabler utan direkta kausala relationer, så till vida att kovarians förklaras av latenta gemensamma orsaker. Ett ramverk för att modellera den datagenererande processen ges av Neuro-Causal Factor Analysis (NCFA). Den grafiska modellen består av en riktad graf med kanter som pekar från de latenta orsaksvariablerna till mätvariablerna; funktionssambanden uppskattas med en begränsad Variational Auto-Encoder (VAE). Vi förbättrar uppskattningen av den grafiska modellen genom att utveckla en MCMC algoritm över Bayesianska nätverk från vilka vi läser de obetingade beroendesambanden mellan mätvariablerna. Till skillnad från traditionella oberoendetest så garanterar denna metod en identifierbar grafisk modell. Vår algoritm är konkurrenskraftig jämfört med referensmetoderna, och den tillåter ytterligare flexibilitet via hyperparametrar som är naturliga för metoden. Optimal justering av dessa hyperparametrar resulterar i att vår metod överträffar referensmetoderna. Vi tränar den förbättrade NCFA modellen på data om användarbeteende på Spotify. Modellen är konkurrenskraftig jämfört med en standard VAE vad gäller rekonstruktion av data, och den tillåter dessutom kausal tolkning och identifierbarhet. Vi analyserar representationen av Spotify-användarna i termer av de latenta orsaksvariablerna. Specifikt så karakteriserar vi grupper av liknande användare samt observerar utfall av en randomiserad kontrollerad studie.

The integration of Vision Transformers (ViTs) using Masked Autoencoder pre-training (ViTMAE) into real estate valuation is investigated in this Master’s thesis, addressing the challenge of effectively analyzing visual information from real estate images. This integration aims to enhance the accuracy and efficiency of valuation, a task traditionally dependent on realtor expertise. The research involved developing a model that combines ViTMAE-extracted visual features from real estate images with traditional property data. Focusing on residential properties in Sweden, the study utilized a dataset of images and metadata from online real estate listings. An adapted ViTMAE model, accessed via the Hugging Face library, was trained on the dataset for feature extraction, which was then integrated with metadata to create a comprehensive multimodal valuation model. Results indicate that including ViTMAE-extracted image features improves prediction accuracy in real estate valuation models. The multimodal approach, merging visual and traditional metadata, improved accuracy over metadata-only models. This thesis contributes to real estate valuation by showcasing the potential of advanced image processing techniques in enhancing valuation models. It lays the groundwork for future research in more refined holistic valuation models, incorporating a wider range of factors beyond visual data. / Detta examensarbete undersöker integrationen av Vision Transformers (ViTs) med Masked Autoencoder pre-training (ViTMAE) i bostadsvärdering, genom att addressera utmaningen att effektivt analysera visuell information från bostadsannonser. Denna integration syftar till att förbättra noggrannheten och effektiviteten i fastighetsvärdering, en uppgift som traditionellt är beroende av en fysisk besiktning av mäklare. Arbetet innefattade utvecklingen av en modell som kombinerar bildinformation extraherad med ViTMAE från fastighetsbilder med traditionella fastighetsdata. Med fokus på bostadsfastigheter i Sverige använde studien en databas med bilder och metadata från bostadsannonser. Den anpassade ViTMAE-modellen, tillgänglig via Hugging Face-biblioteket, tränades på denna databas för extraktion av bildinformation, som sedan integrerades med metadata för att skapa en omfattande värderingsmodell. Resultaten indikerar att inklusion av ViTMAE-extraherad bildinformation förbättrar noggranheten av bostadssvärderingsmodeller. Den multimodala metoden, som kombinerar visuell och traditionell metadata, visade en förbättring i noggrannhet jämfört med modeller som endast använder metadata. Denna uppsats bidrar till bostadsvärdering genom att visa på potentialen hos avancerade bildanalys för att förbättra värderingsmodeller. Den lägger grunden för framtida forskning i mer raffinerade holistiska värderingsmodeller som inkluderar ett bredare spektrum av faktorer utöver visuell data.

New regulations are placed on Swedish Telecom Service Providers (TSPs) due to a rising concern for safeguarding network security and privacy in the face of ever-evolving cyber threats. These regulations demand that Swedish telecom companies expand their data security strategies with proactive security measures. Logs, serving as digital footprints in IT infrastructure, play a crucial role in identifying anomalies that could indicate security breaches. Deep Learning (DL) has been used to detect anomalies in logs due to its ability to discern intricate patterns within the data. By leveraging deep learning-based models, it is not only possible to identify anomalies but also to predict and mitigate potential threats within the telecom network. An LSTM autoencoder was implemented to detect anomalies in two separate multivariate temporal log datasets; the BETH cybersecurity dataset, and a Cisco log dataset that was created specifically for this thesis. The empirical results in this thesis show that the LSTM autoencoder reached an ROC AUC of 99.5% for the BETH dataset and 76.6% for the Cisco audit dataset. The use of an additional anomaly detection aid in the Cisco audit dataset let the model reach an ROC AUC of 99.6%. The conclusion that could be drawn from this work was that the systematic approach to developing a deep learning model for anomaly detection in log data was efficient. However, the study’s findings raise crucial considerations regarding the appropriateness of various log data for deep learning models used in anomaly detection. / Nya föreskrifter har införts för svenska telekomtjänsteleverantörer på grund av en ökad angelägenhet av att säkerställa nätverkssäkerhet och integritet inför ständigt föränderliga cyberhot. Dessa föreskrifter kräver att svenska telekomföretag utvidgar sina dataskyddsstrategier med proaktiva säkerhetsåtgärder. Loggar, som fungerar som digitala fotspår inom IT-infrastruktur, spelar en avgörande roll för att identifiera avvikelser som kan tyda på säkerhetsintrång. Djupinlärning har använts för att upptäcka avvikelser i loggar på grund av dess förmåga att urskilja intrikata mönster inom data. Genom att utnyttja modeller baserade på djupinlärning är det inte bara möjligt att identifiera avvikelser utan även att förutsäga samt mildra konsekvenserna av potentiella hot inom telekomnätet. En LSTM-autoencoder implementerades för att upptäcka avvikelser i två separata multivariata tidsserielogguppsättningar; BETH-cybersäkerhetsdatauppsättningen och en Cisco-loggdatauppsättning som skapades specifikt för detta arbete. De empiriska resultaten i denna avhandling visar att LSTM-autoencodern uppnådde en ROC AUC på 99.5% för BETH-datauppsättningen och 76.6% för Cisco-datauppsättningen. Användningen av ett ytterligare avvikelsedetekteringsstöd i Cisco-datauppsättningen möjliggjorde att modellen uppnådde en ROC AUC på 99.6%. Slutsatsen som kunde dras från detta arbete var att den systematiska metoden för att utveckla en djupinlärningsmodell för avvikelsedetektering i loggdata var effektiv. Dock väcker studiens resultat kritiska överväganden angående lämpligheten av olika loggdata för djupinlärningsmodeller som används för avvikelsedetektering.

Financial trading data presents a unique challenge for anomaly detection due to its high dimensionality and often lack of labelled anomalous examples. Nevertheless, it is of great interest for financial institutions to gain insight into potential trading activities that might lead to financial losses and reputational damage. Given the complexity and unlabelled nature of this financial data, deep learning models such as the Transformer model are particularly suited for this task.   This work investigates the application of a Transformer-based autoencoder for anomaly detection in unlabelled financial transaction data with sequential characteristics. To assess the model's ability to detect anomalies and analyse the effects of class imbalance, synthetic anomalies are injected into the dataset. This creates a controlled environment where the model's performance can be evaluated but also the affects of imbalance can be investigated.    Two approaches are particularly explored for anomaly detection purposes: an unsupervised approach and a semi-supervised approach that explicitly leverages the presence of anomalies in the training data. Experiments suggest that while the unsupervised approach can detect anomalies with distinctive features, its performance suffers when anomalies are included in the training data since the model tends to reconstruct them. Conversely, the semi-supervised approach effectively addresses this limitation, demonstrating a clear advantage in the presence of class imbalance. While synthetic anomalies enable controlled evaluation and class imbalance analysis, generalizability to real-world financial data requires true anomalies.

<div> <p>Sound and vibration analysis is a prominent tool used for scientific investigations in various fields such as structural model identification or dynamic behavior studies. In manufacturing fields, the vibration signals collected through commercial sensors are utilized to monitor machine health, for sustainable and cost-effective manufacturing.</p> <p> Recently, the development of commercial sensors and computing environments have encouraged researchers to combine gathered data and Machine Learning (ML) techniques, which have been proven to be efficient for categorical classification problems. These discriminative algorithms have been successfully implemented in monitoring problems in factories, by simulating faulty situations. However, it is difficult to identify all the sources of anomalies in a real environment. </p> <p>In this paper, a Neural Network (NN) application on a KUKA KR6 robot arm is introduced, as a solution for the limitations described above. Specifically, the autoencoder architecture was implemented for anomaly detection, which does not require the predefinition of faulty signals in the training process. In addition, stethoscopes were utilized as alternative sensing tools as they are easy to handle, and they provide a cost-effective monitoring solution. To simulate the normal and abnormal conditions, different load levels were assigned at the end of the robot arm according to the load capacity. Sound signals were recorded from joints of the robot arm, then meaningful features were extracted from spectrograms of the sound signals. The features were utilized to train and test autoencoders. During the autoencoder process, reconstruction errors (REs) between the autoencoder’s input and output were computed. Since autoencoders were trained only with features corresponding to normal conditions, RE values corresponding to abnormal features tend to be higher than those of normal features. In each autoencoder, distributions of the RE values were compared to set a threshold, which distinguishes abnormal states from the normal states. As a result, it is suggested that the threshold of RE values can be utilized to determine the condition of the robot arm.</p> </div> <br>

Gene expression analysis provides genome-wide insights into the transcriptional activity of a cell. One of the first computational steps in exploration and analysis of the gene expression data is clustering. With a number of standard clustering methods routinely used, most of the methods do not take prior biological information into account. Here, we propose a new approach for gene expression clustering analysis. The approach benefits from a new deep learning architecture, Robust Autoencoder, which provides a more accurate high-level representation of the feature sets, and from incorporating prior system-wide biological information into the clustering process. We tested our approach on two gene expression datasets and compared the performance with two widely used clustering methods, hierarchical clustering and k-means, and with a recent deep learning clustering approach. Our approach outperformed all other clustering methods on the labeled yeast gene expression dataset. Furthermore, we showed that it is better in identifying the functionally common clusters than k-means on the unlabeled human gene expression dataset. The results demonstrate that our new deep learning architecture can generalize well the specific properties of gene expression profiles. Furthermore, the results confirm our hypothesis that the prior biological network knowledge is helpful in the gene expression clustering.

Deep learning is usually applied to static datasets. If used for classification based on data streams, it is not easy to take into account a non-stationarity. This thesis presents work in progress on a new method for online deep classifi- cation learning in data streams with slow or moderate drift, highly relevant for the application domain of malware detection. The method uses a combination of multilayer perceptron and variational autoencoder to achieve constant mem- ory consumption by encoding past data to a generative model. This can make online learning of neural networks more accessible for independent adaptive sys- tems with limited memory. First results for real-world malware stream data are presented, and they look promising. 1

Predictive Maintenance (PM) has been increasingly adopted in the Automotive industry, in the recent decades along with conventional approaches such as the Preventive Maintenance and Diagnostic/Corrective Maintenance, since it provides many advantages to estimate the failure before the actual occurrence proactively, and also being adaptive to the present status of the vehicle, in turn allowing flexible maintenance schedules for efficient repair or replacing of faulty components. PM necessitates the storage and analysis of large amounts of sensor data. This requirement can be a challenge in deploying this method on-board the vehicles due to the limited storage and computational power on the hardware of the vehicle. Hence, this thesis seeks to obtain low dimensional descriptive features from high dimensional data using Representation Learning. This low dimensional representation will be used for predicting vehicle faults, specifically Turbocharger related failures. Since the Logged Vehicle Data (LVD) was base on all the data utilized in this thesis, it allowed for the evaluation of large populations of trucks without requiring additional measuring devices and facilities. The gradual degradation methodology is considered for describing vehicle condition, which allows for modeling the malfunction/ failure as a continuous process rather than a discrete flip from healthy to an unhealthy state. This approach eliminates the challenge of data imbalance of healthy and unhealthy samples. Two important hypotheses are presented. Firstly, Parallel StackedClassical Autoencoders would produce better representations com-pared to individual Autoencoders. Secondly, employing Learned Em-beddings on Categorical Variables would improve the performance of the Dimensionality reduction. Based on these hypotheses, a model architecture is proposed and is developed on the LVD. The model is shown to achieve good performance, and in close standards to the previous state-of-the-art research. This thesis, finally, illustrates the potential to apply parallel stacked architectures with Learned Embeddings for the Categorical features, and a combination of feature selection and extraction for numerical features, to predict the Remaining Useful Life (RUL) of a vehicle, in the context of the Turbocharger. A performance improvement of 21.68% with respect to the Mean Absolute Error (MAE) loss with an 80.42% reduction in the size of data was observed.