With the inception of music streaming and media content delivery platforms, there has been a tremendous increase in the music available on the internet and the metadata associated with it. In this study, we address the problem of violin artist identification, which tries to classify the performing artist based on the learned features. Even though numerous previous works studied the problem in detail and developed features and deep learning models that can be used, an interesting fact was that most studies focused on artist identification in western popular music and less on Indian classical music. For the same reason, there was no standardized dataset for this purpose. Hence, we curated a new dataset consisting of audio recordings from 6 renowned South Indian Carnatic violin artists. In this study, we explore the use of log-Mel-spectrogram feature and the embeddings generated by a pre-learned VGGish network on a Convolutional Neural Network and Convolutional Recurrent Neural Network Model. From the experiments, we observe that the Convolutional Recurrent Neural Network model trained using the log-Mel-spectrogram feature gave the optimal performance with a classification accuracy of 71.70%. / Med starten av plattformar för musikströmning och leverans av mediainnehåll har det skett en enorm ökning av musiken tillgänglig på internet och den metadata som är associerad med den. I denna studie tar vi upp problemet med fiolkonstnärsidentifikation, som försöker klassificera den utövande konstnären utifrån de inlärda dragen. Även om många tidigare verk studerade problemet i detalj och utvecklade funktioner och modeller för djupinlärning som kan användas, var ett intressant faktum att de flesta studier fokuserade på artistidentifiering i västerländsk populärmusik och mindre på indisk klassisk musik. Av samma anledning fanns det ingen standardiserad datauppsättning för detta ändamål. Därför kurerade vi en ny datauppsättning bestående av ljudinspelningar från 6 kända sydindiska karnatiska violinkonstnärer. I den här studien utforskar vi användningen av log-Melspektrogramfunktionen och inbäddningarna som genereras av ett förinlärt VGGishnätverk på ett Convolutional Neural Network och Convolutional Recurrent Neural Network Model. Från experimenten observerar vi att modellen Convolutional Recurrent Neural Network tränad med hjälp av log-Mel-spektrogramfunktionen gav optimal prestanda med en klassificeringsnoggrannhet på 71,70%.

Record linkage, in a genealogical context, is the process of identifying individuals from multiple sources which refer to the same real-world entity. Current solutions focus on the individuals in question and on complex rules developed by human experts. Genealogical databases are highly-structured with relationships existing between the individuals and other instances. These relationships can be utilized and human involvement greatly minimized by using a filtered structured neural network. These neural networks, using traditional back-propagation methods, are biased in a way to make the network human readable. The results show an increase in precision and recall when pedigree data is available and used.

No description available.

How humans perceive image quality in general (e.g. aesthetics, resolution, lighting) and what constitutes to an image of good quality and aesthetics, is an interesting topic for neural networks to tackle. Lately several different aspects have been investigated in theory to try and build models to accurately predict the quality of an image. The problem is investigated in this thesis by performing an experiment and testing a method that computes the aesthetic value of an image and ranks a number of images according to their aesthetic value. The rankings are then compared to human rankings of the same images to determine if there is a correlation between the two. From the results the null hypothesis (that the correlation is just a coincidence) cannot be rejected but in 92% of the cases there were at least some agreement (same image, same position) between the algorithm and the human ranking. There are multiple similarities between the human rankings and the algorithm rankings which reveals a good potential to further investigate the problem in future. / Hur bildkvalité generellt sett uppfattas av människor (så som estetik, upplösning, ljussättning) och vad som gör att en bild kan klassificeras som en bild av hög kvalité är ett intressant ämnesområde för neurala nätverk. Den senaste forskningen visar att det finns flera metoder och aspekter som påverkar hur kvalitativ en bild är i teorin. Problemet undersöks i denna rapport med ett experiment och metoden som testas är en metod som kan sätta ett estetiskt värde på en bild. Tillsammans med flera olika bilder kunde en rangordning skapas. Denna rangordning jämfördes sedan med en rangordning skapad av människor. Från resultaten så kan inte nollhypotesen förkastas (att korrelationen är ett sammanträffande) men att i 92% av fallen så finns det åtminstånde en likhet mellan algoritmens rangordning och en människas rangordning. Det finns flera likheter mellan de mänskliga rangordningarna och algoritmens rangordning vilket visar en god potential för forsatt undersökning i framtiden.

Although generative adversarial networks (GANs) have achieved state-of-the-art results in generating realistic look- ing images, they are often parameterized by neural net- works with relatively few learnable weights compared to those that are used for discriminative tasks. We argue that this is suboptimal in a generative setting where data is of- ten entangled in high dimensional space and models are ex- pected to benefit from high expressive power. Additionally, in a generative setting, a model often needs to extrapo- late missing information from low dimensional latent space when generating data samples while in a typical discrimina- tive task, the model only needs to extract lower dimensional features from high dimensional space. We evaluate different architectures for GANs with varying model capacities using shortcut connections in order to study the impacts of the capacity on training stability and sample quality. We show that while training tends to oscillate and not benefit from additional capacity of naively stacked layers, GANs are ca- pable of generating samples with higher quality, specifically for images, samples of higher visual fidelity given proper regularization and careful balancing. / Trots att Generative Adversarial Networks (GAN) har lyckats generera realistiska bilder består de än idag av neurala nätverk som är parametriserade med relativt få tränbara vikter jämfört med neurala nätverk som används för klassificering. Vi tror att en sådan modell är suboptimal vad gäller generering av högdimensionell och komplicerad data och anser att modeller med högre kapaciteter bör ge bättre estimeringar. Dessutom, i en generativ uppgift så förväntas en modell kunna extrapolera information från lägre till högre dimensioner medan i en klassificeringsuppgift så behöver modellen endast att extrahera lågdimensionell information från högdimensionell data. Vi evaluerar ett flertal GAN med varierande kapaciteter genom att använda shortcut connections för att studera hur kapaciteten påverkar träningsstabiliteten, samt kvaliteten av de genererade datapunkterna. Resultaten visar att träningen blir mindre stabil för modeller som fått högre kapaciteter genom naivt tillsatta lager men visar samtidigt att datapunkternas kvaliteter kan öka, specifikt för bilder, bilder med hög visuell fidelitet. Detta åstadkoms med hjälp utav regularisering och noggrann balansering.

A new scheme for training Rectified Linear Unit (ReLU) based feedforward neural networks is examined in this thesis. The project starts with the row-by-row updating strategy designed for Single-hidden Layer Feedforward neural Networks (SLFNs). This strategy exploits the properties held by ReLUs and optimizes each row in the input weight matrix individually, under the common optimization scheme. Then the Direct Updating Strategy (DUS), which has two different versions: Vector-Based Method (VBM) and Matrix-Based Method (MBM), is proposed to optimize the input weight matrix as a whole. Finally DUS is extended to Multi-hidden Layer Feedforward neural Networks (MLFNs). Since the extension, for general ReLU-based MLFNs, faces an initialization dilemma, a special structure MLFN is presented. Verification experiments are conducted on six benchmark multi-class classification datasets. The results confirm that MBM algorithm for SLFNs improves the performance of neural networks, compared to its competitor, regularized extreme learning machine. For most datasets involved, MLFNs with the proposed special structure perform better when adding extra hidden layers. / Ett nytt schema för träning av rektifierad linjär enhet (ReLU)-baserade och framkopplade neurala nätverk undersöks i denna avhandling. Projektet börjar med en rad-för-rad-uppdateringsstrategi designad för framkopplade neurala nätverk med ett dolt lager (SLFNs). Denna strategi utnyttjar egenskaper i ReLUs och optimerar varje rad i inmatningsviktmatrisen individuellt, enligt en gemensam optimeringsmetod. Därefter föreslås den direkta uppdateringsstrategin (DUS), som har två olika versioner: vektorbaserad metod (VBM) respektive matrisbaserad metod (MBM), för att optimera ingångsviktmatrisen som helhet. Slutli- gen utvidgas DUS till framkopplade neurala nätverk med flera lager (MLFN). Eftersom utvidgningen för generella ReLU-baserade MLFN står inför ett initieringsdilemma presenteras därför en MLFN med en speciell struktur. Verifieringsexperiment utförs på sex datamängder för klassificering av flera klasser. Resultaten bekräftar att MBM-algoritmen för SLFN förbättrar prestanda hos neurala nätverk, jämfört med konkurrenten, den regulariserade extrema inlärningsmaskinen. För de flesta använda dataset, fungerar MLFNs med den föreslagna speciella strukturen bättre när man lägger till extra dolda lager.

Neural Networks have been applied in numeral broad categories of work. Such as classification, data processing, robotics, systemcontrol e.t.c. This thesis compares using traditional methods of the routing process in chip circuit design to using a Neural Network trained with evolution. Constructing and evaluating a chip design is a complicated thing, where a lot of variables have to be accounted for and therefore a simplified evaluation and design process is used in order to train the network and compare the results. This was done by constructing simple test cases and running the algorithms BFS, A*Star and the neural network and comparing the paths each algorithm found using a fitness function. The results were that BFS and A*Star both performed better on complex circuits, but the neural network was able to create better paths on very small and niche circuits. The conclusion of the study is that the neural network approach is not able to compete with the standard industry methods of the routing process, but we do not exclude the possibility that with a better designed Fitness function, this could be possible. / Neurala Nätverk används i flertal breda kategorier av arbete. Såsom klassificering, databehandling, robotik, systemkontroll e.t.c. Denna avhandling jämför traditionella metoder för routingprocessen i chip-kretsdesign med att använda ett neuralt nätverk utbildat med evolution. Att konstruera och utvärdera en chipdesign är en komplicerad sak, där många variabler måste tas hänsyn till och därför används en förenklad utvärderings- och designprocess för att träna nätverket och jämföra resultaten. Detta gjordes genom att konstruera enkla testfall och köra algoritmerna BFS, A * Star och det neurala nätverket och jämföra de sökvägar som varje algoritm fann med hjälp av en så kallad Fitness-funktion. Resultaten var att BFS och A * Star både fungerade bättre på komplexa kretsar, men det neurala nätverket kunde skapa bättre vägar på mycket små och nischade kretsar. Slutsatsen av studien är att det neurala nätverkssättet inte kan konkurrera med routingprocessens standardindustrimetoder, men vi utesluter inte möjligheten att med en bättre utformad Fitness-funktion skulle detta vara möjligt.

Homography estimation is a fundamental task in many computer vision applications, but many techniques for estimation rely on complicated feature extraction pipelines. We extend research in direct homography estimation (i.e. without explicit feature extraction) by implementing a convolutional network capable of estimating homographies. Previous work in deep learning based homography estimation calculates homographies between pairs of images, whereas our network takes single image input and registers it to a reference view where no image data is available. The application of the work is registering frames from American football video to a top-down view of the field. Our model manages to register frames in a test set with an average corner error equivalent to less than 2 yards. / Homografiuppskattning är ett förkrav för många problem inom datorseende, men många tekniker för att uppskatta homografier bygger på komplicerade processer för att extrahera särdrag mellan bilderna. Vi bygger på tidigare forskning inom direkt homografiuppskattning (alltså, utan att explicit extrahera särdrag) genom att  implementera ett Convolutional Neural Network (CNN) kapabelt av att direkt uppskatta homografier. Arbetet tillämpas för att registrera bilder från video av amerikansk fotball till en referensvy av fotbollsplanen. Vår modell registrerar bildramer från ett testset till referensvyn med ett snittfel i bildens hörn ekvivalent med knappt 2 yards.

The primary aim of this study was to fill a gap in the literature in understanding maximal BH in untrained, non-divers by generating a predictive numerical model for PAO2 and PACO2 throughout BH. There have been little to no previous attempts at explicitly characterizing the influence of impermanent personal or environmental factors on PAO2 or PACO2 at BH breakpoint. The metabolic human consumption of O2 and production of CO2 as measured through alveolar partial pressures was observed over time during a voluntary maximum breath-hold for 18 members of the general population. The coefficient of determination was used to determine accuracy of the model in fitting participants’ BH data from this experiment. The volume of the last inhalation pre-BH, time to subjective breakpoint, and breath-to-breath calculated respiratory exchange ratio (RER) were identified as the most influential combination of key variables that improved PAO2 model fit (from R2 = 0.8591 to R2 = 0.8840). Clustering methods coupled with two sample t-tests or ANOVA were then used to identify survey responses most correlated to inter-BH similarities. These were barometric pressure, age, height, weight, resting HR, smoker/ freediver/scuba experience, and weekly frequency of engaging in physical exercise. The model was validated on testing data from an experienced free-diver, from non-rebreathing trials of a sample of the participants, and from simulated dives of 5 participants from research in the Environmental Physiology Department of Karolinska in 1994 [1]. It has been suggested that the presented model can be a valuable tool in developing safer free diving practices. Furthermore, interesting trends in continuous HR, starting PACO2 values, and O2 consumption were observed and analyzed using statistical analysis. Findings were discussed with connection to the underlying physiological principles that might explain the results obtained.

It is often that images generated from Synthetic Aperture Radar (SAR) are noisy, distorted, or incomplete pictures of a target or target region. As the goal for most SAR research pertains to automatic target recognition (ATR), extensive filtering and image processing is required in order to extract the features necessary to carry out ATR. This thesis investigates the use of Artificial Neural Networks (ANNs) in order to improve upon the feature extraction process by laying the foundation for ANN SAR ATR algorithms and programs. The first technique investigated is that of an ANN edge detector designed to be invariant to multiplicative speckle noise. The algorithm designed uses the Back Propagation (BP) algorithm to teach a multi-layer perceptron network to detect edges. In order to do so, several parameters within a Sliding Window (SW), are calculated as the inputs to the ANN. The ANN then outputs an edge map that includes the outer edge features of the target as well as some internal edge features. The next technique that is examined is a pattern recognition and target reconstruction algorithm based off of the associative memory ANN known as the Hopfield Network (HN). For this version of the HN, the network is trained with a collection of varying geometric shapes. The output of the network is a nearest-fit representation of the incomplete image data input. Because of the versatility of this program, it is also able to reconstruct incomplete 3D models determined from SAR data. The final technique investigated is an automatic rotation procedure to detect the change in perspective relative to the platform. This type of detection can prove useful if used for target tracking or 3D modeling where the direction vector or relative angle of the target is a desired piece of information.