No description available.

Polyphonic or multiple music instrument detection is a difficult problem compared to detecting single or solo instruments in an audio recording. As music is time series data it be can modelled using sequence learning methods within deep learning. Recently, temporal convolutional networks (TCN) have shown to outperform conventional recurrent neural networks (RNN) on various sequence modelling tasks. Though there have been significant improvements in deep learning methods, data scarcity becomes a problem in training large scale models. Weakly labelled data is an alternative where a clip is annotated for presence or absence of instruments without specifying the times at which an instrument is sounding. This study investigates how TCN model compares to a Long Short-Term Memory (LSTM) model while trained on weakly labelled dataset. The results showed successful training of both models along with generalisation on a separate dataset. The comparison showed that TCN performed better than LSTM, but only marginally. Therefore, from the experiments carried out it could not be explicitly concluded if TCN is convincingly a better choice over LSTM in the context of instrument detection, but definitely a strong alternative. / Polyfonisk eller multipel musikinstrumentdetektering är ett svårt problem jämfört med att detektera enstaka eller soloinstrument i en ljudinspelning. Eftersom musik är tidsseriedata kan den modelleras med hjälp av sekvensinlärningsmetoder inom djup inlärning. Nyligen har ’Temporal Convolutional Network’ (TCN) visat sig överträffa konventionella ’Recurrent Neural Network’ (RNN) på flertalet sekvensmodelleringsuppgifter. Även om det har skett betydande förbättringar i metoder för djup inlärning, blir dataknapphet ett problem vid utbildning av storskaliga modeller. Svagt märkta data är ett alternativ där ett klipp kommenteras för närvaro av frånvaro av instrument utan att ange de tidpunkter då ett instrument låter. Denna studie undersöker hur TCN-modellen jämförs med en ’Long Short-Term Memory’ (LSTM) -modell medan den tränas i svagt märkta datasätt. Resultaten visade framgångsrik utbildning av båda modellerna tillsammans med generalisering i en separat datasats. Jämförelsen visade att TCN presterade bättre än LSTM, men endast marginellt. Därför kan man från de genomförda experimenten inte uttryckligen dra slutsatsen om TCN övertygande är ett bättre val jämfört med LSTM i samband med instrumentdetektering, men definitivt ett starkt alternativ.

The success of recognizing periodic actions in single-person-simple-background datasets, such as Weizmann and KTH, has created a need for more complex datasets to push the performance of action recognition systems. In this work, we create a new synthetic action dataset and use it to highlight weaknesses in current recognition systems. Experiments show that introducing background complexity to action video sequences causes a significant degradation in recognition performance. Moreover, this degradation cannot be fixed by fine-tuning system parameters or by selecting better feature points. Instead, we show that the problem lies in the spatio-temporal cuboid volume extracted from the interest point locations. Having identified the problem, we show how improved results can be achieved by simple modifications to the cuboids. For the above method however, one requires near-perfect localization of the action within a video sequence. To achieve this objective, we present a two stage weakly supervised probabilistic model for simultaneous localization and recognition of actions in videos. Different from previous approaches, our method is novel in that it (1) eliminates the need for manual annotations for the training procedure and (2) does not require any human detection or tracking in the classification stage. The first stage of our framework is a probabilistic action localization model which extracts the most promising sub-windows in a video sequence where an action can take place. We use a non-linear classifier in the second stage of our framework for the final classification task. We show the effectiveness of our proposed model on two well known real-world datasets: UCF Sports and UCF11 datasets. iii Another application of the weakly supervised probablistic model proposed above is in the gaming environment. An important aspect in designing interactive, action-based interfaces is reliably recognizing actions with minimal latency. High latency causes the system’s feedback to lag behind and thus significantly degrade the interactivity of the user experience. With slight modification to the weakly supervised probablistic model we proposed for action localization, we show how it can be used for reducing latency when recognizing actions in Human Computer Interaction (HCI) environments. This latency-aware learning formulation trains a logistic regression-based classifier that automatically determines distinctive canonical poses from the data and uses these to robustly recognize actions in the presence of ambiguous poses. We introduce a novel (publicly released) dataset for the purpose of our experiments. Comparisons of our method against both a Bag of Words and a Conditional Random Field (CRF) classifier show improved recognition performance for both pre-segmented and online classification tasks.

As imaging systems become ubiquitous, the ability to recognize human actions is becoming increasingly important. Just as in the object detection and recognition literature, action recognition can be roughly divided into classification tasks, where the goal is to classify a video according to the action depicted in the video, and detection tasks, where the goal is to detect and localize a human performing a particular action. A growing literature is demonstrating the benefits of localizing discriminative sub-regions of images and videos when performing recognition tasks. In this thesis, we address the action detection and recognition problems. Action detection in video is a particularly difficult problem because actions must not only be recognized correctly, but must also be localized in the 3D spatio-temporal volume. We introduce a technique that transforms the 3D localization problem into a series of 2D detection tasks. This is accomplished by dividing the video into overlapping segments, then representing each segment with a 2D video projection. The advantage of the 2D projection is that it makes it convenient to apply the best techniques from object detection to the action detection problem. We also introduce a novel, straightforward method for searching the 2D projections to localize actions, termed TwoPoint Subwindow Search (TPSS). Finally, we show how to connect the local detections in time using a chaining algorithm to identify the entire extent of the action. Our experiments show that video projection outperforms the latest results on action detection in a direct comparison. Second, we present a probabilistic model learning to identify discriminative regions in videos from weakly-supervised data where each video clip is only assigned a label describing what action is present in the frame or clip. While our first system requires every action to be manually outlined in every frame of the video, this second system only requires that the video be given a single highlevel tag. From this data, the system is able to identify discriminative regions that correspond well iii to the regions containing the actual actions. Our experiments on both the MSR Action Dataset II and UCF Sports Dataset show that the localizations produced by this weakly supervised system are comparable in quality to localizations produced by systems that require each frame to be manually annotated. This system is able to detect actions in both 1) non-temporally segmented action videos and 2) recognition tasks where a single label is assigned to the clip. We also demonstrate the action recognition performance of our method on two complex datasets, i.e. HMDB and UCF101. Third, we extend our weakly-supervised framework by replacing the recognition stage with a twostage neural network and apply dropout for preventing overfitting of the parameters on the training data. Dropout technique has been recently introduced to prevent overfitting of the parameters in deep neural networks and it has been applied successfully to object recognition problem. To our knowledge, this is the first system using dropout for action recognition problem. We demonstrate that using dropout improves the action recognition accuracies on HMDB and UCF101 datasets.

No description available.

No description available.

This thesis examines the effectiveness of Graph Neural Networks (GNNs) in detecting money laundering activities using transaction data with unreliable labels. It analyses how weakly supervised learning, specifically with GNNs, manages the challenges posed by incomplete and inaccurate labels in anti-money laundering (AML) detection. The thesis utilizes simulated transaction data to compare the performance of GNNs against statistical models. This was done by generating various datasets with the AMLSim tool, and evaluating the node classification performance of different statistical machine learning models and GNNs. The findings indicate that GNNs, due to their ability to find relationships in graph structures, demonstrate superior performance in scenarios with incomplete and inaccurate labels. The findings also indicate that inaccurate positive labels has a great negative effect on the performance, showing the label importance of money launderers in graph data. This research provides possible improvements for anti-money laundering detection by employing GNNs to manage challenges in real-world data.

L'objectif de c e mémoire est d'apporter une contribution à la modélisation et la simulation numérique de la convection thermosolutale de mélanges binaires de gaz parfaits contenus dans des cavités. Un modèle a été élaboré en se basant sur l'approximation de faible compressibilité. Le premier chapitre précise la démarche suivie dans la modélisation et une formulation originale en est déduite afin de traiter les différents types de conditions aux limites et de conditions de références hydrostatiques analysés dans le mémoire. Les variations de masse volumique sont déduites de la loi des gaz parfaits et la pression thermodynamique est calculée à partir de la conservation de la masse totale. La méthode numérique repose sur la méthode des volumes finis mise en uvre sur des maillages décalés. Le couplage vitesse-pression est traité par un nouvel algorithme dont l'efficacité est discutée en détail. La démarche numérique est validée via des comparaisons avec des solutions de références, en régime stationnaire comme en régime transitoire pour des écoulements transitionnels. Dans la seconde partie du mémoire, on considère d'abord la convection thermosolutale dans une cavité rectangulaire verticale dans le cas où les écoulements sont induits par des gradients horizontaux de température et de concentration. On discute en particulier les limites de l'approximation d'extrême dilution. La condensation de vapeur d'eau et l'évaporation d'un film d'eau liquide sur les parois d'une cavité sont ensuite étudiées en régime transitoire. Ces changements de phase surfaciques sont associés à la convection naturelle dans une cavité dont les températures des quatre parois varient au cours du temps / The aim of this dissertation is at modeling and numerically simulating thermosolutal convection within cavities filled by binary gas mixtures of ideal gases. A new problem formulation, based on the weakly compressible approximation, has been derived in order to account for the changes in density as well as thermodynamic pressure. The ideal gas law and global conservation of mass are invoked for the predictions of density field and thermodynamic pressure. The first part of the manuscript is devoted to the mathematical derivation of the governing equations and boundary conditions, numerical procedure implemented and, checks of the accuracy of the results through deeply comparisons with updated benchmark solutions. The emphasis is put on the efficiency of algorithm used for solving the pressure-velocity coupling for unsteady, transitional flow regimes. Thermosolutal convection without phase changes at the cavity surfaces is first considered in the second part of the manuscript. We investigated the case of vertical cavities with horizontal gradients of concentration and temperature. The results clearly show that the dilute approximation fails to be valid as soon as the maximum concentration difference exceeds a critical value function of the flow parameters. Surface condensation of water vapor or evaporation of liquid water film in vertical cavities is then considered. The specific thermal boundary conditions of uniform but time-varying temperatures of the four walls are considered. Periodic variations of the flow, temperature and relative humidity fields are discussed in detail. The evolutions of thicknesses of the water film over the four walls are also predicted

In the framework of quantum theory the characteristics of synchrotron radiation (SR) are considered. In order to simplify theoretical description the process of radiation is restricted to single-photon emission. For arbitrary quantum transitions the spectral-angular distributions of SR power are given in exact analytical form. Scalar particles (bosons) and particles with spin $\\hbar/2$ (electrons) are treated separately. Special attention is given to the particular transitions, namely, to the transitions to first excited and ground states. It is shown that the components of linear polarization of radiation from electron switch places due to the orientation of spin when the electron jumps to the ground state. This fact can be considered an analytical proof for the presence of $\\pi$-component of quantum radiation in the plane of motion. The radiation emitted from weakly excited particles is thoroughly analysed. To describe the evolution of the profiles of angular distributions various functions are introduced both for two- and three-level systems. For quantum transitions from the first excited state to the ground state the comparative analysis of radiation from bosons and electrons is performed, which helps to estimate the influence of spin and its direction on the characteristics of radiation. The radiation from unpolarized electron is considered separately. Tracking the behavior of effective angles allows to discover the inconsistency of well-known classical conclusion about the concentration of total (summed over spectrum) ultrarelativistic radiation in the plane of motion. It is shown that the effective angles of quantum radiation tend to finite values and do not vanish in ultrarelativistic region. A brief review of classical theory includes an introduction of the new concept, $n$-part of spectrum. In order to find an adequate classical analogue for the radiation from weakly excited particles, the idea to reduce classical spectrum was developed. It turns out that the characteristics of radiation calculated for reduced classical spectrum stay in good quantitative and qualitative agreement with their quantum analogues, at least for single-harmonic and two-harmonic quantum spectra, and classical theory of a reduced spectrum can be claimed representational in this sense. The evolution of maximum in radiation spectrum is considered in separate chapter. A well-known approximation obtained for critical frequency in the framework of classical theory is invalid when quantum corrections enter the picture. But there appears a possibility to find the conditions for the maximum to shift to the highest harmonic of finite quantum spectrum. It is shown that the shifts occur successively starting with primary harmonic in non-relativistic case, and this result remains valid independently of spin. For a scalar particle there exists a fixed set of numbers, which are the critical values of external field, such that the shift of radiation maximum in the spectrum of boson can only happen when the intensity of external field is greater than certain critical value related to corresponding harmonic. If this condition is not satisfied, the position of maximum remains unchanged. It turns out that the presence of spin perturbs this picture, so that the critical values of field intensity depend on the number of initial level. / Consideramos as características da radiação sincrotron (RS) no âmbito da teoria quântica. Para simplificar a descrição teórica do processo de radiação restringimos à consideração da emissão de único fóton. Para transições quânticas arbitrárias, as distribuições espectrais e angulares da potência da RS são dadas de forma analítica exata. Tratamos separadamente partículas escalares (bósons) e com spin ½ (elétrons). Atenção especial é dada às transições particulares, a saber, as transições ao primeiro estado excitado e estado fundamental. É mostrado que os componentes de polarização linear da radiação de elétron se trocam em relação à orientação de spin quando o elétron passa para o estado fundamental. Este fato pode ser considerado como uma comprovação analítica para a presença de -componente da radiação quântica no plano de movimento. Analisamos minuciosamente a radiação emitida pela partícula fracamente excitada. Várias funções são introduzidas para descrever a evolução dos perfis de distribuições angulares para sistemas de dois e três níveis. Para transições quânticas do primeiro estado excitado ao estado fundamental a análise comparativa da radiação de bósons e elétrons é realizada, e isso ajuda à estimar a influência de spin e sua direção sobre as características da RS. A radiação de elétrons não polarizados é considerada separadamente. Observando o comportamento dos ângulos efetivos, é fácil perceber a inconsistência da conclusão clássica bem conhecida sobre a concentração de radiação ultra-relativista total no plano do movimento. Mostramos que os ângulos efetivos da radiação quântica tendem aos valores finitos e não desaparecem na região ultrarelativista. Uma revisão breve da teoria clássica inclui a introdução do conceito novo, isto é a n-parte do espectro. A fim de encontrar um análogo clássico adequado para a radiação das partículas fracamente excitados, a ideia de reduzir o espectro clássico foi desenvolvida. Constatamos que as características da radiação calculadas para o espectro clássico reduzido permanecem em boa concordância, tanto quantitativa quanto qualitativa, com os seus análogos quânticos, pelo menos no que diz respeito aos espectros quânticos de uma ou duas harmônicas. Neste sentido, a teoria clássica do espectro reduzido pode ser chamada de representativa. A evolução do máximo no espectro da radiação é considerada em capítulo separado. A aproximação, comumente considerada na teoria classica para frequência crítica, é inválida quando as correções quânticas entram em cena. Mas existe uma possibilidade de encontrar as condições para o máximo transferir-se à harmônica maior do espectro quântico. É mostrado que as transferências ocorrem sucessivamente, comecando com a harmônica principal no caso não relativístico, e este resultado permanece válido, independentemente de spin. Para uma partícula escalar existe um conjunto fixo dos valores críticos do campo externo, de tal modo que a transferência do máximo da radiação entre duas harmônicas específicas pode acontecer somente quando a intensidade do campo externo é maior do que o valor crítico associado com essas harmônicas. Se essa condição não for satisfeita, a posição do máximo permanece inalterada. Verificamos que a presença de spin perturba esta condição, no caso do elétron os valores críticos da intensidade do campo dependem de número do nível inicial.

Neste trabalho apresentamos uma versão do Teorema de Poincaré-Bendixson para campos vetoriais contínuos na garrafa de Klein. Como conseqüência, mostramos que a garrafa de Klein não possui campo vetorial contínuo com trajetória injetiva recorrente / We present a version of the Poincaré-Bendixson Theorem on the Klein bottle for continuous vector fields. As a consequence, we obtain the fact that the Klein bottle does not admit continuous vector fields having a recurrent injective trajectory