Makams are well-defined modes of classical Turkish music. They can be taken as the Turkish music counterparts of Western music tonal structures at a certain level. Nevertheless, makams have additional features such as the usage of specific notes resulting from their different architecture and the special use of scales (i.e. progression). The main goal of this study is to construct a platform for identifying makams through a computer program by proposing a machine learning mechanism. There are restrictionson the mechanism related to the characteristics of the task. Such a mechanism should represent real-time sequential input with continuous values, should handle possible errors in this input and show immediate learning with limited data. These restrictions are valid and necessary for an analogy with the act of listening to music. A Boltzmann machine, modified for this purpose is designed, implemented and used in this study as this learning mechanism. Two characteristics of this study define its significance. First, this study is on the structural features of makams of classical Turkish music. Second, the identifying mechanism is a Boltzmann machine having a different schema than statistical identification tasks in tonality induction.

Submitted by Daniel Felipe Silva Santos null (danielfssantos@yahoo.com.br) on 2017-08-29T19:56:20Z No. of bitstreams: 1 ReconhecedorDeVeiculos2D-DBM.pdf: 3800862 bytes, checksum: 46f12ff55f4e0680833b9b1b184ad505 (MD5) / Approved for entry into archive by Luiz Galeffi (luizgaleffi@gmail.com) on 2017-08-29T20:19:13Z (GMT) No. of bitstreams: 1 santos_dfs_me_sjrp.pdf: 3800862 bytes, checksum: 46f12ff55f4e0680833b9b1b184ad505 (MD5) / Made available in DSpace on 2017-08-29T20:19:13Z (GMT). No. of bitstreams: 1 santos_dfs_me_sjrp.pdf: 3800862 bytes, checksum: 46f12ff55f4e0680833b9b1b184ad505 (MD5) Previous issue date: 2017-08-14 / Neste trabalho é proposto um método para reconhecer veículos em imagens coloridas baseado em uma rede neural Perceptron Multicamadas pré-treinada por meio de técnicas de aprendizado em profundidade, sendo uma das técnicas composta por Máquinas de Boltzmann Profundas e projeção bilinear e a outra composta por Máquinas de Boltzmann Profundas Multinomiais e projeção bilinear. A proposição deste método justifica-se pela demanda cada vez maior da área de Sistemas de Transporte Inteligentes. Para se obter um reconhecedor de veículos robusto, a proposta é utilizar o método de treinamento inferencial não-supervisionado Divergência por Contraste em conjunto com o método inferencial Campos Intermediários, para treinar múltiplas instâncias das redes profundas. Na fase de pré-treinamento local do método proposto são utilizadas projeções bilineares para reduzir o número de nós nas camadas da rede. A junção das estruturas em redes profundas treinadas separadamente forma a arquitetura final da rede neural, que passa por uma etapa de pré- treinamento global por Campos Intermediários. Na última etapa de treinamentos a rede neural Perceptron Multicamadas (MLP) é inicializada com os parâmetros pré-treinados globalmente e a partir deste ponto, inicia-se um processo de treinamento supervisionado utilizando gradiente conjugado de segunda ordem. O método proposto foi avaliado sobre a base BIT-Vehicle de imagens frontais de veículos coletadas de um ambiente de tráfego real. Os melhores resultados obtidos pelo método proposto utilizando rede profunda multinomial foram de 81, 83% de acurácia média na versão aumentada da base original e 91, 10% na versão aumentada da base combinada (Carros, Caminhões e Ônibus). Para a abordagem de redes profundas não multinomiais os melhores resultados foram de 81, 42% na versão aumentada da base original e 91, 13% na versão aumentada da base combinada. Com a aplicação da projeção bilinear, houve um decréscimo considerável nos tempos de treinamento das redes profundas multinomial e não multinomial, sendo que no melhor caso o tempo de execução do método proposto foi 5, 5 vezes menor em comparação com os tempos das redes profundas sem aplicação de projeção bilinear. / In this work it is proposed a vehicle recognition method for color images based on a Multilayer Perceptron neural network pre-trained through deep learning techniques (one technique composed by Deep Boltzmann Machines and bilinear projections and the other composed by Multinomial Deep Boltzmann Machines and bilinear projections). This proposition is justified by the increasing demand in Traffic Engineering area for the class of Intelligent Transportation Systems. In order to create a robust vehicle recognizer, the proposal is to use the inferential unsupervised training method of Contrastive Divergence together with the Mean Field inferential method, for training multiple instances of deep models. In the local pre-training phase of the proposed method, bilinear projections are used to reduce the number of nodes of the neural network. The combination of the separated trained deep models constitutes the final recognizer’s architecture, that yet will be global pre-trained through Mean Field. In the last phase of training the Multilayer Perceptron neural network is initialized with globally pre-trained parameters and from this point, a process of supervised training starts using second order conjugate gradient. The proposed method was evaluated over the BIT-Vehicle database of frontal images of vehicles collected from a real road traffic environment. The best results obtained by the proposed method that used multinomial deep models were 81.83% of mean accuracy in the augmented original database version and 91.10% in the augmented combined database version (Cars, Trucks and Buses). For the non-multinomial deep models approach, the best results were 81.42% in the augmented version of the original database and 91.13% in the augmented version of the combined database. It was also observed a significant decreasing in the training times of the multinomial deep models and non-multinomial deep models with bilinear projection application, where in the best case scenario the execution time of the proposed method was 5.5 times lower than the deep models that did not use bilinear projection.

This thesis considers statistical modelling of natural image data. Obtaining advances in this field can have significant impact for both engineering applications, and for the understanding of the human visual system. Several recent advances in natural image modelling have been obtained with the use of unsupervised feature learning. We consider a class of such models, restricted Boltzmann machines (RBMs), used in many recent state-of-the-art image models. We develop extensions of these stochastic artificial neural networks, and use them as a basis for building more effective image models, and tools for computational vision. We first develop a novel framework for obtaining Boltzmann machines, in which the hidden unit activations co-transform with transformed input stimuli in a stable and predictable way throughout the network. We define such models to be transformation equivariant. Such properties have been shown useful for computer vision systems, and have been motivational for example in the development of steerable filters, a widely used classical feature extraction technique. Translation equivariant feature sharing has been the standard method for scaling image models beyond patch-sized data to large images. In our framework we extend shallow and deep models to account for other kinds of transformations as well, focusing on in-plane rotations. Motivated by the unsatisfactory results of current generative natural image models, we take a step back, and evaluate whether they are able to model a subclass of the data, natural image textures. This is a necessary subcomponent of any credible model for visual scenes. We assess the performance of a state- of-the-art model of natural images for texture generation, using a dataset and evaluation techniques from in prior work. We also perform a dissection of the model architecture, uncovering the properties important for good performance. Building on this, we develop structured extensions for more complicated data comprised of textures from multiple classes, using the single-texture model architecture as a basis. These models are shown to be able to produce state-of-the-art texture synthesis results quantitatively, and are also effective qualitatively. It is demonstrated empirically that the developed multiple-texture framework provides a means to generate images of differently textured regions, more generic globally varying textures, and can also be used for texture interpolation, where the approach is radically dfferent from the others in the area. Finally we consider visual boundary prediction from natural images. The work aims to improve understanding of Boltzmann machines in the generation of image segment boundaries, and to investigate deep neural network architectures for learning the boundary detection problem. The developed networks (which avoid several hand-crafted model and feature designs commonly used for the problem), produce the fastest reported inference times in the literature, combined with state-of-the-art performance.

<div>When a user types in a search query in an Information Retrieval system, a list of top ‘n’ ranked documents relevant to the query are returned by the system. Relevant means not just returning documents that belong to the same category as that of the search query, but also returning documents that provide a concise answer to the search query. Determining the relevance of the documents is a significant challenge as the classic indexing techniques that use term/word frequencies do not consider the term (word) dependencies or the impact of previous terms on the current words or the meaning of the words in the document. There is a need to model the dependencies of the terms in the text data and learn the underlying statistical patterns to find the similarity between the user query and the documents to determine the relevancy.</div><div><br></div><div>This research proposes a solution based on Markov Random Fields (MRF) and Restricted Boltzmann Machines (RBM) to solve the problem of term dependencies and learn the underlying patterns to return documents that are very similar to the user query.</div>

Thesis (MScEng)--Stellenbosch University, 2014. / ENGLISH ABSTRACT: Process monitoring and fault diagnosis are used to detect abnormal events in processes. The early detection of such events or faults is crucial to continuous process improvement. Although principal component analysis and partial least squares are widely used for process monitoring and fault diagnosis in the metallurgical industries, these models are linear in principle; nonlinear approaches should provide more compact and informative models. The use of auto associative neural networks or auto encoders provide a principled approach for process monitoring. However, until very recently, these multiple layer neural networks have been difficult to train and have therefore not been used to any significant extent in process monitoring. With newly proposed algorithms based on the pre-training of the layers of the neural networks, it is now possible to train neural networks with very complex structures, i.e. deep neural networks. These neural networks can be used as auto encoders to extract features from high dimensional data. In this study, the application of deep auto encoders in the form of Restricted Boltzmann machines (RBM) to the extraction of features from process data is considered. These networks have mostly been used for data visualization to date and have not been applied in the context of fault diagnosis or process monitoring as yet. The objective of this investigation is therefore to assess the feasibility of using Restricted Boltzmann machines in various fault detection schemes. The use of RBM in process monitoring schemes will be discussed, together with the application of these models in automated control frameworks. / AFRIKAANSE OPSOMMING: Prosesmonitering en fout diagnose word gebruik om abnormale gebeure in prosesse op te spoor. Die vroeë opsporing van sulke gebeure of foute is noodsaaklik vir deurlopende verbetering van prosesse. Alhoewel hoofkomponent-analise en parsiële kleinste kwadrate wyd gebruik word vir prosesmonitering en fout diagnose in die metallurgiese industrieë, is hierdie modelle lineêr in beginsel; nie-lineêre benaderings behoort meer kompakte en insiggewende modelle te voorsien. Die gebruik van outo-assosiatiewe neurale netwerke of outokodeerders bied 'n beginsel gebaseerder benadering om dit te bereik. Hierdie veelvoudige laag neurale netwerke was egter tot onlangs moeilik om op te lei en is dus nie tot ʼn beduidende mate in die prosesmonitering gebruik nie. Nuwe, voorgestelde algoritmes, gebaseer op voorafopleiding van die lae van die neurale netwerke, maak dit nou moontlik om neurale netwerke met baie ingewikkelde strukture, d.w.s. diep neurale netwerke, op te lei. Hierdie neurale netwerke kan gebruik word as outokodeerders om kenmerke van hoë-dimensionele data te onttrek. In hierdie studie word die toepassing van diep outokodeerders in die vorm van Beperkte Boltzmann Masjiene vir die onttrekking van kenmerke van proses data oorweeg. Tot dusver is hierdie netwerke meestal vir data visualisering gebruik en dit is nog nie toegepas in die konteks van fout diagnose of prosesmonitering nie. Die doel van hierdie ondersoek is dus om die haalbaarheid van die gebruik van Beperkte Boltzmann Masjiene in verskeie foutopsporingskemas te assesseer. Die gebruik van Beperkte Boltzmann Masjiene se eienskappe in prosesmoniteringskemas sal bespreek word, tesame met die toepassing van hierdie modelle in outomatiese beheer raamwerke.

The decline of Moore's law has catalyzed a significant effort to identify beyond-CMOS devices and architectures for the coming decades. A multitude of classical and quantum systems have been proposed to address this challenge, and spintronics has emerged as a promising approach for these post-Moore systems. Many of these architectures are tailored specifically for applications in combinatorial optimization and machine learning. Here we propose the use of spintronics for such applications by exploring two distinct but related computing paradigms. First, the use of spin-currents to manipulate and control quantum information is investigated with demonstrated high-fidelity gate operation. This control is accomplished through repeated entanglement and measurement of a stationary qubit with a flying-spin through spin-torque like effects. Secondly, by transitioning from single-spin quantum bits to larger spin ensembles, we then explore the use of stochastic nanomagnets to realize a probabilistic system that is intrinsically governed by Boltzmann statistics. The nanomagnets explore the search space at rapid speeds and can be used in a wide-range of applications including optimization and quantum emulation by encoding the solution to a given problem as the ground state of the equivalent Boltzmann machine. These applications are demonstrated through hardware emulation using an all-digital autonomous probabilistic circuit.

In this thesis we develop a class of nonlinear generative models for high-dimensional time series. The first key property of these models is their distributed, or "componential" latent state, which is characterized by binary stochastic variables which interact to explain the data. The second key property is the use of an undirected graphical model to represent the relationship between latent state (features) and observations. The final key property is composability: the proposed class of models can form the building blocks of deep networks by successively training each model on the features extracted by the previous one. We first propose a model based on the Restricted Boltzmann Machine (RBM) that uses an undirected model with binary latent variables and real-valued "visible" variables. The latent and visible variables at each time step receive directed connections from the visible variables at the last few time-steps. This "conditional" RBM (CRBM) makes on-line inference efficient and allows us to use a simple approximate learning procedure. We demonstrate the power of our approach by synthesizing various motion sequences and by performing on-line filling in of data lost during motion capture. We also explore CRBMs as priors in the context of Bayesian filtering applied to multi-view and monocular 3D person tracking. We extend the CRBM in a way that preserves its most important computational properties and introduces multiplicative three-way interactions that allow the effective interaction weight between two variables to be modulated by the dynamic state of a third variable. We introduce a factoring of the implied three-way weight tensor to permit a more compact parameterization. The resulting model can capture diverse styles of motion with a single set of parameters, and the three-way interactions greatly improve its ability to blend motion styles or to transition smoothly among them. In separate but related work, we revisit Products of Hidden Markov Models (PoHMMs). We show how the partition function can be estimated reliably via Annealed Importance Sampling. This enables us to demonstrate that PoHMMs outperform various flavours of HMMs on a variety of tasks and metrics, including log likelihood.

The Internet of Things (IoTs) has triggered rapid advances in sensors, surveillance devices, wearables and body area networks with advanced Human-Computer Interfaces (HCI). Neural Networks optimized algorithmically for high accuracy and high representation power are very deep and require tremendous storage and processing capabilities leading to higher area and power costs. For developing smart front-ends for ‘always on’ sensor nodes we need to optimize for power and area. This requires considering trade-offs with respect to various entities such as resource utilization, processing time, area, power, accuracy etc. Our experimental results show that there is presence of a network configuration with minimum energy given the input constraints of an application in consideration. This presents the need for a hardware-software co-design approach. We present a highly parameterized hardware design on an FPGA allowing re-configurability and the ability to evaluate different design choices in a short amount of time. We also describe the capability of extending our design to offer run time configurability. This allows the design to be altered for different applications based on need and also allows the design to be used as a cascaded classifier beneficial for continuous sensing for low power applications. This thesis aims to evaluate the use of Restricted Boltzmann Machines for building such re-configurable low power front ends. We develop the hardware architecture for such a system and provide experimental results obtained for the case study of Posture detection for body worn cameras used for law enforcement.

In this thesis we develop a class of nonlinear generative models for high-dimensional time series. The first key property of these models is their distributed, or "componential" latent state, which is characterized by binary stochastic variables which interact to explain the data. The second key property is the use of an undirected graphical model to represent the relationship between latent state (features) and observations. The final key property is composability: the proposed class of models can form the building blocks of deep networks by successively training each model on the features extracted by the previous one. We first propose a model based on the Restricted Boltzmann Machine (RBM) that uses an undirected model with binary latent variables and real-valued "visible" variables. The latent and visible variables at each time step receive directed connections from the visible variables at the last few time-steps. This "conditional" RBM (CRBM) makes on-line inference efficient and allows us to use a simple approximate learning procedure. We demonstrate the power of our approach by synthesizing various motion sequences and by performing on-line filling in of data lost during motion capture. We also explore CRBMs as priors in the context of Bayesian filtering applied to multi-view and monocular 3D person tracking. We extend the CRBM in a way that preserves its most important computational properties and introduces multiplicative three-way interactions that allow the effective interaction weight between two variables to be modulated by the dynamic state of a third variable. We introduce a factoring of the implied three-way weight tensor to permit a more compact parameterization. The resulting model can capture diverse styles of motion with a single set of parameters, and the three-way interactions greatly improve its ability to blend motion styles or to transition smoothly among them. In separate but related work, we revisit Products of Hidden Markov Models (PoHMMs). We show how the partition function can be estimated reliably via Annealed Importance Sampling. This enables us to demonstrate that PoHMMs outperform various flavours of HMMs on a variety of tasks and metrics, including log likelihood.

Usually, human beings are able to quickly distinguish between different places, solely from their visual appearance. This is due to the fact that they can organize their space as composed of discrete units. These units, called ''semantic places'', are characterized by their spatial extend and their functional unity. Such a semantic category can thus be used as contextual information which fosters object detection and recognition. Recent works in semantic place recognition seek to endow the robot with similar capabilities. Contrary to classical localization and mapping works, this problem is usually addressed as a supervised learning problem. The question of semantic places recognition in robotics - the ability to recognize the semantic category of a place to which scene belongs to - is therefore a major requirement for the future of autonomous robotics. It is indeed required for an autonomous service robot to be able to recognize the environment in which it lives and to easily learn the organization of this environment in order to operate and interact successfully. To achieve that goal, different methods have been already proposed, some based on the identification of objects as a prerequisite to the recognition of the scenes, and some based on a direct description of the scene characteristics. If we make the hypothesis that objects are more easily recognized when the scene in which they appear is identified, the second approach seems more suitable. It is however strongly dependent on the nature of the image descriptors used, usually empirically derived from general considerations on image coding.Compared to these many proposals, another approach of image coding, based on a more theoretical point of view, has emerged the last few years. Energy-based models of feature extraction based on the principle of minimizing the energy of some function according to the quality of the reconstruction of the image has lead to the Restricted Boltzmann Machines (RBMs) able to code an image as the superposition of a limited number of features taken from a larger alphabet. It has also been shown that this process can be repeated in a deep architecture, leading to a sparse and efficient representation of the initial data in the feature space. A complex problem of classification in the input space is thus transformed into an easier one in the feature space. This approach has been successfully applied to the identification of tiny images from the 80 millions image database of the MIT. In the present work, we demonstrate that semantic place recognition can be achieved on the basis of tiny images instead of conventional Bag-of-Word (BoW) methods and on the use of Deep Belief Networks (DBNs) for image coding. We show that after appropriate coding a softmax regression in the projection space is sufficient to achieve promising classification results. To our knowledge, this approach has not yet been investigated for scene recognition in autonomous robotics. We compare our methods with the state-of-the-art algorithms using a standard database of robot localization. We study the influence of system parameters and compare different conditions on the same dataset. These experiments show that our proposed model, while being very simple, leads to state-of-the-art results on a semantic place recognition task.