No description available.

京都大学 / 新制・課程博士 / 博士(理学) / 甲第23677号 / 理博第4767号 / 新制||理||1683(附属図書館) / 京都大学大学院理学研究科数学・数理解析専攻 / (主査)准教授 SVADLENKA Karel, 教授 泉 正己, 教授 坂上 貴之 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DFAM

The mechanism and the oxidation potential of the dissociative single electron transfer for tetra-n-butylammonium acetate has been investigated via conventional (cyclic voltammetry) and convolution voltammetry. The oxidation potential for tetra-n-butylammonium acetate was determined to be 0.60 ± 0.10 (vs. Ag/ (0.1 M) AgNO₃) in anhydrous acetonitrile. The results also indicated the mechanism of oxidation was concerted dissociative electron transfer (cDET), rather than stepwise as was previously reported. To further investigate the mechanism, a series of aliphatic and aromatic tetra-n butylammonium carboxylates were synthesized and investigated via convolution and conventional methods under anhydrous conditions (propionate, pivalate, phenyl acetate, and benzoate). The reported results showed high reproducibility and consistency with a concerted dissociative electron transfer for aliphatic carboxylates with a systematic shift in the oxidation potentials (0.60 ± 0.09 V for acetate, 0.47 ± 0.05 V for propionate, and 0.40 ± 0.05 V for pivalate) within the series which is expected trend based on radical stabilization energies of the alkyl groups on the aliphatic carboxylates. Hydrogen bonding was investigated as a possible source for the discrepancy between our results and the reported mechanism of the dissociative electron transfer. Because of the extreme hygroscopic nature of carboxylate salts, it was hypothesized that the presence of small amounts of water might alter the reaction mechanism. Deionized water and deuterium oxide additions to anhydrous acetonitrile were performed to test this hypothesis. The mechanism was noted to shift towards a stepwise mechanism as water was added. In addition, the derived oxidation potentials became more positive with increasing concentrations of water. Several explanations are presented with regards to water effects on the shift in the electron transfer mechanism. Indirect electrolysis (homogeneous redox catalysis) was also employed as an alternative and independent approach to quantify the oxidation potentials of carboxylates. A series of substituted ferrocenes were investigated as mediators for the oxidation of tetra-n-butylammonium acetate. Preliminary data showed redox catalysis was feasible for these systems. Further analyses of the electrochemical results suggested a follow-up chemical step (addition to mediator) that competes with the redox catalysis mechanism. As predicted from theoretical working curves, a plateau region in the i_p/i_pd plots (where no meaningful kinetic information could be obtained) was observed. Products mixture analyses verified the consumption of the mediator upon electrolysis, but no further information with regards to the nature of the mechanism was deduced. In a related study the effects of hydrogen bonding and ions on the reactivity of neutral free radicals were examined by laser flash photolysis. The rate of the β-scission of the cumyloxyl radical is influenced by cations (Li⁺ > Mg²⁺ ≈ Na⁺ > ⁿBu₄N⁺) due to stabilizing ion-dipole interactions in the transition state of the developing carbonyl group. Experimental findings are in a good agreement with theoretical work suggesting metal ion complexation can cause radical clocks to run fast with a more significant effect if there is an increase in dipole moment going from the reactant to the transition state. / Ph. D. / Our work focuses on employing electrochemical techniques to investigate single electron transfer processes, which lead to unstable organic species that contain an odd number of electrons called radicals and radical ions. Many essential biological and environmental pathways are found to occur via radicals, i.e. photosynthesis, atmospheric degradation, enzyme catalyzed reactions in biology, autooxidation, DNA mutations, and more. Electrochemical techniques permit us to investigate the scientific fundamentals of radical processes by generating radicals and radical ions in a controlled manner with a higher efficiency. We have combined electrochemical techniques with established physical organic theories of electron transfer to allow us to determine of the rate and mechanism of electron transfer for a selective group of chemical compounds, specifically anions derived from carboxylic acids (carboxylates). A fundamental understanding of single electron transfer processes for carboxylates allows for a prediction of chemical behavior and the future design of novel chemical compounds for alternative chemical functionality. Our findings are the first to report experimental evidence for a so-called concerted dissociative electron transfer mechanism for carboxylates, where the transfer of an electron is accompanied by the simultaneous breakage of a carbon-carbon bond yielding a radical and carbon dioxide. The mechanism has been shown to proceed in a stepwise fashion only in the presence of water. Our work highlights the environmental effects on radical stability such as water and metals ions.

In the past two decades, there has been huge amount of interest in Multiresolution Analysis Algorithms (MAAs) and their applications. Processing some of their applications such as medical imaging are computationally intensive, power hungry and requires large amount of memory which cause a high demand for efficient algorithm implementation, low power architecture and acceleration. Recently, some MAAs such as Finite Ridgelet Transform (FRIT) Haar Wavelet Transform (HWT) are became very popular and they are suitable for a number of image processing applications such as detection of line singularities and contiguous edges, edge detection (useful for compression and feature detection), medical image denoising and segmentation. Efficient hardware implementation and acceleration of these algorithms particularly when addressing large problems are becoming very chal-lenging and consume lot of power which leads to a number of issues including mobility, reliability concerns. To overcome the computation problems, Field Programmable Gate Arrays (FPGAs) are the technology of choice for accelerating computationally intensive applications due to their high performance. Addressing the power issue requires optimi- sation and awareness at all level of abstractions in the design flow. The most important achievements of the work presented in this thesis are summarised here. Two factorisation methodologies for HWT which are called HWT Factorisation Method1 and (HWTFM1) and HWT Factorasation Method2 (HWTFM2) have been explored to increase number of zeros and reduce hardware resources. In addition, two novel efficient and optimised architectures for proposed methodologies based on Distributed Arithmetic (DA) principles have been proposed. The evaluation of the architectural results have shown that the proposed architectures results have reduced the arithmetics calculation (additions/subtractions) by 33% and 25% respectively compared to direct implementa-tion of HWT and outperformed existing results in place. The proposed HWTFM2 is implemented on advanced and low power FPGA devices using Handel-C language. The FPGAs implementation results have outperformed other existing results in terms of area and maximum frequency. In addition, a novel efficient architecture for Finite Radon Trans-form (FRAT) has also been proposed. The proposed architecture is integrated with the developed HWT architecture to build an optimised architecture for FRIT. Strategies such as parallelism and pipelining have been deployed at the architectural level for efficient im-plementation on different FPGA devices. The proposed FRIT architecture performance has been evaluated and the results outperformed some other existing architecture in place. Both FRAT and FRIT architectures have been implemented on FPGAs using Handel-C language. The evaluation of both architectures have shown that the obtained results out-performed existing results in place by almost 10% in terms of frequency and area. The proposed architectures are also applied on image data (256 £ 256) and their Peak Signal to Noise Ratio (PSNR) is evaluated for quality purposes. Two architectures for cyclic convolution based on systolic array using parallelism and pipelining which can be used as the main building block for the proposed FRIT architec-ture have been proposed. The first proposed architecture is a linear systolic array with pipelining process and the second architecture is a systolic array with parallel process. The second architecture reduces the number of registers by 42% compare to first architec-ture and both architectures outperformed other existing results in place. The proposed pipelined architecture has been implemented on different FPGA devices with vector size (N) 4,8,16,32 and word-length (W=8). The implementation results have shown a signifi-cant improvement and outperformed other existing results in place. Ultimately, an in-depth evaluation of a high level power macromodelling technique for design space exploration and characterisation of custom IP cores for FPGAs, called func-tional level power modelling approach have been presented. The mathematical techniques that form the basis of the proposed power modeling has been validated by a range of custom IP cores. The proposed power modelling is scalable, platform independent and compares favorably with existing approaches. A hybrid, top-down design flow paradigm integrating functional level power modelling with commercially available design tools for systematic optimisation of IP cores has also been developed. The in-depth evaluation of this tool enables us to observe the behavior of different custom IP cores in terms of power consumption and accuracy using different design methodologies and arithmetic techniques on virous FPGA platforms. Based on the results achieved, the proposed model accuracy is almost 99% true for all IP core's Dynamic Power (DP) components.

Convolution and multiplication operations in the filtering process can be optimized by minimizing the resource utilization using Field Programmable Gate Arrays (FPGA) and separable filter kernels. An FPGA architecture for separable convolution is proposed to achieve reduction of on-chip resource utilization and external memory bandwidth for a given processing rate of the convolution unit. Multiplication in integer number system can be optimized in terms of resources, operation time and power consumption by converting to logarithmic domain. To achieve this, a method altering the filter weights is proposed and implemented for error reduction. The results obtained depict significant error reduction when compared to existing methods, thereby optimizing the multiplication in terms of the above mentioned metrics. Underwater video and still images are used by many programs within National Oceanic Atmospheric and Administration (NOAA) fisheries with the objective of identifying, classifying and quantifying living marine resources. They use underwater cameras to get video recording data for manual analysis. This process of manual analysis is labour intensive, time consuming and error prone. An efficient solution for this problem is proposed which uses Gabor filters for feature extraction. The proposed method is implemented to identify two species of fish namely Epinephelus morio and Ocyurus chrysurus. The results show higher rate of detection with minimal rate of false alarms.

Le graphe est un concept puissant pour la représentation des relations entre des paires d'entités. Les données ayant une structure de graphes sous-jacente peuvent être trouvées dans de nombreuses disciplines, décrivant des composés chimiques, des surfaces des modèles tridimensionnels, des interactions sociales ou des bases de connaissance, pour n'en nommer que quelques-unes. L'apprentissage profond (DL) a accompli des avancées significatives dans une variété de tâches d'apprentissage automatique au cours des dernières années, particulièrement lorsque les données sont structurées sur une grille, comme dans la compréhension du texte, de la parole ou des images. Cependant, étonnamment peu de choses ont été faites pour explorer l'applicabilité de DL directement sur des données structurées sous forme des graphes. L'objectif de cette thèse est d'étudier des architectures de DL sur des graphes et de rechercher comment transférer, adapter ou généraliser à ce domaine des concepts qui fonctionnent bien sur des données séquentielles et des images. Nous nous concentrons sur deux primitives importantes : le plongement de graphes ou leurs nœuds dans une représentation de l'espace vectorielle continue (codage) et, inversement, la génération des graphes à partir de ces vecteurs (décodage). Nous faisons les contributions suivantes. Tout d'abord, nous introduisons Edge-Conditioned Convolutions (ECC), une opération de type convolution sur les graphes réalisés dans le domaine spatial où les filtres sont générés dynamiquement en fonction des attributs des arêtes. La méthode est utilisée pour coder des graphes avec une structure arbitraire et variable. Deuxièmement, nous proposons SuperPoint Graph, une représentation intermédiaire de nuages de points avec de riches attributs des arêtes codant la relation contextuelle entre des parties des objets. Sur la base de cette représentation, l'ECC est utilisé pour segmenter les nuages de points à grande échelle sans sacrifier les détails les plus fins. Troisièmement, nous présentons GraphVAE, un générateur de graphes permettant de décoder des graphes avec un nombre de nœuds variable mais limité en haut, en utilisant la correspondance approximative des graphes pour aligner les prédictions d'un auto-encodeur avec ses entrées. La méthode est appliquée à génération de molécules / Graph is a powerful concept for representation of relations between pairs of entities. Data with underlying graph structure can be found across many disciplines, describing chemical compounds, surfaces of three-dimensional models, social interactions, or knowledge bases, to name only a few. There is a natural desire for understanding such data better. Deep learning (DL) has achieved significant breakthroughs in a variety of machine learning tasks in recent years, especially where data is structured on a grid, such as in text, speech, or image understanding. However, surprisingly little has been done to explore the applicability of DL on graph-structured data directly.The goal of this thesis is to investigate architectures for DL on graphs and study how to transfer, adapt or generalize concepts working well on sequential and image data to this domain. We concentrate on two important primitives: embedding graphs or their nodes into a continuous vector space representation (encoding) and, conversely, generating graphs from such vectors back (decoding). To that end, we make the following contributions.First, we introduce Edge-Conditioned Convolutions (ECC), a convolution-like operation on graphs performed in the spatial domain where filters are dynamically generated based on edge attributes. The method is used to encode graphs with arbitrary and varying structure.Second, we propose SuperPoint Graph, an intermediate point cloud representation with rich edge attributes encoding the contextual relationship between object parts. Based on this representation, ECC is employed to segment large-scale point clouds without major sacrifice in fine details.Third, we present GraphVAE, a graph generator allowing to decode graphs with variable but upper-bounded number of nodes making use of approximate graph matching for aligning the predictions of an autoencoder with its inputs. The method is applied to the task of molecule generation

Le développement de méthodes de la détection en temps réel de cellules anormales (pouvant être considérées comme des cellules cancéreuses) par captures et traitements bio-images sont des axes de recherche importants dans le domaine biomédical car cela contribue à diagnostiquer un cancer. C’est dans ce contexte que se situe ces travaux de thèse. Plus précisément, les travaux présentés dans ce manuscrit, se focalise sur le développement de procédures de lecture, de détection et de classification automatiques de bio-images de cellules anormales considérées comme des cellules cancéreuses. Par conséquent, une première étape du travail à consister à déterminer une solution de détection, à partir d’images microscopiques multispectrales permettant une répétitivité d’images sur une gamme de longueurs d'ondes de certains types de bio-images anormales associées à différents stades ou évolutions de cellules cancéreuses. L’approche développée dans ces travaux repose sur l’exploitation d’une nouvelle méthode de segmentation basée sur l'intensité de la couleur et pouvant être appliquée sur des séquences d'objets dans une image en reformant de manière adaptative et itérative la localisation et la couverture de contours réels de cellules. Cette étape préalable de segmentation est primordiale et permet une classification des tissus anormaux en utilisant la méthode de réseau de neurones à convolution (CNN) appliqué sur les images microscopiques segmenté de type snake. L’approche permet d’obtenir de bas résultats comparativement à une approche basée sur d’autres méthodes de segmentation de la littérature. En effet, cette méthode de classification atteint des valeurs de performance de 100% pour la phase d’apprentissage et de 99.168 % pour les phases de test. Cette méthode est comparée à différents travaux antérieurs et basée sur différentes fonctionnalités d'extraction, et a prouvé son efficacité par rapport à ces autres méthodes. En terme de perspectives, les travaux futurs visent à valider notre approche sur des ensembles de données plus larges, et à explorer différentes architectures CNN selon différents critères d’optimisation / Development of methods for help diagnosis of the real time detection of abnormal cells (which can be considered as cancer cells) through bio-image processing and detection are most important research directions in information science and technology. Our work has been concerned by developing automatic reading procedures of the normal and abnormal bio-images tissues. Therefore, the first step of our work is to detect a certain type of abnormal bio-images associated to many types evolution of cancer within a Microscopic multispectral image, which is an image, repeated in many wavelengths. And using a new segmentation method that reforms itself in an iterative adaptive way to localize and cover the real cell contour, using some segmentation techniques. It is based on color intensity and can be applied on sequences of objects in the image. This work presents a classification of the abnormal tissues using the Convolution neural network (CNN), where it was applied on the microscopic images segmented using the snake method, which gives a high performance result with respect to the other segmentation methods. This classification method reaches high performance values, where it reaches 100% for training and 99.168% for testing. This method was compared to different papers that uses different feature extraction, and proved its high performance with respect to other methods. As a future work, we will aim to validate our approach on a larger datasets, and to explore different CNN architectures and the optimization of the hyper-parameters, in order to increase its performance, and it will be applied to relevant medical imaging tasks including computer-aided diagnosis

Deep learning has in recent years come to dominate the previously separate fields of research in machine learning, computer vision, natural language understanding and speech recognition. Despite breakthroughs in training deep networks, there remains a lack of understanding of both the optimization and structure of deep networks. The approach advocated by many researchers in the field has been to train monolithic networks with excess complexity, and strong regularization --- an approach that leaves much to desire in efficiency. Instead we propose that carefully designing networks in consideration of our prior knowledge of the task and learned representation can improve the memory and compute efficiency of state-of-the art networks, and even improve generalization --- what we propose to denote as structural priors. We present two such novel structural priors for convolutional neural networks, and evaluate them in state-of-the-art image classification CNN architectures. The first of these methods proposes to exploit our knowledge of the low-rank nature of most filters learned for natural images by structuring a deep network to learn a collection of mostly small, low-rank, filters. The second addresses the filter/channel extents of convolutional filters, by learning filters with limited channel extents. The size of these channel-wise basis filters increases with the depth of the model, giving a novel sparse connection structure that resembles a tree root. Both methods are found to improve the generalization of these architectures while also decreasing the size and increasing the efficiency of their training and test-time computation. Finally, we present work towards conditional computation in deep neural networks, moving towards a method of automatically learning structural priors in deep networks. We propose a new discriminative learning model, conditional networks, that jointly exploit the accurate representation learning capabilities of deep neural networks with the efficient conditional computation of decision trees. Conditional networks yield smaller models, and offer test-time flexibility in the trade-off of computation vs. accuracy.

<p>In recent years, deep learning models have become popular in the real-time embedded application, but there are many complexities for hardware deployment because of limited resources such as memory, computational power, and energy. Recent research in the field of deep learning focuses on reducing the model size of the Convolution Neural Network (CNN) by various compression techniques like Architectural compression, Pruning, Quantization, and Encoding (e.g., Huffman encoding). Network pruning is one of the promising technique to solve these problems.</p> <p>This thesis proposes methods to prune the convolution neural network (SqueezeNet) without introducing network sparsity in the pruned model. </p> <p>This thesis proposes three methods to prune the CNN to decrease the model size of CNN without a significant drop in the accuracy of the model.</p> <p>1: Pruning based on Taylor expansion of change in cost function Delta C.</p> <p>2: Pruning based on L₂ normalization of activation maps.</p> <p>3: Pruning based on a combination of method 1 and method 2.</p><p>The proposed methods use various ranking methods to rank the convolution kernels and prune the lower ranked filters afterwards SqueezeNet model is fine-tuned by backpropagation. Transfer learning technique is used to train the SqueezeNet on the CIFAR-10 dataset. Results show that the proposed approach reduces the SqueezeNet model by 72% without a significant drop in the accuracy of the model (optimal pruning efficiency result). Results also show that Pruning based on a combination of Taylor expansion of the cost function and L₂ normalization of activation maps achieves better pruning efficiency compared to other individual pruning criteria and most of the pruned kernels are from mid and high-level layers. The Pruned model is deployed on BlueBox 2.0 using RTMaps software and model performance was evaluated.</p><p></p>

In the development of advanced driver assistance systems, computer vision problemsneed to be optimized to run efficiently on embedded platforms. Convolutional neural network(CNN) accelerators have proven to be very efficient for embedded camera platforms,such as the ones used for automotive vision systems. Therefore, the focus of this thesisis to evaluate the efficiency of a CNN on a future embedded heterogeneous processingdevice. The memory size in an embedded system is often very limited, and it is necessary todivide the input into multiple tiles. In addition, there are power and speed constraintsthat needs to be met to be able to use a computer vision system in a car. To increaseefficiency and optimize the memory usage, different methods for CNN layer fusion areproposed and evaluated for a variety of tile sizes. Several different layer fusion methods and input tile sizes are chosen as optimal solutions,depending on the depth of the layers in the CNN. The solutions investigated inthe thesis are most efficient for deep CNN layers, where the number of channels is high.