Dissertations / Theses on the topic 'CNRN'

2011/2012
Lo studio della fisica nucleare in Italia ebbe il suo mito fondativo nelle vicende dei “ragazzi di via Panisperna”, dal nome della via romana in cui sorgevano i laboratori diretti da Enrico Fermi. Dopo aver raggiunto la fama mondiale (in particolare con il Nobel per la fisica di Fermi nel 1938), il gruppo fu disperso a causa della politica (sia razziale che scientifica) del regime fascista. Mentre Fermi ed altri, espatriati in America, davano il proprio determinante contributo alla realizzazione della bomba atomica, in Italia rimase il solo Edoardo Amaldi che, nel dopoguerra, si trovò ad essere, nel Paese e fuori, un fondamentale punto di riferimento per la fisica italiana. Nell’immediato dopoguerra, a fronte di un sostanziale disinteresse del Governo italiano in materia di ricerca, furono le industrie elettriche private a muovere i primi passi verso la ricerca e lo sviluppo della tecnologia nucleare, concedendo il proprio appoggio ad alcuni giovani ricercatori del Politecnico di Milano che diedero vita al CISE (Centro Informazioni Studi Esperienze). Parallelamente, la “comunità dei fisici” iniziava a ritagliarsi un proprio spazio autonomo di manovra. Nel 1951 i gruppi universitari che si occupavano di fisica fondamentale diedero vita all’Istituto Nazionale di Fisica Nucleare (INFN) mentre l'anno successivo, non senza attriti con il CISE, il Ministero dell'Industria appoggiò la creazione del Comitato Nazionale per le Ricerche Nucleari (CNRN), incaricato di promuovere e occuparsi della fisica nucleare applicata. Alla presidenza fu nominato Francesco Giordani, chimico napoletano legato all'IRI ed agli ambienti del neo meridionalismo. Il Comitato, che solo nel 1960 fu mutato in CNEN (Comitato Nazionale per l'Energia Nucleare) ottenendo la necessaria personalità giuridica, dovette costantemente far fronte alle difficoltà derivanti dalla propria fragilità istituzionale e dalle continue tensioni con l'industria elettrica privata. Ciononostante, sotto la guida del suo Segretario Generale Felice Ippolito, riuscì a dar vita ad importanti realizzazioni (come il Sincrotrone di Frascati o il Centro di ricerche nucleari di Ispra) e diede l'impulso fondamentale che portò alla costruzione, nei primi anni '60, delle prime centrali nucleari in Italia. Questo “periodo aureo” della fisica nucleare applicata iniziò a finire nell’agosto del 1963 quando una dura campagna stampa prese a mettere in discussione le gestione di Ippolito che si ritrovò al centro di un “caso” che da mediatico si fece ben presto giuridico e portò, nel marzo del 1964, all'arresto del Segretario Generale del CNEN per irregolarità amministrative. “Il caso Ippolito”, lungi dall’essere solo un processo per un isolato caso di malversazione, di fatto sancì la fine della ricerca e dello sviluppo del nucleare in Italia, facendo piazza pulita non solo dei progetti di nuove centrali atomiche, ma anche di un certo tipo di gestione degli enti pubblici che aveva fatto dell’elasticità amministrativa il proprio punto di forza, laddove in seguito si impose la burocratizzazione e la lottizzazione politica. Nella tesi in esame ho tentato di individuare, all'interno di un campo di ricerca così vario e così ricco di spunti collocati a cavallo di più discipline storiche (storia e filosofia della scienza, dell'economica e dell'industria, della cultura e della politica, delle relazioni internazionali), alcuni snodi focali ed emblematici che permettessero di sviluppare un percorso di indagine su quello che appariva come un meraviglioso tentativo di far recuperare all'Italia il tempo perduto a causa del regime fascista, in termini di sviluppo tecnologico e scientifico, ma anche culturale e politico. Tale tentativo ottenne risultati di rilievo mondiale nel dopoguerra ma andò incontro ad una nuova sconfitta, nei primi anni '60, quando emerse l'incapacità dello Stato di riformare se stesso per tener dietro ai rapidi mutamenti, non solo tecnici, che la tecnologia d'eccellenza pretende per mantenersi tale. Fin dall'inizio ho individuato in Felice Ippolito il trait d'union tra i fatti caratterizzanti le vicende trattate. Interessato alla ricerca nucleare quale geologo esperto in prospezioni minerarie, in seguito venne nominato Segretario Generale del CNRN e si trovò a rivestire un ruolo chiave, emblematico e rappresentativo, all'interno di un complesso ambiente culturale composto da intellettuali, scienziati ed alti funzionari che parteciparono ad una rete di rapporti all'interno della quale si elaborarono delle organiche strategie di sviluppo per il Paese. Ippolito divenne referente e portavoce di una comunità scientifica che si caratterizzava in quegli anni per il suo rapporto estremamente dialettico e consapevole con tutte le componenti della società, dalla classe politica al mondo dell'industria e dell'economica, dal mondo della cultura alle classi subalterne. Per comprendere l'incontro tra Ippolito e la comunità dei fisici, ho ritenuto di iniziare la tesi con un accenno all'esperienza dei “ragazzi di via Panisperna” e di Enrico Fermi, in particolare. La partenza in treno di Fermi per Stoccolma, il 6 dicembre 1938, dove avrebbe ritirato il premio Nobel prima di espatriare negli Stati Uniti (in fuga dalle leggi razziali ma soprattutto dall'incapacità del regime fascista di comprenderne e sostenerne le iniziative), è stata presentata come evento simbolico e metaforico della perdita di un primo “treno per la modernità” da parte dell'Italia. L'attenzione è stata posta soprattutto su chi rimase sulla banchina di quella stazione, ovvero Edoardo Amaldi, che pur con molti dubbi alla fine scelse di rimanere in Italia diventando il punto di riferimento per eccellenza, in virtù del suo carisma scientifico ed umano, della comunità dei fisici italiani nel dopoguerra. In particolare ho messo in evidenza il rafforzarsi in Amaldi di un punto di vista autonomo su quello che doveva essere il rapporto tra la ricerca scientifica ed il mondo della politica e dell'industria. Mentre oltreoceano Fermi delegava al Governo la valutazione etica e la gestione dei risultati del proprio lavoro scientifico, in Italia il suo allievo Amaldi fin dal dopoguerra iniziò a tessere una rete di rapporti, con l'industria e le aziende controllate dallo Stato, caratterizzati da alcuni principi imprescindibili. Quando gli industriali elettrici privati lo chiamarono al CISE, Amaldi pose perentorie condizioni alla propria partecipazione, come la difesa della sua autonomia scientifica, il rifiuto di ogni principio di segretezza, ed il fatto che la ricerca doveva andare a beneficio dell'intera collettività e non a vantaggio di pochi gruppi privati. Dopo aver delineato alcuni elementi della figura di Amaldi, ho concentrato il mio interesse su Ippolito e sui suoi rapporti con l'ambiente culturale napoletano, liberale e meridionalista, di cui anche Francesco Giordani faceva parte. Attraverso la bibliografia e gli archivi dell'ente, ho esaminato la nascita del CNRN sull'asse Ippolito-Giordani-Pietro Campilli (il Ministro dell'Industria che sostenne il progetto) e di seguito l'insorgere delle tensioni con il CISE e l'industria privata. L'obiettivo è stato di mettere in evidenza l'estrema “coerenza” dell'incontro tra i fisici rappresentati da Amaldi e la politica scientifica portata avanti da Ippolito e Giordani, capaci di soddisfarne sia le ambizioni tecnico scientifiche che etiche e politiche. Con un capitolo intermedio, su tematiche di politica nucleare internazionale, ho introdotto il tema dell'iniziatica Atoms for peace, lanciata dal Presidente americano Eisenhower, che prospettava una politica di disarmo atomico fondata sulla socializzazione della tecnologia nucleare ad uso civile. Rinunciando a proporre un inquadramento storiografico e critico complessivo, ho scelto di render conto della rappresentazione offerta da uno dei protagonisti di quegli anni, ovvero il francese Bertrand Goldschmidt, che influenzò grandemente il punto di vista di Ippolito e degli Amici del Mondo (cui Ippolito si legò) e che oggi testimonia in maniera particolarmente efficace il clima di “euforia atomica” che determinò allora fondamentali scelte di politica energetica europea. L'iniziativa Atoms for Peace diede l'occasione ad Ippolito di avviare un'intesa collaborazione con l'ambiente culturale che ruotava attorno alla rivista «Il Mondo» diretta da Mario Pannunzio e che in quel momento si presentava come la fucina, di stampo liberale radicale, dei progetti politici che portarono in seguito al Centrosinistra. Ripercorrendo le pagine della rivista ho messo in evidenza un percorso di progressiva presa di coscienza sulla questione nucleare. Se fino all'iniziativa Atoms for Peace erano considerate solo le applicazioni militari di tale tecnologia, in seguito e anche grazie all'intervento di Ippolito, il dibattito sul nucleare venne connesso alla questione della produzione energetica vista nella prospettiva della lotta contro i monopoli e per la nazionalizzazione del settore. Su questi temi centrali in quella fase politica (sulla nazionalizzazione del settore elettrico si giocò la battaglia fondamentale per il Centrosinistra), Ippolito in particolare, a metà degli anni '50, iniziò a tessere un discorso unitario tra crescente richiesta energetica, sviluppo della tecnologia nucleare e necessaria nazionalizzazione. Coerenti a questa linea iniziarono ad apparire su «Il Mondo» i “Dialoghi plutonici” di Ernesto Rossi che testimoniavano i rapporti sempre più stretti tra Ippolito e la rivista, nel contesto delle vicissitudini politiche che portarono alla nascita del Partito Radicale ed ai convegni degli Amici del Mondo “La lotta contro i monopoli” e “Atomo ed elettricità”. Usando gli atti dei convegni e analizzando i molti articoli in merito apparsi sulla rivista, ho messo in evidenza il processo che portò, a partire dalle posizioni antistataliste sempre sostenute sulle pagine di «Il Mondo» in particolare da Rossi, al definirsi della presa di posizione nazionalizzatrice espressa durante il convegno “La lotta contro i monopoli”. Del seminario “Atomo ed elettricità” ho ritenuto di particolare interesse l'identificazione operata dai relatori tra esigenze tecnico-scientifiche dell'energia nucleare e opzione nazionalizzatrice che portò ad una lettura prettamente politica delle scelte tecniche da operare in materia di filiere tecnologiche. Lettura che, come evidenzieremo, Ippolito non condividerà a favore di un approccio che preferisce le soluzioni particolari alle analisi universali. Atoms for Peace comporta un rilancio generale della politica nucleare italiana anche in termini di “gara atomica” tra ricerca e sviluppo pubblici e privati. In particolare, ho esaminato il crescente clima di ostilità tra il CNRN e l'industria privata (l'Edison in particolare) e le cause che portarono alle dimissioni di Giordani dalla Presidenza del Comitato. In un capitolo titolato “Come Mattei all'Agip” ho delineato le difficoltà istituzionali che dovette affrontare Ippolito da segretario plenipotenziario del CNRN ed il conseguente sviluppo di un modus operandi problematico che ebbe importanti conseguenze nella creazione del “caso” che sarebbe esploso. Tra le molte vicissitudini del CNRN ho seguito soprattutto il processo che portò alla costruzione delle prime centrali atomiche in Italia con particolar attenzione alla collaborazione tra CNRN e Banca Mondiale che portò alla costruzione della centrale di Garigliano e che sintetizzò istanze meridionaliste e nucleariste. Con il capitolo “Dal CNRN al CNEN” ho esaminato il percorso politico che portò alla nascita del CNEN nel contesto delle trattative per il primo Governo di Centrosinistra e della nazionalizzazione dell'energia elettrica. L'obiettivo è stato in particolare mettere in evidenza le tensioni che andarono delineandosi all'interno del nuovo ente elettrico, l'ENEL, tra le posizioni rappresentate dal Direttore Generale Angelini ed il consigliere Ippolito. Negli ultimi due capitoli ho riassunto in modo antologico l'aspetto più ampiamente trattato dalla storiografia esistente sul tema, ovvero il “caso” mediatico e giuridico che prese il nome del Segretario Generale del CNEN e che portò alla sua incarcerazione. Oltre alla fase processuale, ho ricostruito il quadro politico e gli avvenimenti che portarono alla messa in stato di accusa di Ippolito, nell'estate del 1963, ed alla sua incarcerazione l'anno successivo, che ebbero come diretta conseguenza il drastico ridimensionamento dei programmi nucleari del CNEN. Infine ho proposto un'analisi delle ipotesi interpretative date al “caso Ippolito” evidenziando anche alcuni aspetti che, per varie ragioni, non sono stati ancora indagati. In ultima analisi il presente studio tenta di mettere in luce la complessità della materia trattata che, pur prestandosi per molte ragioni alle semplificazioni complottistiche e dietrologiche di stampo giornalistico, risulta incomprensibile senza una contestualizzazione capace di connettere il percorso della fisica nucleare italiana (che a partire dall'esperienza dei “ragazzi di via Panisperna” tende a pensarsi e muoversi come una “comunità” portatrice di propri interessi e ideali), il dibattito filosofico, culturale e tecnico sulle ragioni e sui mezzi dell'intervento dello Stato nell'economia e sul ruolo di intellettuali e scienziati nella società, ed infine la storia politica italiana, europea ed internazionale che portò alla nascita del Centrosinistra.
XXIV Ciclo
1975

The content of this thesis illustrates the six months work done during my internship at TKH Security Solutions - Siqura B.V. in Gouda, Netherlands. The aim of this thesis is to investigate on convolutional neural networks possible usage, from two different point of view: first we propose a novel algorithm for person re-identification, second we propose a deployment chain, for bringing research concepts to product ready solutions. In existing works, the person re-identification task is assumed to be independent of the person detection task. In this thesis instead, we consider the two tasks as linked. In fact, features produced by an object detection convolutional neural network (CNN) contain useful information, which is not being used by current re-identification methods. We propose several solutions for learning a metric on CNN features to distinguish between different identities. Then the best of these solutions is compared with state of the art alternatives on the popular Market-1501 dataset. Results show that our method outperforms them in computational efficiency, with only a reasonable loss in accuracy. For this reason, we believe that the proposed method can be more appropriate than current state of the art methods in situations where the computational efficiency is critical, such as embedded applications. The deployment chain we propose in this thesis has two main goals: it must be flexible for introducing new advancement in networks architecture, and it must be able to deploy neural networks both on server and embedded platforms. We tested several frameworks on several platforms and we ended up with a deployment chain that relies on the open source format ONNX.

O objetivo deste trabalho foi realizar a dosimetria de filtros dinâmicos aprimorados (“Enhanced Dynamic Wedge” – EDW). Isso foi feito com dois intuitos: 1) a avaliação da implementação clínica do filtro dinâmico aprimorado considerando-se suas vantagens e desvantagens; e 2) a comparação entre a eficiência e a melhor aplicabilidade de três sistemas dosimétricos diferentes em conjunto com o filtro dinâmico aprimorado. Os sistemas dosimétricos utilizados foram: o Sistema de Imagem de Feixe (“Beam Image System” – BIS2G), da Scanditronix-Wellhöfer, uma câmara de ionização cilíndrica de 0,6 cm3 do tipo Farmer, e um arranjo linear de câmaras de ionização múltiplas (“Ionization Chamber Linear Detector Array”) com 23 câmaras de ionização cilíndricas de 0,14 cm3, tipo Farmer. Com o Sistema de Imagem de Feixe – BIS2G foi possível obter curvas de fluência que permitiram analisar a simetria dos perfis de fluência, linearidade da dose, efeito da taxa de dose e efeito da gravidade, características diretamente ligadas à qualidade do funcionamento do filtro dinâmico aprimorado, além de ter sido possível também calcular o fator filtro para campos simétricos. Com a câmara de ionização mediu-se a porcentagem de dose em profundidade - PDP e calculou-se o fator filtro para campos simétricos e campos assimétricos (uma das vantagens do EDW). Com o arranjo linear de câmaras de ionização múltiplas foi possível obter curvas de isodose que permitiram a verificação da simetria e a comparação com os resultados obtidos com o BIS2G, além do cálculo do fator filtro para campos simétricos. O fator filtro calculado através de medidas tomadas com os três sistemas foi também comparado com resultados obtidos por dois formalismos, de Gibbons1 e de Klein et al 2, bem conhecidos da literatura e que serviram para a validação dos resultados e deste trabalho. Destas comparações verificou-se que os ângulos do filtro dinâmico aprimorado que apresentam maior diferença entre resultados obtidos e os formalismos foram os ângulos de 45º e de 60º em relação ao método de Klein et al que apresentaram diferenças de 2,1% em relação ao BIS2G, 1,9% em relação à câmara de ionização e 1,1% em relação ao arranjo linear de câmaras de ionização.
The purpose of this work was to accomplish the dosimetry of Enhanced Dynamic Wedge (EDW). It was carried out with two intents: 1) the evaluation of the clinical implementation of the EDW considering its advantages and disadvantages, and 2) the comparison among the efficiency and the better applicability of three different dosimetric systems in a set with the EDW. The dosimetric systems used were: the Beam Image System (BIS2G), from Scanditronix-Wellhöfer, a cylindrical ionization chamber of 0.6 cm3, type Farmer, and an ionization chamber linear detector array, composed by 23 cylindrical ionization chamber of 0.14 cm3, type Farmer. With the BIS2G was possible to plot graphs of fluence that allowed to analyze the symmetry of the graphs of fluence, linearity of the dose, dose rate effect, and gravity effect, features directly connected to the quality of the working of the EDW, further on having been possible to calculate the wedge factor for symmetric fields. With the ionization chamber it was measured the PDP and it was calculated the wedge factor for symmetric and asymmetric fields (one advantage of EDW). With the array it was possible to plot graphs of isodosis that allowed the checking of the symmetry and the comparison with the results obtained with the BIS2G, besides the calculation of the wedge factor for symmetric fields. The wedge factors obtained from the measurements of the three systems were also compared with the results of the formalisms, of Gibbons1 and of Klein et al 2, both well known if the literature, and that were used to validate the results and the work. From these comparisons was verified that the angles of the EDW that show more difference between the results and the formalisms were the 46º and 60º angles related to the method of Klein et al, which show differences of 2.12% related to BIS2G, 1.92% related to the ionization chamber, and 1.10% related to ionization chamber linear detector array.

This thesis presents the results of an architectural study on the design of FPGA- based architectures for convolutional neural networks (CNNs). We have analyzed the memory access patterns of a Convolutional Neural Network (one of the biggest networks in the family of deep learning algorithms) by creating a trace of a well-known CNN architecture and by developing a trace-driven DRAM simulator. The simulator uses the traces to analyze the effect that different storage patterns and dissonance in speed between memory and processing element, can have on the CNN system. This insight is then used create an initial design for a layer architecture for the CNN using an FPGA platform. The FPGA is designed to have multiple parallel-executing units. We design a data layout for the on-chip memory of an FPGA such that we can increase parallelism in the design. As the number of these parallel units (and hence parallelism) depends on the memory layout of input and output, particularly if parallel read and write accesses can be scheduled or not. The on-chip memory layout minimizes access contention during the operation of parallel units. The result is an SoC (System on Chip) that acts as an accelerator and can have more number of parallel units than previous work. The improvement in design was also observed by comparing post synthesis loop latency tables between our design and one with a single unit design. This initial design can help in designing FPGAs targeted for deep learning algorithms that can compete with GPUs in terms of performance.

Genetic variation within the cannabinoid 1 receptor (CB1R) locus (CNR1) has been repeatedly associated with drug addiction pathologies. Genomic annotation of CNR1 indicates the vast majority of this genetic variation likely results in altered transcriptional regulation of the CNR1 gene as a mechanistic link to the disease phenotype. There is a lack of information describing the regulation of CNR1 transcription and the potential impact of disease-associated variation within the CNR1 locus on its transcriptional regulation. This study investigates the impact of an evolutionary conserved regulatory region of CNR1, termed ECR1, and the disease-associated variation contained within, on the transcriptional activity of the cognate CNR1 promoter region. Reporter assays conducted in primary hippocampal cells demonstrate that CNR1 promoter exhibits variable transcriptional activity during periods of CB1R signalling and cell depolarisation. Coupled to allelic variants of ECR1, the CNR1 promoter shows significant changes in transcriptional activity under resting conditions indicating that disease-associated variation within ECR1 may decrease CNR1 transcription. Further, alleles of ECR1 can drive allele-specific transcriptional responses from the CNR1 promoter during periods of CB1R stimulation and cell depolarisation. The results highlight the potential for disease-associated regulatory variation of the CNR1 locus to create stratified transcriptional responses to specific cell signalling scenarios and putatively to clinical strategies employing pharmacological agents. Furthermore, investigation of DNA-protein interactions at the allelic ECR1 region demonstrate that disease-associated variation within ECR1 alters DNA-protein interactions within the nucleus consistent with a decrease in transcriptional activity in the disease-associated allele variant. Collectively the current work supports the hypothesis that disease-associated variation within the ECR1 regulatory region of the CNR1 locus has the capacity to significantly impact on CNR1 promoter transcriptional activity. It is posited that allele-specific transcriptional effects may have a major impact on the susceptibility of individuals to drug addiction or on responses to clinical pharmacological treatments.

Les bacteries propioniques laitieres et en particulier celles de l'espece propionibacterium freudenreichii, jouent un role majeur dans l'etape de l'affinage des fromages a pate pressee cuite, type emmental par leur intervention dans l'ouverture de la pate et certainement dans la proteolyse du caille. Or la proteolyse due en partie a des peptidases intracellulaires implique la liberation de ces enzymes dans la pate par eclatement des bacteries nommees autolyse. Lors de cette these nous avons determine des parametres inducteurs de l'autolyse par une approche biochimique et ultrastructurale. Ainsi la capacite a l'autolyse depend de la souche consideree, du stade de croissance cellulaire, de l'absence de nutriment carbone, des conditions physicochimiques du milieu environnant (ph, nature des sels, temperature,). Les enzymes responsables de l'autolyse cellulaire (nommees autolysines), au nombre de 8 chez p. Freudenreichii, ont ete partiellement caracterisees (pm, resistance thermique, glycosylation). Des etapes de purification de ces autolysines ont ete franchies. Les resultats obtenus contribueront a un choix moins empirique des levains propioniques utilises en fromagerie

Breve introduzione alle reti neurali e al deep learning con descrizione dei sistemi utilizzati per i modelli e i test effettuati. Spiegazione del funzionamento dei sistemi creati ed esposizione dei risultati ottenuti.

A definite trend in Biomedical Imaging is the one towards the integration of increasingly complex interpretative layers to the pure data acquisition process. One of the most interesting and looked-forward goals in the field is the automatic segmentation of objects of interest in extensive acquisition data, target that would allow Biomedical Imaging to look beyond its use as a purely assistive tool to become a cornerstone in ambitious large-scale challenges like the extensive quantitative study of the Human Brain. In 2019 Convolutional Neural Networks represent the state of the art in Biomedical Image segmentation and scientific interests from a variety of fields, spacing from automotive to natural resource exploration, converge to their development. While most of the applications of CNNs are focused on single-image segmentation, biomedical image data -being it MRI, CT-scans, Microscopy, etc- often benefits from three-dimensional volumetric expression. This work explores a reformulation of the CNN segmentation problem that is native to the 3D nature of the data, with particular interest to the applications to Fluorescence Microscopy volumetric data produced at the European Laboratories for Nonlinear Spectroscopy in the context of two different large international human brain study projects: the Human Brain Project and the White House BRAIN Initiative.

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.

Quando um embalado é transportado, as pessoas que trabalham, residem ou circulam no itinerário percorrido pelo transporte de materiais radioativos são irradiadas, assim como aquelas pessoas que estão dentro dos veículos que viajam no mesmo sentido ou no sentido contrário. Desta forma, o trabalho proposto tem como principal objetivo estimar o impacto radiológico resultante do transporte de radiofármacos do IPEN - CNEN/SP até alguns destinos previamente definidos. Para isso, foram estimadas as doses nos indivíduos do público que estão nas ruas e nos veículos que trafegam próximos dos meios de transporte, ao longo do itinerário percorrido pelos embalados, durante a realização dos transportes dos radiofármacos. Foram estimadas as doses, também, para os trabalhadores, resultantes tanto da operação de condução do meio de transporte como das operações de carga e descarga dos embalados, já que estas tarefas são executadas pelo mesmo trabalhador.
When a package is shipped, people who work, live or travel in the route used for transportation of radioactive materials are irradiated, as well as those who are inside vehicles that travel in the same or opposite directions. Therefore, the proposed work has as main objective to estimate the radiological impact of the transportation of radiopharmaceuticals of IPENCNEN/ SP to some predefined destinations. The doses in individuals who are in the public streets and in vehicles that travel close to the means of transportation, along the route traveled by packages, during the transport of radiopharmaceuticals were estimated. The doses were also estimated for drivers, from both the operation of driving the vehicle, and the loading and unloading of packages, becouse these tasks are performed by the drivers.

Made available in DSpace on 2014-10-09T12:26:29Z (GMT). No. of bitstreams: 0
Made available in DSpace on 2014-10-09T14:04:25Z (GMT). No. of bitstreams: 0
Tese (Doutoramento)
IPEN/T
Instituto de Pesquisas Energeticas e Nucleares - IPEN-CNEN/SP

Streptococcus thermophilus est une bacterie lactique qui ne possede pas de protease de paroi et dont le systeme peptidasique a ete peu caracterise jusqu'a present. Notre travail a donc consiste a dresser un inventaire des peptidases presentes chez une souche de s. Thermophilus. Apres chromatographie d'echanges d'ions d'un extrait intracellulaire de la souche cnrz 302, nous avons mis en evidence la presence d'au moins 10 peptidases. Par la suite, deux aminopeptidases de cette souche, apii et apiii, ont ete purifiees. Il s'agit de deux metalloenzymes presentant une activite optimale a ph 7 et 8,5 a 36 et 60c, respectivement. Apii est une enzyme monomerique de masse moleculaire 97 kda alors qu'apiii est multimerique (360 kda). Apii est une aminopeptidase de large specificite alors qu'apiii est une glutamyl aminopeptidase specifique des acides amines acides. Elles presentent des homologies avec les aminopeptidases pepn (apii) et pepa (apiii) des lactocoques. Le gene codant pour une aminopeptidase generale de s. Thermophilus cnrz 302 a ensuite ete clone dans e. Coli. Cette aminopeptidase, surexprimee chez e. Coli, semble identique a l'aminopeptidase pepc des lactocoques. Nos resultats montrent que le systeme peptidasique de s. Thermophilus est tres analogue a celui des lactocoques

The topic of this paper is how CNN and BBC, two of the largest media companies in the world, presented Islam in the Palestinian civil war during the years 2006-2007. Articles that CNN and BBC published on the Palestinian civil war have been analyzed in order to answer this question. The purpose is to see if Islam is portrayed in an Islamophobic way by CNN and BBC and if it is possible to find discursive tracks from Clash of Civilizations-theory in the analyzed articles. The findings indicate that there are elements of Islamophobia and discursive tracks of Clash of Civilizations when it comes to presenting islam during the Palestinian civil war. Another conclusion is also that CNN and BBC presented islam in different ways during the civil war.

The interest in autonomous vehicles has recently increased and as a consequence many companies and researchers have begun working on their own solutions to many of the issues that ensue when a car has to handle complicated decisions on its own. This project looks into the possibility of relegating as many decisions as possible to only one sensor and engine control unit (ECU) — in this work, by letting a Raspberry Pi with a camera attached control a vehicle following a road. To solve this problem, image processing, or more specifically, machine learning’s convolutional neural networks (CNN) are utilized to steer a car by monitoring the path with a single camera. The proposed CNN is designed and implemented using a machine learning library for Python known as Keras. The design of the network is based on the famous Lenet, but has been downscaled to increase computation speed and to reduce memory size while still maintaining a sufficient accuracy. The network was run on the ECU, which in turn was fastened to a RC car together with the camera. For control purposes wires were soldered to the remote controller and connected to the Raspberry Pi. As concerns steering, a simple bang-bang controller was implemented. Glass box testing was used to assess the effectiveness of the code, and to guarantee a continuous evaluation of the results. To satisfy the network’s requirements in terms of both accuracy and computation speed larger experiments were performed. The final experiments showed that the network achieved sufficient accuracy and performance to steer the prototype car in real time tasks, such as following model roads and stopping at the end of the path, as planned. This shows that despite being small with moderate accuracy, this CNN can handle the task of lane-keeping using only the data of one single camera. Since the CNN could do this while running on a small computer such as the Raspberry Pi, it has been observed that using a CNN for a lane-keeping algorithm in an embedded system looks promising.
På senare tid så har intresset angående självkörande bilar ökat. Detta har lett till att många företag och forskare har börjat jobbat på sina egna lösningar till den myriad av problem som upstår när en bil behöver ta komplicerade beslut på egen hand. Detta projekt undersöker möjligheten att lämna så många av dessa beslut som möjligt till en enda sensor och processor. I detta fall så blir det en Raspberry Pi (RPI) och en kamera som sätts på en radiostyrd bil och skall följa en väg. För att implementera detta så används bildbehandling, eller mer specifikt, convolutional neural networks (CNN) från maskininlärning för att styra bilen med en enda kamera. Det utvecklade nätverket är designat och implementerat med ett bibliotek för maskininlärning i Python som kallas för Keras. Nätverkets design är baserat på det berömda Lenet men den har skalats ner för att öka prestandan och minska storleken som nätverket tar men fortfarande uppnå en anständing träffsäkerhet. Nätverket körs på RPIn, vilken i sin tur är fastsatt på en radiostyrd bil tillsammans med kameran. Kablar har kopplats och blivit lödda mellan RPIn och handkontrollen till radiostyrda bilen så att RPIn kan styra bilen. Själva styrningen lämnats åt en simpel "Bang Bang controller". Utvärdering av nätvärket och prototypen utfördes löpande under projektets gång, enhetstester gjordes enligt glasboxmetoden för att testa och verifiera olika delar av koden. Större experiment gjordes för att säkerställa att nätverket presterar som förväntat i olika situationer. Det slutgiltiga experimentet fastställde att nätverket uppfyller en acceptabel träffsäkerhet och kan styra prototypen utan problem när denne följer olika vägar samt att den kan stanna i de fall den behöver. Detta visar att trots den begränsade storleken på nätverket så kunde det styra en bil baserat på datan från endast en sensor. Detta var dessutom möjligt när man körde nätverket på en liten och svag dator som en RPI, detta visar att CNN var kraftfulla nog i det här fallet.

Pour obtenir des hauts taux de détection, les CNNs requièrent d'un grand nombre de paramètres à stocker, et en fonction de l'application, aussi un grand nombre d'opérations. Cela complique gravement le déploiement de ce type de solutions dans les systèmes embarqués. Ce manuscrit propose plusieurs solutions à ce problème en visant une coadaptation entre l'algorithme, l'application et le matériel.Dans ce manuscrit, les principaux leviers permettant de fixer la complexité computationnelle d'un détecteur d'objets basé sur les CNNs sont identifiés et étudies. Lorsqu'un CNN est employé pour détecter des objets dans une scène, celui-ci doit être appliqué à travers toutes les positions et échelles possibles. Cela devient très coûteux lorsque des petits objets doivent être trouvés dans des images en haute résolution. Pour rendre la solution efficiente et ajustable, le processus est divisé en deux étapes. Un premier CNN s'especialise à trouver des régions d'intérêt de manière efficiente, ce qui permet d'obtenir des compromis flexibles entre le taux de détection et le nombre d’opérations. La deuxième étape comporte un CNN qui classifie l’ensemble des propositions, ce qui réduit la complexité de la tâche, et par conséquent la complexité computationnelle.De plus, les CNN exhibent plusieurs propriétés qui confirment leur surdimensionnement. Ce surdimensionnement est une des raisons du succès des CNN, puisque cela facilite le processus d’optimisation en permettant un ample nombre de solutions équivalentes. Cependant, cela complique leur implémentation dans des systèmes avec fortes contraintes computationnelles. Dans ce sens, une méthode de compression de CNN basé sur une Analyse en Composantes Principales (ACP) est proposé. L’ACP permet de trouver, pour chaque couche du réseau, une nouvelle représentation de l’ensemble de filtres appris par le réseau en les exprimant à travers d’une base ACP plus adéquate. Cette base ACP est hiérarchique, ce qui veut dire que les termes de la base sont ordonnés par importance, et en supprimant les termes moins importants, il est possible de trouver des compromis optimales entre l’erreur d’approximation et le nombre de paramètres. À travers de cette méthode il es possible d’obtenir, par exemple, une réduction x2 sur le nombre de paramètres et opérations d’un réseau du type ResNet-32, avec une perte en accuracy <2%. Il est aussi démontré que cette méthode est compatible avec d’autres méthodes connues de l’état de l’art, notamment le pruning, winograd et la quantification. En les combinant toutes, il est possible de réduire la taille d’un ResNet-110 de 6.88 Mbytes à 370kBytes (gain mémoire x19) avec une dégradation d’accuracy de 3.9%.Toutes ces techniques sont ensuite misses en pratique dans un cadre applicatif de détection de vissages. La solution obtenue comporte une taille de modèle de 29.3kBytes, ce qui représente une réduction x65 par rapport à l’état de l’art, à égal taux de détection. La solution est aussi comparé a une méthode classique telle que Viola-Jones, ce qui confirme autour d’un ordre de magnitude moins de calculs, au même temps que l’habilité d’obtenir des taux de détection plus hauts, sans des hauts surcoûts computationnels Les deux réseaux sont en suite évalues sur un multiprocesseur embarqué, ce qui permet de vérifier que les taux de compression théoriques obtenues restent cohérents avec les chiffres mesurées. Dans le cas de la détection de vissages, la parallélisation du réseau comprimé par ACP sûr 8 processeurs incrémente la vitesse de calcul d’un facteur x11.68 par rapport au réseau original sûr un seul processeur
Recently, Convolutional Neural Networks have become the state-of-the-art soluion(SOA) to most computer vision problems. In order to achieve high accuracy rates, CNNs require a high parameter count, as well as a high number of operations. This greatly complicates the deployment of such solutions in embedded systems, which strive to reduce memory size. Indeed, while most embedded systems are typically in the range of a few KBytes of memory, CNN models from the SOA usually account for multiple MBytes, or even GBytes in model size. Throughout this thesis, multiple novel ideas allowing to ease this issue are proposed. This requires to jointly design the solution across three main axes: Application, Algorithm and Hardware.In this manuscript, the main levers allowing to tailor computational complexity of a generic CNN-based object detector are identified and studied. Since object detection requires scanning every possible location and scale across an image through a fixed-input CNN classifier, the number of operations quickly grows for high-resolution images. In order to perform object detection in an efficient way, the detection process is divided into two stages. The first stage involves a region proposal network which allows to trade-off recall for the number of operations required to perform the search, as well as the number of regions passed on to the next stage. Techniques such as bounding box regression also greatly help reduce the dimension of the search space. This in turn simplifies the second stage, since it allows to reduce the task’s complexity to the set of possible proposals. Therefore, parameter counts can greatly be reduced.Furthermore, CNNs also exhibit properties that confirm their over-dimensionment. This over-dimensionement is one of the key success factors of CNNs in practice, since it eases the optimization process by allowing a large set of equivalent solutions. However, this also greatly increases computational complexity, and therefore complicates deploying the inference stage of these algorithms on embedded systems. In order to ease this problem, we propose a CNN compression method which is based on Principal Component Analysis (PCA). PCA allows to find, for each layer of the network independently, a new representation of the set of learned filters by expressing them in a more appropriate PCA basis. This PCA basis is hierarchical, meaning that basis terms are ordered by importance, and by removing the least important basis terms, it is possible to optimally trade-off approximation error for parameter count. Through this method, it is possible to compress, for example, a ResNet-32 network by a factor of ×2 both in the number of parameters and operations with a loss of accuracy <2%. It is also shown that the proposed method is compatible with other SOA methods which exploit other CNN properties in order to reduce computational complexity, mainly pruning, winograd and quantization. Through this method, we have been able to reduce the size of a ResNet-110 from 6.88Mbytes to 370kbytes, i.e. a x19 memory gain with a 3.9 % accuracy loss.All this knowledge, is applied in order to achieve an efficient CNN-based solution for a consumer face detection scenario. The proposed solution consists of just 29.3kBytes model size. This is x65 smaller than other SOA CNN face detectors, while providing equal detection performance and lower number of operations. Our face detector is also compared to a more traditional Viola-Jones face detector, exhibiting approximately an order of magnitude faster computation, as well as the ability to scale to higher detection rates by slightly increasing computational complexity.Both networks are finally implemented in a custom embedded multiprocessor, verifying that theorical and measured gains from PCA are consistent. Furthermore, parallelizing the PCA compressed network over 8 PEs achieves a x11.68 speed-up with respect to the original network running on a single PE

Images are often corrupted by noise which reduces their visual quality and interferes with analysis. Convolutional Neural Networks (CNNs) have become a popular method for denoising images, but their training typically relies on access to thousands of pairs of noisy and clean versions of the same underlying picture. Unsupervised methods lack this requirement and can instead be trained purely using noisy images. This thesis evaluated two different unsupervised denoising algorithms: Noise2Self (N2S) and Parametric Probabilistic Noise2Void (PPN2V), both of which train an internal CNN to denoise images. Four different CNNs were tested in order to investigate how the performance of these algorithms would be affected by different network architectures. The testing used two different datasets: one containing clean images corrupted by synthetic noise, and one containing images damaged by real noise originating from the camera used to capture them. Two of the networks, UNet and a CBAM-augmented UNet resulted in high performance competitive with the strong classical denoisers BM3D and NLM. The other two networks - GRDN and MultiResUNet - on the other hand generally caused poor performance.

Changes in the expression of the cannabinoid receptor 1 (CB1) has been associated with the progression of many human diseases such as anxiety and depression, schizophrenia, epilepsy, Huntington’s disease, Parkinson’s disease, obesity and osteoporosis. However, little is known about the mechanisms that control CB1 expression or how they may be affected by human polymorphic variation to bring about disease susceptibility. In order to elucidate the regulatory systems controlling the tissue specific transcriptional regulation of the CNR1 gene, that encodes CB1, and the effects of human polymorphisms on these systems, this study combined comparative genomics, molecular biology, cell biology and pharmacology with human genetics. Comparative genomics was used to identify five evolutionary conserved regions (ECR) 5’ of the CNR1 gene that had been conserved between humans and birds (310 million years). Primary cell culture, using luciferase reporter constructs and transgenic studies suggested that some of these regions acted as enhancers which controlled gene expression in a tissue specific manner. In addition, it was demonstrated that the activity of these ECRs can be altered using different signal transduction agonists. More importantly, evidence is provided to suggest that ECR1 and the endogenous CNR1 promoter work synergistically in primary dorsal root ganglia (DRG) neurons and that ECR2 and the CNR1 promoter work synergistically in primary hypothalamic neurons in response to MAPK and ER agonists. Furthermore, a validated human SNP within ECR1 can interfere with a putative AP-1 transcription factor binding site when the major allele is present to block the up-regulation seen with a MAPK agonist in hippocampal neurons. The discovery of such novel cell specific regulatory pathways and the recognition of the cell specific effects of polymorphisms on these pathways may lead to a better understanding of how CNR1 mis-expression can contribute to disease and will enhance our ability to develop novel therapies.

Muitos pesquisadores e gerentes têm se empenhado na busca pela descoberta dos fatores que afetam o sucesso dos processos de mudança organizacional. Um dos fatores mais apontados na literatura sobre este tema diz respeito à resistência à mudança por parte de alguns atores organizacionais, em especial os pertencentes aos níveis hierárquicos mais baixos. Mas como será que estes indivíduos percebem a si mesmos e aos outros num processo de mudança organizacional? Como eles percebem o conteúdo e o gerenciamento da mudança? Qual o papel da comunicação nesse contexto? O presente trabalho tem como objetivo principal aprofundar a compreensão do papel da comunicação na construção do significado para os indivíduos e grupos em mudanças planejadas, e sua relação com aspectos comportamentais apresentados por estes. Para tanto, foi conduzido um estudo de caso longitudinal em uma empresa pública da esfera federal que vem passando por um processo de intervenção organizacional na área de Tecnologia da Informação com a implantação de um sistema de gerenciamento eletrônico de documentos e do fluxo de trabalho. A metodologia da pesquisa baseou-se na triangulação de métodos, ou seja, na análise de documentos oficiais da organização, aplicação de questionários e entrevistas semi-estruturadas aos empregados em dois momentos distintos, e na observação-participante do pesquisador utilizando-se o método Pesquisa- Ação (THIOLLENT, 1997 e 2005), cuja adoção justifica-se pelo alto grau de envolvimento do autor com o projeto em estudo. Os resultados comprovam que a mudança organizacional é sensível ao aspecto relacional e temporal e que a ação e o comportamento humano nas organizações devem ser pensados como frutos da comunicação enquanto fator determinante na construção dos significados e das identidades sociais. Os resultados suscitam também alguns questionamentos acerca da problemática e conflitos vividos pelos médios gerentes num contexto de mudança e apontam para a necessidade de atentarmos para o fato de que eles são agentes, mas também são gente.
A lot of researchers and managers have been trying to find the factors that affect the success of organizational change processes. One of the most mentioned factors in literature about this theme is the resistance to change on the part of some organizational actors, in special the ones belonging to the lower hierarchical levels. How these individuals realize themselves and the other in organizational change processes? How they realize the content and management of the change? What is the role of communication in a change context? This work aims to contribute to the comprehension of the communication role in the meaning construction for individuals and groups in planned organizational changes, as well as to understand its relationship with behavior aspects presented for them. For this purpose, it was conducted a longitudinal case study in a public federal organization that is performing an organizational change in the Technology of Information area: implantation of an electronic management system of documents and workflow. The research methodology was based on methods triangulation that comprehended the analysis of official organizational documents, application of questionnaires and semi-structured interviews to the employers in two different moments, and researcher participant observation by means of Action Research (THIOLLENT, 1997 and 2005) due to the high level of relationship between the researcher and the change project. The results showed that the organizational change is sensible to relational and temporal aspects and that human action and behavior in organizations should be thought as the results of communication as determinant to the meaning construction and social identity. The results also present some questionings about the conflicts experienced by middle managers in a change context and point to the necessity of realizing the fact that middle managers are change agents but they are also human beings.

A fim de implementar o novo código de prática da Agência Internacional de Energia Atômica (AIEA) no Brasil, o Laboratório de Calibração de Instrumentos (LCI/IPEN) está realizando calibrações em termos de kerma no ar e em termos de dose absorvida na água, em feixes de 60Co. As razões ND,w/NK obtidas são comparadas com valores da literatura, obtendo-se concordância satisfatória. As diferenças entre os valores das razões CK (ND,w/NK) obtidas no presente trabalho e os valores da literatura são devidas a vários fatores. Estes fatores podem ser as variações entre as câmaras, que devem ser objeto de diferenças nas incertezas estimadas pelos PSDLs, e as diferenças nos padrões utilizados por cada Instituto de Metrologia Nacional (NMI) ou Laboratório de Dosimetria. No entanto, se forem conhecidas as razões entre cada NMI e o BIPM, para os padrões de kerma no ar e de dose absorvida na água, é possível a normalização das razões CK medidas para as razões equivalentes ao BIPM. Todos os resultados de razões CK obtidos foram convertidos para as razões equivalentes ao BIPM para facilitar a comparação. Neste trabalho é discutida a utilização da razão CK como parâmetro de controle de qualidade na verificação de resultados das calibrações rotineiras. Para avaliação dos procedimentos de calibração adotados no LCI, foram realizados testes de estabilidade a longo prazo com as câmaras de ionização padrões. Além disso, foi criado um banco de dados para armazenamento dos dados de calibração, com o objetivo de melhorar a qualidade do procedimento de calibração.
In order to implement the new IAEA code of practice in Brazil the national calibration laboratories, the Laboratory of Instrument Calibration (LCI/IPEN) is calibrating clinical dosimeters in terms of both air kerma and absorbed dose to water in a 60Co gamma ray beam. The ND,w/NK ratios thus obtained are then compared with the literature values; a satisfactory agreement has been found. The differences between the CK values obtained in the present work and the literature values may be due to several components. These could be the chamber-to-chamber variations which may be subject to different estimated uncertainties by PSDLs, and the variations in the standards used by each National Metrology Institute (NMI) or Dosimetry Laboratory. However, since the ratio of both air kerma and absorbed dose to water standards for each NMI to those of the BIPM are known, it is possible to reduce the measured ratios to the expected BIPM equivalent ratios. All these CK results have been converted to the BIPM equivalent values to facilitate comparison. The use of the CK value as a and quality control parameter to verify the results of the routine calibrations is discussed in this work. The long term stability of wich standard ionization chamber was checked in order to evaluate the calibration procedures followed at LCI. Moreover, a database was created to store the calibration data aiming the quality improvement of the calibration procedure.

Made available in DSpace on 2014-10-09T12:52:05Z (GMT). No. of bitstreams: 0
Made available in DSpace on 2014-10-09T13:57:44Z (GMT). No. of bitstreams: 0
Dissertação (Mestrado)
IPEN/D
Instituto de Pesquisas Energeticas e Nucleares - IPEN/CNEN-SP

Made available in DSpace on 2014-10-09T12:46:40Z (GMT). No. of bitstreams: 0
Made available in DSpace on 2014-10-09T13:58:16Z (GMT). No. of bitstreams: 1 07535.pdf: 5658873 bytes, checksum: bfc9a035eb51ae184a489a4570c8c827 (MD5)
Dissertacao (Mestrado)
IPEN/D
Instituto de Pesquisas Energeticas e Nucleares - IPEN/CNEN-SP

Ce travail a été effectué au sein du département de valorisation de la recherche du plus grand organisme public de recherche français et européen: Le CNRS, Centre National de la Recherche Scientifique, Direction de la Politique Industrielle (DPI). A l'aune des théories économiques sur la croissance et l'innovation, des « nouvelles » politiques et réorganisations du système d'innovation national Français et du concept « d'Open Science », le Transfert de Technologie (TTO : Technology Transfer Office) de la Recherche Fondamentale Académique peine à trouver sa place. Ce travail se propose de discuter le paradigme de « l'inéfficience » du transfert de technologie en dégageant les causes endogènes et exogènes de la problématique de la valorisation de la recherche publique. Cette thèse montre l'existence d'un réel « Paradoxe de la valorisation et de l'innovation» en éclairant deux éléments fondamentaux de la difficulté du transfert de technologie public: la structure de la connaissance et le processus de création (genèse) des inventions. A partir de résultats empiriques, quantitatifs et qualitatifs, cette thèse propose une autre vision du transfert de technologie public. Ce modèle constitue une approche de « vectorisation quantique » du transfert de technologie. Il s'inspire de la théorie de construction de « La Pensée Vivante » (Rohrbach -1993), et de « l'Esprit Scientifique » (Bachelard -1961) sur le mode de construction de la pensée: constituée de la Pensée Mémoire, la Pensée Inventive, la Pensée Réflective et la Pensée Créatrice. Il s'inspire également de la «Gestion des savoirs» (Knowledge Management, Nonaka -1994 et Prax -2001) et nous permet de proposer les éléments de compréhension du processus de genèse de l'invention
This thesis a been performed at CNRS (Centre National de la Recherche Scientifique) in Directorate of Industrial Policy. If we take in consideration the economic theories on growth and innovation, the "new" politics and innovation of the French national system of innovation and the concept of "Open Science", the Technology Transfer Office (TTO) of the fundamental academic research struggle to find its place. This work offers to adjust the paradigm of the "inefficience" of the academic technology transfer by bringing out the endogenous and exogenous causes of the problematic of the "valorisation" of public research. We describe a real « Paradox ofValorisation and Innovation » by focusing on two fundamental facts of the difficulty of the academic technology transfer: the structure of knowledge and the proccss of creation (genesis) of the inventions. These empirical, quantitative and qualifying results claim for anothcr vision of the public technology transfer. This model constitutes an approach of "Quantic Vectorisation" of the transfer of technology. It is inspired by the theory of construction of the "La Pensée Vivante" ("Living Thought" - Rohrbach -1993) and by the "L'Esprit Scientifique" ("Scientific Spirit" - Bachelard -1961) on the way of construction of the thinking: made by "La Pensée Mémoire" (Memory Thinking), "La Pensée Inventive" (Inventive Thinking), "La Pensée Réflective" (Reflective Thinking) and "La Pensée Créatrice" (Creative Thinking). It is also inspired by the "Knowledge Management" (Nonaka -1994 and Prax -2001) and it allows us to propose elements to understand the process of the genesis of the invention

Författare: Magnus Gustafsson Niclas Hagel

Handledare: Thomas Knoll

Examinator: Martin Danielsson

Titel: Al-Jazeera och CNN - En jämförande fallstudie i krigsjournalistik

Typ av rapport: C - uppsats

Ämne: Medie- och Kommunikationsvetenskap

År: Höstterminen 2008

Sektion: Sektionen för Hälsa och Samhälle

Syfte: Vårt syfte är att studera och jämföra al-Jazeeras och CNN:s

bevakning av en händelse i Afghanistankonflikten för att kunna

redogöra för eventuella skillnader. Vi vill se hur olika faktorer

påverkar journalistiken. En analys ur ett genusperspektiv

kommer också att göras.

Metod: Fallstudie har tillämpats som huvudsaklig metod och vid analys

av material har innehållsanalys och kritisk diskursanalys använts.

Slutsatser: Efter att ha jämfört de två nyhetskanalerna kan vi tydligt se att

det finns stora skillnader i rapporteringen av ett amerikanskt

flyganfall mot en afghansk by. CNN som amerikansk

nyhetskanal visar att deras rapportering påverkas av det

amerikanska medieklimatet där en neutral krigsrapportering kan

ses som stötande och journalister ständigt utsätts för

påtryckningar. Ur ett genusperspektiv ser vi dock tydliga

likheter mellan kanalerna.

En svårighet med att använda Artificiell Intelligens, är resurserna som krävs för att utföra beräkningarna under en acceptabel tidsram, men också med en bra träffsäkerhet. Målet med denna uppsats är att jämföra olika modeller av convolutional neural networks, mellan träffsäkerhet och hastighet, för att hitta den modell som är mest effektiv. Dessutom evalueras den mest effektiva modellen genom en webblösning, som kan markera bilder med text. Resultatet visar att varje modell har olika fördelar i hastighet och träffsäkerhet, men att VGG16 har nära till bäst resultat utan de problem som andra modeller har.

Alex Krizhevsky and his colleagues changed the world of machine vision and image processing in 2012 when their deep learning model, named Alexnet, won the Im- ageNet Large Scale Visual Recognition Challenge with more than 10.8% lower error rate than their closest competitor. Ever since, deep learning approaches have been an area of extensive research for the tasks of object detection, classification, pose esti- mation, etc...This thesis presents a comprehensive analysis of different deep learning models and architectures that have delivered state of the art performances in various machine vision tasks. These models are compared to each other and their strengths and weaknesses are highlighted. We introduce a new approach for human head and shoulder detection from RGB- D data based on a combination of image processing and deep learning approaches. Candidate head-top locations(CHL) are generated from a fast and accurate image processing algorithm that operates on depth data. We propose enhancements to the CHL algorithm making it three times faster. Different deep learning models are then evaluated for the tasks of classification and detection on the candidate head-top loca- tions to regress the head bounding boxes and detect shoulder keypoints. We propose 3 different small models based on convolutional neural networks for this problem. Experimental results for different architectures of our model are highlighted. We also compare the performance of our model to mobilenet. Finally, we show the differences between using 3 types of inputs CNN models: RGB images, a 3-channel representation generated from depth data (Depth map, Multi-order depth template, and Height difference map or DMH), and a 4 channel input composed of RGB+D data.

Cannabinoid receptors play an important role in regulating bone mass and bone turnover. Studies in our laboratories have shown that young mice lacking type 1 cannabinoid receptor (CNR1-/-) had increased bone mass and were resistant to ovariectomy-induced bone loss. Other workers have reported that type 2 cannabinoid receptor knockout mice (CNR2-/-) develop age-related osteoporosis. The aim of this PhD thesis was to further investigate the role of CNR2 in bone metabolism in vitro and in vivo, using genetic and pharmacological approaches. This study showed that CNR2-/- mice had normal bone mass and bone turnover at 3 months of age, but following ovariectomy, CNR2-/- mice were partially protected from bone loss, because of a mild defect in osteoclast formation and bone resorption. In keeping with this, studies in vitro showed that RANKL-stimulated bone marrow cultures from CNR2-/- mice had fewer osteoclasts than cultures from wild type littermates. The CNR2-selective antagonist/inverse agonist AM630, inhibited osteoclast formation in wild type bone marrow cultures in vitro and prevented ovariectomy-induced bone loss in wild type mice in vivo. In contrast, osteoclast cultures from CNR2-/- mice were resistant to the inhibitory effects of AM630 at low concentrations and CNR2-/- ovariectomised mice did not respond to its protective effects at low doses, consistent with a CNR2- mediated effect. These results indicate that CNR2 regulates bone loss under conditions of increased bone turnover, such as ovariectomy, by affecting osteoclast differentiation and function. CNR2-deficient mice developed accelerated age-related osteoporosis and by 12 months of age they had a significant reduction in osteoblast numbers and bone formation, whereas osteoclast numbers remained comparable to wild type littermates. In agreement with this, osteoblasts derived from bone marrow of CNR2-/- mice had reduced PTHstimulated alkaline phosphatase activity and ability to form bone nodules, when compared with wild type cultures. The CNR2-selective agonist, HU308, stimulated bone nodule formation in wild type calvarial osteoblast cultures in vitro and reversed ovariectomy-induced bone loss in wild type mice in vivo. HU308 had blunted effects on bone nodule formation in cultures from CNR2-/- mice and no significant effects on ovariectomy-induced bone loss in CNR2-/- mice, indicating a CNR2-mediated effect. These studies demonstrate that CNR2 protects against age-related bone loss by mainly enhancing osteoblast differentiation and bone formation. In conclusion, type 2 cannabinoid receptors protect from bone loss by maintaining bone remodelling at balance. In addition, type 2 cannabinoid receptor agonists show evidence of anabolic activity, whereas antagonists/inverse agonists show evidence of antiosteoclastic activity in vitro and in vivo.

Ce document comporte trois parties qui synthétisent environ dix années de mes recherches. J'ai voulu mettre en valeur mes activités de recherche et d'administration de la recherche principales en faisant apparaître les collaborations que j'ai pu avoir avec d'autres laboratoires en France ou à l'étranger, ainsi que les travaux des doctorants que j'ai eu l'occasion de co-encadrer. Il est structuré de la façon suivante. Dans la première partie du document je présente un rapport de mes activités nationales et internationales ainsi qu'une liste complète de mes publications. Dans la deuxième partie je développe les travaux de recherches les plus importants que j'ai effectué depuis mon entrée au CNRS. Enfin, dans la troisième partie sont insérées quelques unes de mes publications significatives utiles pour la compréhension de mes travaux.

Na presente dissertação, foi desenvolvido um equipamento para tomografia com nêutrons que está operacional e instalado no canal de irradiação 14 do Reator Nuclear de Pesquisas IEA-R1 do IPEN-CNEN/SP. As imagens apresentadas neste trabalho, são de objetos que foram selecionados de modo a realçarem uma das principais aplicações da técnica, que é o estudo de materiais hidrogenados mesmo se envoltos por espessa camada de alguns metais. Neste equipamento, uma tomografia completa pode ser obtida em 400 s, com uma resolução espacial máxima de 205 μm. Estas características são comparáveis às dos equipamentos mais desenvolvidos em operação em outros países, e propiciam imagens com qualidade suficiente para que sejam realizadas análises tanto qualitativas quanto quantitativas dos objetos inspecionados. A implementação da técnica da tomografia com nêutrons abre a possibilidade de novas linhas de pesquisa, pois disponibiliza uma nova ferramenta para inspeção de objetos, que fornece uma visão da sua estrutura interna, que nem sempre é possível por métodos de imageamento bidimensional.
In the work reported in this dissertation, a facility for neutron tomography was developed and installed at the irradiation channel #14 of the IEA-R1 nuclear research reactor of IPEN-CNEN/SP. Several selected objects were inspected, and the obtained images demonstrate the main characteristic of the present technique that is its capability to visualize hydrogenous rich substances. In such facility, a tomography can be obtained in 400 s with a spatial resolution of 205 μm, and the obtained images have sufficient quality to allow qualitative and quantitative analysis. These characteristics are very similar to the ones of the top facilities around the world, and the quality of the provided images are sufficient to allow qualitative and quantitative analysis of the inspected object. The implementation of the neutron tomography technique opens up the possibility of new research as it provides a new tool for inspection of objects, which provides a view of its internal structure, which is not always possible for two-dimensional imaging methods.

Made available in DSpace on 2014-10-09T12:49:08Z (GMT). No. of bitstreams: 0
Made available in DSpace on 2014-10-09T14:01:10Z (GMT). No. of bitstreams: 1 09815.pdf: 1279672 bytes, checksum: b0b13ac00bb4c31120dba8064c665328 (MD5)
Dissertacao (Mestrado)
IPEN/D
Instituto de Pesquisas Energeticas e Nucleares - IPEN/CNEN-SP

Submitted by Claudinei Pracidelli (cpracide@ipen.br) on 2016-06-22T12:49:00Z No. of bitstreams: 0
Made available in DSpace on 2016-06-22T12:49:00Z (GMT). No. of bitstreams: 0
Dissertação (Mestrado em Tecnologia Nuclear)
IPEN/D
Instituto de Pesquisas Energeticas e Nucleares - IPEN-CNEN/SP

Made available in DSpace on 2009-11-18T18:56:47Z (GMT). No. of bitstreams: 1 ACFD99.pdf: 1217256 bytes, checksum: b8516b0a13b0a920b33f9d7ff454ffcb (MD5) Previous issue date: 2007
This research had the objective to verify if the current organization design of the Comissão Nacional de Energia Nuclear (CNEN) facilitates or makes it difficult the reach of the organization performance. In order to make the analysis, the used ways had been bibliographic, documental and field researches. The bibliographical survey allowed to conclude that it has a necessity of searching a balance between the objectivist and subjectivist approaches of the theory of the organization design. The documental survey supplied elements that they make possible to understand the main phases of the development of the CNEN in elapsing of the time and that they explain the current organization design. Finally, the research of field, carried through with the main managers of the institution, made possible the data-collecting that contributed to ratify the initial assumption of the research of that the current organization design of the CNEN makes it difficult the performance of the organization.
Esta pesquisa teve por objetivo verificar se a atual modelagem organizacional da Comissão Nacional de Energia Nuclear (CNEN) facilita ou dificulta o alcance do desempenho organizacional. Para possibilitar a análise, os meios utilizados foram as pesquisas bibliográfica, documental e de campo. O levantamento bibliográfico permitiu concluir que há uma necessidade de se buscar um equilíbrio entre as vertentes objetivista e subjetivista da teoria da modelagem organizacional. A pesquisa documental forneceu elementos que possibilitaram compreender as fases principais do desenvolvimento da CNEN no decorrer do tempo e que explicam a atual modelagem organizacional. Por fim, a pesquisa de campo, realizada com os principais gerentes da instituição, possibilitou o levantamento de dados que contribuiu para ratificar a suposição inicial da pesquisa de que a atual modelagem organizacional da CNEN dificulta o desempenho da organização.

Thesis (M.A.)--Georgia State University, 2005.
Title from title screen. Merrill Morris, committee chair; Marian Meyers, Douglas Barthlow, committee members. Electronic text (98 p.) : digital, PDF file. Description based on contents viewed June 21, 2007. Includes bibliographical references (p. 89-96).

News provides us with information about our world so we can make decisions about the matters that affect our daily lives—both for our personal and the public good. Television news is a pervasive force in our society, and it is important to study because of the influence it exerts on human action. But news is produced by human beings, and those human beings must make selections and rejections regarding what makes it into a newscast and what doesn’t. In addition, decisions have to be made on how to frame, present, order, word, edit, shape what news items are included. Many forces influence these decisions throughout the complex television news process. Media sociology scholars urge researchers to examine these influences at five levels: the individual, newsroom, organization, extra-organization and societal or cultural levels. This gatekeeping study examined this complex news process at work and revealed the complex set of forces that influence news decisions by news producers at CNN, a global 24-hour news network. By exposing the processes by which the news is made, one can better understand the influences that shape the end product—the news.

Detta projekt studerar effekterna av transfer learning på inhämtandet av information från CNN-baserade ljuddatarepresentationer. Flera otränade CNN- nätverk matas med melspektrogrammatriser och tränas på tre olika uppgifter, nämligen ’genre’, ’region’ och ’year’ och klassifikationsprestandan mäts. Efter detta appliceras transfer learning och klassifikationsprestandan mäts igen. F1- score för individuella klasser inom de olika uppgifterna mäts också. Genom att jämföra resultaten visas att transfer learning är applicerbart på denna domän.
This project studies the effects of transfer learning on music information retrieval tasks of CNN-based audio data representations. Several neural networks are fed melspectrogram matrices and trained with random initial weights on three different classification tasks including ’genre’, ’region’ and ’year’ and classification performance is measured, after which transfer learning is utilized and classification performance is measured again. F1-score for individual classes within the different tasks is also measured. Comparing the results shows that transfer learning is applicable in this task domain.

Mestrado em Engenharia Eletrónica e Telecomunicações
O presente trabalho propõe-se implementar duas ferramentas em software, um deles para aquisição e processamento de um sinal de satélite a 20GHz e canais meteorológicos, o outro para processamento de um sinal de satélite a 40GHz. Em primeiro lugar desenvolveu-se o software que recebe o sinal de 40GHz, denominado detetor digital, em MATLAB. A aquisição dos dados é feita com base num recetor de rádio digital, o processamento dos dados é feito através de deteção espetral (FFTs) e uma malha de sincronização de fase (PLL). A par da aquisição de dados é feito também um apontamento periódico da antena de receção. Todos os dados processados são guardados em ficheiros de saída devidamente estruturados e datados. De seguida desenvolveu-se o software de aquisição e processamento do sinal de satélite a 20GHz, denominado software de aquisição, em LabVIEW. Neste caso, a aquisição do sinal de satélite é feita através de um recetor superheteródino que converte o sinal para a banda de áudio, sinal que depois é digitalizado através de uma placa de aquisição de dados. Os canais meteorológicos são também digitalizados na mesma placa. O software de aquisição recebe os dados provenientes da placa de aquisição e, para fazer o processamento do sinal utiliza deteção espetral (FFTs), para fazer o processamento dos dados meteorológicos, usa apenas médias reduzir a taxa de amostragem. Todos os dados processados são guardados em ficheiros de saída devidamente estruturados e datados. Finalmente, são analisados alguns resultados obtidos pelo detetor digital, por forma a comparar o desempenho dos métodos de deteção.
The aim of this thesis is to implement two software tools: one of them is to acquire and process a satellite beacon signal at 20GHz alongside with some meteorological channels, the other is to process a satellite beacon signal at 40GHz. The first one, developed in MATLAB, is the software for the 40GHz beacon, named Digital Detector. The data acquisition is done with a Software Digital Radio (SDR) receiver, the data processing is done through spectral detection (FFTs) and a phase locked-loop (PLL). In parallel with data acquisition, the digital detector is capable of, periodically, control the antenna pointing system. All processed data is stored in output files properly structured and dated. The next step was the development, in LabVIEW, of the acquisition and processing software for the 20GHz beacon, named acquisition software. In this case, the acquisition of the satellite signal is made using a superheterodyne receiver which converts the signal to the audio band; the signal is then digitized by a data acquisition board. The weather channels are also scanned on the same board. The acquisition software receives data from the acquisition card and uses spectral detection (FFTs) to detect the signal. Regarding the processing of meteorological data, it displays only averages to control the data rate. All processed data is stored in output files properly structured and dated. Finally, some results obtained with the digital detector are analyzed in order to compare the performance of both detection methods implemented.

An interesting but largely unanswered question in biology is how eukaryotic organisms precisely regulate the size of multicellular tissues or groups of cells. Developing Dictyostelium cells aggregate in dendritic streams using relayed pulses of adenosine 3'5'-cyclic monophosphate (cAMP) as a chemoattractant to form ∼20,000-cell fruiting bodies. Counting factor (CF), a secreted protein complex, regulates group size by increasing cell motility and decreasing cell-cell adhesion to induce the breakup of excessively large aggregation streams. We used a second-site suppressor screen by conducting random insertional mutagenesis in smlA- cells which over-secrete CF to search for CF signal transduction components, and found that an insertion in the cnrN gene affects group size. Cells lacking CnrN (cnrN- cells) form small aggregation territories with few streams, which then form small fruiting bodies. Expressing CnrN in cnrN- cells rescues the abnormal phenotype. Computer simulations suggested that in the absence of stream formation, CF should be unable to affect group size. As predicted, cnrN- group size is insensitive to the addition or depletion of CF, even though CF regulates the motility of cnrN - cells. cnrN- cells have excessively large cAMP-stimulated cAMP pulses, and the small territory phenotype can be rescued by developing cells in the presence of the cAMP-hydrolyzing enzyme cAMP phosphodiesterase or simply by starving cells at low densities. The predicted amino acid sequence of CnrN has similarity to phosphatase and tensin homologs (PTENs). PTENs play a key role in inhibiting phosphatidylinositol 3' kinase (PI3K) dependent pathways. In Dictyostelium, cAMP pulses activate PI3Ks, and activated PI3Ks in turn stimulate adenylyl cyclase to produce cAMP. As indicated by the sequence similarity of CnrN to PTEN, in response to cAMP stimulation, cnrN- cells show elevated and extended activation of PI3K-dependent pathways, including PIP3 accumulation, Akt activation, actin polymerization, and adenylyl cyclase-catalysed cAMP production. Our results suggest that CnrN has functional similarity to PTEN and regulates the cAMP pulse size by negatively regulating PI3K-dependent pathways, and that CnrN-mediated regulation on territory and stream formation may use a CF-independent mechanism.

Le système endocannaboïde (eCB) est constitué des ligands, des récepteurs – les plus étudiés étant les récepteurs CB1 et CB2 – et les enzymes de synthèse et de dégradation. Les ligands étant lipophiles, ils ne sont pas encapsulés dans des vésicules, ce qui place les enzymes de synthèse et de dégradation dans une position de régulateurs clés. Plusieurs études démontrent une participation du système eCB à des processus de développement dans le système nerveux central (SNC). La rétine est un modèle important pour l’étude de ces processus car elle contient plusieurs types cellulaires bien connus, dont le patron de développement est clairement établi. Pour l’instant très peu est connu sur l’expression du système eCB durant le développement rétinien. C’est dans ce cadre que les patrons d’expression du récepteur CB1 et de l’enzyme de dégradation FAAH ont été étudiés pendant le développement rétinien postnatal chez le rat. Pour identifier les types cellulaires exprimant ces protéines, des co-marquages ont été accomplis pour le récepteur CB1 ou FAAH et des marqueurs des types cellulaires rétiniens. À P1, les cellules ganglionnaires, amacrines, horizontales et mitotiques expriment le récepteur CB1. Les cellules ganglionnaires et amacrines cholinergiques sont FAAH-positives. Au cours du développement, certains types cellulaires démontrent une expression transitoire de ces deux protéines, suggérant une implication du système eCB dans les processus de développement. Nos données démontrent également une importante expression du système eCB dans la rétine adulte, ce qui soutient l’hypothèse de son implication dans la réponse rétinienne. En bref, des études fonctionnelles in vitro sur des rétines de non-mammifères ont révélées que le récepteur CB1 modulait la réponse des cônes et des cellules bipolaires. Malgré la récente démonstration de sa présence dans la rétine, il n’existe pas de d’étude sur le rôle du récepteur CB2 dans la rétine. Dans cette thèse, les conséquences fonctionnelles de l’élimination des récepteurs CB1 ou CB2 ont été évaluées chez des souris transgéniques. Les réponses rétiniennes ont été enregistrées par électrorétinographie chez des souris cnr1-/- (CB1R-KO) et cnr2-/- (CB2R-KO). Nos données suggèrent une implication différente pour chaque récepteur dans la formation de la réponse rétinienne
The endocannabinoid (eCBs) system is composed of the ligands, the receptors - the most studied are CB1R and CB2R – and the synthesizing and degradative enzymes. The lipophilic ligands are not stored in vesicles, thereby placing the synthesizing and degradative enzymes as key regulators of the receptor function. The eCB system is thought to participate to developmental processes in the central nervous system (CNS). The rodent retina is a valuable model to study CNS development, as it contains well identified cell types with established developmental timelines. Very little is known about the distribution of this neuromodulationsystem in the developing retina. In this thesis, the expression patterns of CB1R and eCB degradative enzyme FAAH were investigated in the rat retina during postnatal development. To identify the cells expressing these proteins, co-stainings were carried out for CB1R or FAAH and retinal cell type markers. At P1, CB1R was expressed in ganglion, amacrine, horizontal and mitotic cells, whereas FAAH was present in ganglion and cholinergic amacrine cells. In the course of development, both CB1R and FAAH were transiently expressed in some cell type, suggesting a role of the eCB system in developmental processes. Furthermore, our data demonstrated an important expression of both proteins in adult animals, supporting the hypothesis that the eCB system is involved in retinal functions. Briefly, functional in vitro studies on non-mammalian retinae have revealed an effect of CB1R on cone photoreceptors and bipolar cells response. Despite the recent demonstration of CB2R mRNA and protein presence in the retina, there are no data on CB2R functional role in retina have been published. In this thesis, the consequences of removing either CB1R or CB2R from the retina of transgenic mice were evaluated. Retinal response was recorded by electroretinogram in cnr1-/- (CB1R-KO) and cnr2-/- (CB2R-KO) mice. This data suggests that both receptors are involved in shaping the retinal response to light and they have different roles in this process.

碩士
國立臺灣大學
資訊工程學研究所
105
With the large size of the convolutional neural network (CNN), performance and energy efficiency of CNN accelerator become an important problem. From previous works, we can find that DRAM accesses took a large part in energy consumption. To reduce DRAM accesses, we observe the computation behavior of convolutional layer, and many parameters are shared between computation. Those data may be loaded on-chip repeatedly with the limitation of on-chip buffer size in an accelerator. We would like to capture data reuse via the on-chip buffer to reduce DRAM accesses of CNN computation. There are three kinds of data reuse can be captured, and those data will be kept by on-chip buffer and be evicted when not needed. The first kind of data reuse is input feature map reuse, the next is filter reuse and the other is intermediate feature map reuse. Each layer in a CNN model may favor different data reuse policy based on the size of its input, output, and filters. But existing CNN accelerators only focus on one type of data reuse through CNN processing. To have flexibility using different data reuse policy for each layer in CNN processing, we would like to propose a reconfigurable CNN accelerator design, which can be configured to capture different types of reuse with the objective of minimizing off-chip memory accesses. With separating the CNN processing into several computation primitives which are units of convolution with different inputs and filters, we can reuse different data by arranging the computation ordering of those computation primitives in our accelerator. And our accelerator will execute based on the instructions generated by off-line generator considering the optimal reuse policy and hardware constraints. Our work shows that with our reconfigurable design, DRAM accesses can be reduced, and compare the execution time and the energy when using different data reuse policy. We also analyze the effect of the different configuration in our CNN accelerator design.

碩士
元智大學
通訊工程學系
104
Purpose This is an attempt to unravel the problem of human face keypoints recognition. In the new area of machine learning research called deep learning. Different approaches to this problem were evaluated and proposed one system to implement using python libraries for computational skills. Methodology Face keypoints detection was achieved by using a template algorithm. Using GPU instances and convolutional networks consisting of multiple levels. The key idea is to pre-train models in completely unsupervised way and finally they can be fine-tuned for the task at hand using supervised learning using Nesterov Gradient Algorithm to create the perceptron units. Manual detection was used to test implemented face keypoints recognition system. Findings Successful results were obtained for automated face keypoints recognition under robust and controlled conditions. The experimental results show that the model provides better results than publicly available benchmarks for the dataset. Originality/Value Discuss different machine learning techniques used for face keypoints detection and provide a description why most algorithms are based in neural networks. Keywords Convolution Neural network, face keypoints recognition.

碩士
國立中央大學
資訊管理學系在職專班
107
This paper proposes a set of "CNN-based identity recognition system" for identity recognition using a computer vision library OpenCV and deep learning technology and webcam. It is expected to be applied to access control and regional security. Monitoring, advertising, or other related systems that need to be enhanced by confirming their identity. This thesis is based on Python and TensorFlow's built-in GoogLeNet CNN model. Supervised learning is used to obtain facial image features and classified by identity. This paper uses self-organizing face image data and compares GoogLeNet. The identification rate of the three versions of the model, in the neural network architecture with the highest recognition rate in the experiment, can increase the recognition rate by adding the residual network experiment. Using the neural network model of the above-mentioned best recognition rate, the OpenCV is used to load the movie to instantly recognize the character in the film to verify the practicability of the neural network model of the research training. In the verification part of the results, the paper has self-organized 14 public figures, and each public figure has at least 130 face images as training and test and verification samples, among which the best recognition rate of the neural network is in 1260 images. The recognition rate of the training sample is 100%, and the image recognition rate of the 450 images is 99.11%. The time of the instant image recognition from the face image in the film to the completion identity is about 0.1 second.

碩士
國立中央大學
資訊工程學系
107
The visual is one of the most important senses that a human receives outside information. The visual helps us explore the world, receive new knowledge, and communicate with computer. As contactless human-computer interaction (HCI) model continues to develop, the technology of communicating with gaze behavior has become a highlight in this field. There have been many applications in the fields of education, advertising, nursing, entertainment or virtual reality. In general, most of the eye tracking devices need calibration in advance or fixing head. There are still many restrictions on usage specification. To solve the above problems, this study uses the ResNet model as the core of classification to construct Gaze Block Estimation Model (GBE Model). It can estimate the gaze block of the user without calibration process. Moreover, only an RGB camera device without depth information is used to capture the image, such as a webcam, a built-in camera on a laptop, or front-facing camera of a smartphone. The deep learning approach is data-driven. It needs a large amount of correctly labeled training data to train a stable and compliant model. However, the existing public dataset of visual behavior has different application scenarios. Resulting in images of the dataset does not apply to all application domains. Therefore, this study collects and builds up to a dataset of eye images of up to 300000 images. According to the experimental results, the GBE Model can estimate gaze block of the user without calibration process and allow the head moving. Even in the real-life testing, it can reach 85.1% accuracy. The experimental results prove the proposed method can let user use gaze block to control the screen, and achieve the goal of HCI application scenario.

碩士
逢甲大學
水利工程與資源保育學系
107
During the last two decades, the application of artificial intelligence in the field of flood forecasting has increased noticeably. Since the information of flood forecasting is the most important part of disaster management, also the emergency response and the mechanism of Recurrent Neural Network (RNN) include the behavior of the time series, this study attempt to adopt the Gated Recurrent Unit (GRU) which is a type of RNN used to develop a rainfall-runoff model for the mentioned purpose above. In this research RNN is using Gated Recurrent Unit (GRU). In each field, applicability of GRU is still in researching. Thereby, this paper will discuss the application GRU in the flood forecast. In order to improve the prediction accuracy of the GRU, the data is processed by using the Convolutional Neural Network (CNN) and then input into the GRU for prediction, called CNN-GRU. In the past, most studies used to extract every rainfall from the data before learning artificial neural networks for flood flow prediction. However this study will use a different approach, because GRU cell can remember the status from past. In addition, optimal hyperparameters setting for artificial neural networks will be found by genetic algorithm (GA) to modeling Dali River hourly rainfall-runoff model. Evaluation indicators show that CNN-GRU is better than GRU, the evaluation indicators show that CNN-GRU is better than GRU, because CNN-GRU uses CNN to extract eigenvalues from input data before using GRU for prediction.

碩士
國立臺灣大學
資訊工程學研究所
106
In Mandarin Chinese system, the tone plays an important role. Different tone patterns of the same syllable may result in different meanings. People whose native language aren’t Mandarin can be distinguished by their tone patterns. Therefore, we propose a method for tone classification. First, we convert the audio signal into the spectrogram. We treat the spectrogram as images, apply them as the image inputs for image recognition convolutional neural networks, and create tone classification models. We compare different image recognition models for tone classification. This approach can achieve good accuracy without too many processes on the audio signal. The tone classification architecture can be applied to Chinese teaching methods which will lead to educational success.

碩士
國立臺灣海洋大學
資訊工程學系
107
This paper proposes a method that uses deep learning and convolution neural network (CNN) for handwritten invoice recognition, this method can help enterprises solve that enterprises use only handwritten invoices and reduce labor costs of sorting invoices. Invoice recognition contains invoice number, buyer's government uniform invoice number, seller's government uniform invoice number, digital total amount, and Chinese total amount. Models train by different content, analyze and calculate the best results based on the labels, coordinates and scores of the model detection results. Besides, total amount result use digital total amount and Chinese total amount to correct, which increase 3% accuracy of total amount. The experiment use about 500 labeled invoices to train models, use models to recognize that randomly selected 1000 invoices, according to research results, the overall recognition accuracy over 95%.