• Refine Query
  • Source
  • Publication year
  • to
  • Language
  • 4
  • 1
  • 1
  • 1
  • 1
  • 1
  • Tagged with
  • 9
  • 9
  • 3
  • 3
  • 3
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 1
  • 1
  • 1
  • 1
  • About
  • The Global ETD Search service is a free service for researchers to find electronic theses and dissertations. This service is provided by the Networked Digital Library of Theses and Dissertations.
    Our metadata is collected from universities around the world. If you manage a university/consortium/country archive and want to be added, details can be found on the NDLTD website.
1

A reliability assessment methodology for distribution systems with distributed generation

Duttagupta, Suchismita Sujaya 16 August 2006 (has links)
Reliability assessment is of primary importance in designing and planning distribution systems that operate in an economic manner with minimal interruption of customer loads. With the advances in renewable energy sources, Distributed Generation (DG), is forecasted to increase in distribution networks. The study of reliability evaluation of such networks is a relatively new area. This research presents a new methodology that can be used to analyze the reliability of such distribution systems and can be applied in preliminary planning studies for such systems. The method uses a sequential Monte Carlo simulation of the distribution system’s stochastic model to generate the operating behavior and combines that with a path augmenting Max flow algorithm to evaluate the load status for each state change of operation in the system. Overall system and load point reliability indices such as hourly loss of load, frequency of loss of load and expected energy unserved can be computed using this technique. On addition of DG in standby mode of operation at specific locations in the network, the reliability indices can be compared for different scenarios and strategies for placement of DG and their capacities can be determined using this methodology.
2

Maksimalaus srauto tinkle radimo algoritmų analizė / Analysis of max flow algorithms

Ūsas, Algimantas 18 January 2005 (has links)
Graphs are a pervasive data structure in computer science, and algorithms for working with them are fundamental to the field. There are hundreds of interesting computational problems defined in terms of graphs. This time we’ll touch one of them – the problem of maximum net flow. One of the main problems in the theory of graphs is the problem of maximum flow. The purpose is to calculate the biggest amount of matter, which can be relayed from the source to the flow. This is one of the easier tasks, which can be solved with the help of algorithms. Beside that the basic algorithms of maximum flow can be used for solving other problems about net flows. The results of this work are: - created the program equipment, realizing four classic methods of calculating maximum flow – Ford-Falkerson, Edmonds-Karp, Dinic and Goldberg; - with the help of this program equipment is collected the statistic of these methods efficiency. Results showed, that every method can be realized within shorter time as was proved earlier.
3

Segmentace 3D obrazových dat s využitím grafové reprezentace / Segmentation of 3D image data utilising graph representation

Demel, Jan January 2014 (has links)
This thesis deals with the application of graph theory in image segmentation. There are specifically presented method utilizing graph cuts and extensions of this method. In the first chapter thera are initially explained basics of graph theory that are essential for understanding of the presented method. It is described in the second chapter, including its extensions that use shape priors. In the third chapter there is presented solution which is used for vertebrae lesion segmentation in the CT data sets. Final function is implemented into the program but it can be used also separately. Success rate is described using sensitivity and specificity in the last chapter, there are also examples of results.
4

Towards Data-Driven I/O Load Balancing in Extreme-Scale Storage Systems

Banavathi Srinivasa, Sangeetha 15 June 2017 (has links)
Storage systems used for supercomputers and high performance computing (HPC) centers exhibit load imbalance and resource contention. This is mainly due to two factors: the bursty nature of the I/O of scientific applications; and the complex and distributed I/O path without centralized arbitration and control. For example, the extant Lustre parallel storage system, which forms the backend storage for many HPC centers, comprises numerous components, all connected in custom network topologies, and serve varying demands of large number of users and applications. Consequently, some storage servers can be more loaded than others, creating bottlenecks, and reducing overall application I/O performance. Existing solutions focus on per application load balancing, and thus are not effective due to the lack of a global view of the system. In this thesis, we adopt a data-driven quantitative approach to load balance the I/O servers at extreme scale. To this end, we design a global mapper on Lustre Metadata Server (MDS), which gathers runtime statistics collected from key storage components on the I/O path, and applies Markov chain modeling and a dynamic maximum flow algorithm to decide where data should be placed in a load-balanced fashion. Evaluation using a realistic system simulator shows that our approach yields better load balancing, which in turn can help yield higher end-to-end performance. / Master of Science / Critical jobs such as meteorological prediction are run at exa-scale supercomputing facilities like Oak Ridge Leadership Computing Facility (OLCF). It is necessary for these centers to provide an optimally running infrastructure to support these critical workloads. The amount of data that is being produced and processed is increasing rapidly necessitating the need for these High Performance Computing (HPC) centers to design systems to support the increasing volume of data. Lustre is a parallel filesystem that is deployed in HPC centers. Lustre being a hierarchical filesystem comprises of a distributed layer of Object Storage Servers (OSSs) that are responsible for I/O on the Object Storage Targets (OSTs). Lustre employs a traditional capacity-based Round-Robin approach for file placement on the OSTs. This results in the usage of only a small fraction of OSTs. Traditional Round-Robin approach also increases the load on the same set of OSSs which results in a decreased performance. Thus, it is imperative to have a better load balanced file placement algorithm that can evenly distribute the load across all OSSs and the OSTs in order to meet the future demands of data storage and processing. We approach the problem of load imbalance by splicing the whole system into two views: filesystem and applications. We first collect the current usage statistics of the filesystem by means of a distributed monitoring tool. We then predict the applications’ I/O request pattern by employing a Markov Chain Model. Finally, we make use of both these components to design a load balancing algorithm that eventually evens out the load on both the OSSs and OSTs. We evaluate our algorithm on a custom-built simulator that simulates the behavior of the actual filesystem.
5

Achievable rates for Gaussian Channels with multiple relays

Coso Sánchez, Aitor del 12 September 2008 (has links)
Los canales múltiple-entrada-múltiple-salida (MIMO) han sido ampliamente propuestos para superar los desvanecimientos aleatorios de canal en comunicaciones inalámbricas no selectivas en frecuencia. Basados en equipar tanto transmisores como receptores con múltiple antenas, sus ventajas son dobles. Por un lado, permiten al transmisor: i) concentrar la energía transmitida en una dirección-propia determinada, o ii) codificar entre antenas con el fin de superar desvanecimientos no conocidos de canal. Por otro lado, facilitan al receptor el muestreo de la señal en el dominio espacial. Esta operación, seguida por la combinación coherente de muestras, aumenta la relación señal a ruido de entrada al receptor. De esta forma, el procesado multi-antena es capaz de incrementar la capacidad (y la fiabilidad) de la transmisión en escenarios con alta dispersión.Desafortunadamente, no siempre es posible emplazar múltiples antenas en los dispositivos inalámbricos, debido a limitaciones de espacio y/o coste. Para estos casos, la manera más apropiada de explotar el procesado multi-antena es mediante retransmisión, consistente en disponer un conjunto de repetidores inalámbricos que asistan la comunicación entre un grupo de transmisores y un grupo de receptores, todos con una única antena. Con la ayuda de los repetidores, por tanto, los canales MIMO se pueden imitar de manera distribuida. Sin embargo, la capacidad exacta de las comunicaciones con repetidores (así como la manera en que este esquema funciona con respeto al MIMO equivalente) es todavía un problema no resuelto. A dicho problema dedicamos esta tesis.En particular, la presente disertación tiene como objetivo estudiar la capacidad de canales Gaussianos asistidos por múltiples repetidores paralelos. Dos repetidores se dicen paralelos si no existe conexión directa entre ellos, si bien ambos tienen conexión directa con la fuente y el destino de la comunicación. Nos centramos en el análisis de tres canales ampliamente conocidos: el canal punto-a-punto, el canal de múltiple-acceso y el canal de broadcast, y estudiamos su mejora de funcionamiento con repetidores. A lo largo de la tesis, se tomarán las siguientes hipótesis: i) operación full-duplex en los repetidores, ii) conocimiento de canal tanto en transmisión como en recepción, y iii) desvanecimiento sin memoria, e invariante en el tiempo.En primer lugar, analizamos el canal con múltiples repetidores paralelos, en el cual una única fuente se comunica con un único destino en presencia de N repetidores paralelos. Derivamos límites inferiores de la capacidad del canal por medio de las tasas de transmisión conseguibles con distintos protocolos: decodificar-y-enviar, decodificar-parcialmente-y-enviar, comprimir-y-enviar, y repetición lineal. Asimismo, con un fin comparativo, proveemos un límite superior, obtenido a través del Teorema de max-flow-min-cut. Finalmente, para el número de repetidores tendiendo a infinito, presentamos las leyes de crecimiento de todas las tasas de transmisión, así como la del límite superior.A continuación, la tesis se centra en el canal de múltiple-acceso (MAC) con múltiples repetidores paralelos. El canal consiste en múltiples usuarios comunicándose simultáneamente con un único destino en presencia de N repetidores paralelos. Derivamos una cota superior de la región de capacidad de dicho canal utilizando, de nuevo, el Teorema de max-flow-min-cut, y encontramos regiones de tasas de transmisión conseguibles mediante: decodificar-y-enviar, comprimir-y-enviar, y repetición lineal. Asimismo, se analiza el valor asintótico de dichas tasas de transmisión conseguibles, asumiendo el número de usuarios creciendo sin límite. Dicho estudio nos permite intuir el impacto de la diversidad multiusuario en redes de acceso con repetidores.Finalmente, la disertación considera el canal de broadcast (BC) con múltiples repetidores paralelos. En él, una única fuente se comunica con múltiples destinos en presencia de N repetidores paralelos. Para dicho canal, derivamos tasas de transmisión conseguibles dado: i) codificación de canal tipo dirty paper en la fuente, ii) decodificar-y-enviar, comprimir-y-enviar, y repetición lineal, respectivamente, en los repetidores. Además, para repetición lineal, demostramos que la dualidad MAC-BC se cumple. Es decir, la región de tasas de transmisión conseguibles en el BC es igual a aquélla del MAC con una limitación de potencia suma. Utilizando este resultado, se derivan algoritmos de asignación óptima de recursos basados en teoría de optimización convexa. / Multiple-input-multiple-output (MIMO) channels are extensively proposed as a means to overcome the random channel impairments of frequency-flat wireless communications. Based upon placing multiple antennas at both the transmitter and receiver sides of the communication, their virtues are twofold. On the one hand, they allow the transmitter: i) to concentrate the transmitted power onto a desired eigen-direction, or ii) tocode across antennas to overcome unknown channel fading. On the other hand, they permit the receiver to sample the signal on the space domain. This operation, followed by the coherent combination of samples, increases the signal-to-noise ratio at the input of the detector. In fine, MIMO processing is able to provide large capacity (and reliability) gains within rich-scattered scenarios.Nevertheless, equipping wireless handsets with multiple antennas is not always possible or worthwhile. Mainly, due to size and cost constraints, respectively. For these cases, the most appropriate manner to exploit multi-antenna processing is by means of relaying. This consists of a set of wireless relay nodes assisting the communication between a set of single-antenna sources and a set of single-antenna destinations. With the aid of relays, indeed, MIMO channels can be mimicked in a distributed way. However, the exact channel capacity of single-antenna communications with relays (and how this scheme performs with respect to the equivalent MIMO channel) is a long-standing open problem. To it we have devoted this thesis.In particular, the present dissertation aims at studying the capacity of Gaussian channels when assisted by multiple, parallel, relays. Two relays are said to be parallel if there is no direct link between them, while both have direct link from the source and towards the destination. We focus on three well-known channels: the point-to-point channel, the multi-access channel and the broadcast channel, and study their performance improvement with relays. All over the dissertation, the following assumptions are taken: i) full-duplex operation at the relays, ii) transmit and receive channel state information available at all network nodes, and iii) time-invariant, memory-less fading.Firstly, we analyze the multiple-parallel relay channel, where a single source communicates to a single destination in the presence of N parallel relays. The capacity of the channel is lower bounded by means of the achievable rates with different relaying protocols, i.e. decode-and-forward, partial decode-and-forward, compress-and-forward and linear relaying. Likewise, a capacity upper bound is provided for comparison, derived using the max-flow-min-cut Theorem. Finally, for number of relays growing to infinity, the scaling laws of all achievable rates are presented, as well as the one of the upper bound.Next, the dissertation focusses on the multi-access channel (MAC) with multiple-parallel relays. The channel consists of multiple users simultaneously communicating to a single destination in the presence of N parallel relay nodes. We bound the capacity region of the channel using, again, the max-flow-min-cut Theorem and find achievable rate regions by means of decode-and-forward, linear relaying and compress-and-forward. In addition, we analyze the asymptotic performance of the obtained achievable sum-rates, given the number of users growing without bound. Such a study allows us to grasp the impact of multi-user diversity on access networks with relays.Finally, the dissertation considers the broadcast channel (BC) with multiple parallel relays. This consists of a single source communicating to multiple receivers in the presence of N parallel relays. For the channel, we derive achievable rate regions considering: i) dirty paper encoding at the source, and ii) decode-and-forward, linear relaying and compress-and-forward, respectively, at the relays. Moreover, for linear relaying, we prove that MAC-BC duality holds. That is, the achievable rate region of the BC is equal to that of the MAC with a sum-power constraint. Using this result, the computation of the channel's weighted sum-rate with linear relaying is notably simplified. Likewise, convex resource allocation algorithms can be derived.
6

Objeto de aprendizagem para o ensino de algoritmos solucionadores de problemas de otimização em redes

Lourenço, Wilson Da Silva 26 February 2015 (has links)
Submitted by Nadir Basilio (nadirsb@uninove.br) on 2015-07-17T15:18:49Z No. of bitstreams: 1 Wilson da Silva Lourenco.pdf: 1321079 bytes, checksum: ea090b0df77d0c04ef1dde30e7b41558 (MD5) / Made available in DSpace on 2015-07-17T15:18:49Z (GMT). No. of bitstreams: 1 Wilson da Silva Lourenco.pdf: 1321079 bytes, checksum: ea090b0df77d0c04ef1dde30e7b41558 (MD5) Previous issue date: 2015-02-26 / The network optimization problems (NOP) are common to several areas such as engineering, transport and telecommunications, and have been objects of intense research and studies. Among the classical NOP are the problems of Shortest Path (SPP), Max Flow (MFP) and Traveling Salesman (TSP), which are usually studied in undergraduate and graduate courses such as Industrial Engineering, Computer Science, Information Systems and Logistics, with the use of resources such as chalk and blackboard that hinder the teacher's work, in the sense of showing the functioning of algorithms that solve these problems while maintaining students' motivation for learning. In this context, it is proposed in this research, a computational tool, characterized as a Learning Object (OA) and called TASNOP - Teaching Algorithms for Solving Network Optimization Problems, whose purpose is to contribute to students' understanding about concepts from NOP and, mainly, the functioning of algorithms A*, Greedy Search and Dijkstra used for resolution of SPP, Ford-Fulkerson employed in the resolution of MFP and the Nearest Neighbor to solve the TSP. It is important to highlight that the proposed OA can be accessed through web and also employed in distance learning environments (DLE). Experiments conducted in 2014 with 129 students of Computer Science, from which 51 performed an exercise using the TASNOP and 78 without this tool, confirm that students who used the TASNOP performed better in solving the proposed exercise, corroborating the idea that the OA helped to improve their understanding about the algorithms discussed in this research. In addition, the 51 students who employed the TASNOP answered a questionnaire about it use and, the answers indicated that the TASNOP shows a potential to be used as a learning support tool. / Os problemas de otimização em redes (POR) são comuns a diversas áreas como engenharia, transportes e telecomunicações, e têm sido objetos de intensas pesquisas e estudos. Entre os POR clássicos estão os problemas de Caminho Mínimo (PCM), Fluxo Máximo (PFM) e Caixeiro Viajante (PCV), os quais normalmente são estudados em cursos de graduação e pós-graduação tais como Engenharia de Produção, Ciência da Computação, Sistemas de Informação e Logística, com a utilização de recursos como giz e lousa, o que dificulta o trabalho do professor, no sentido de mostrar o funcionamento dos algoritmos que solucionam esses problemas, mantendo a motivação dos alunos para a aprendizagem. Neste contexto, propõe-se nesta pesquisa, uma ferramenta computacional, caracterizada como um Objeto de Aprendizagem (OA) denominado TASNOP - Teaching Algorithms for Solving Network Optimization Problems, cuja finalidade é contribuir para compreensão dos alunos sobre conceitos de POR e, principalmente, sobre o funcionamento dos algoritmos A*, Busca Gulosa, e Dijkstra, usados para resolução do PCM, Ford-Fulkerson empregado na resolução de PFM e o algoritmo Vizinho mais Próximo para resolução do PCV. É importante ressaltar que o OA proposto pode ser acessado via web e, inclusive, ser acoplado em ambientes de ensino a distância (EaD). Experimentos realizados no ano de 2014 envolvendo 129 alunos do curso de Ciência da Computação, dos quais 51 resolveram um exercício com o uso do TASNOP e 78 sem o seu uso, permitiram verificar que os alunos que utilizaram o TASNOP obtiveram melhor desempenho na resolução do exercício proposto, corroborando a ideia de que o OA contribuiu para melhorar suas compreensões acerca dos algoritmos abordados nesta pesquisa. Em adição, os 51 alunos que usaram o TASNOP responderam a um questionário sobre o seu uso e, com base nessas respostas, ficou evidente o potencial do TASNOP como uma ferramenta de apoio ao ensino.
7

Image Processing Methods for Myocardial Scar Analysis from 3D Late-Gadolinium Enhanced Cardiac Magnetic Resonance Images

Usta, Fatma 25 July 2018 (has links)
Myocardial scar, a non-viable tissue which occurs on the myocardium due to the insufficient blood supply to the heart muscle, is one of the leading causes of life-threatening heart disorders, including arrhythmias. Analysis of myocardial scar is important for predicting the risk of arrhythmia and locations of re-entrant circuits in patients’ hearts. For applications, such as computational modeling of cardiac electrophysiology aimed at stratifying patient risk for post-infarction arrhythmias, reconstruction of the intact geometry of scar is required. Currently, 2D multi-slice late gadolinium-enhanced magnetic resonance imaging (LGEMRI) is widely used to detect and quantify myocardial scar regions of the heart. However, due to the anisotropic spatial dimensions in 2D LGE-MR images, creating scar geometry from these images results in substantial reconstruction errors. For applications requiring reconstructing the intact geometry of scar surfaces, 3D LGE-MR images are more suited as they are isotropic in voxel dimensions and have a higher resolution. While many techniques have been reported for segmentation of scar using 2D LGEMR images, the equivalent studies for 3D LGE-MRI are limited. Most of these 2D and 3D techniques are basic intensity threshold-based methods. However, due to the lack of optimum threshold (Th) value, these intensity threshold-based methods are not robust in dealing with complex scar segmentation problems. In this study, we propose an algorithm for segmentation of myocardial scar from 3D LGE-MR images based on Markov random field based continuous max-flow (CMF) method. We utilize the segmented myocardium as the region of interest for our algorithm. We evaluated our CMF method for accuracy by comparing its results to manual delineations using 3D LGE-MR images of 34 patients. We also compared the results of the CMF technique to ones by conventional full-width-at-half-maximum (FWHM) and signal-threshold-to-reference-mean (STRM) methods. The CMF method yields a Dice similarity coefficient (DSC) of 71 +- 8.7% and an absolute volume error (|VE|) of 7.56 +- 7 cm3. Overall, the CMF method outperformed the conventional methods for almost all reported metrics in scar segmentation. We present a comparison study for scar geometries obtained from 2D vs 3D LGE-MRI. As the myocardial scar geometry greatly influences the sensitivity of risk prediction in patients, we compare and understand the differences in reconstructed geometry of scar generated using 2D versus 3D LGE-MR images beside providing a scar segmentation study. We use a retrospectively acquired dataset of 24 patients with a myocardial scar who underwent both 2D and 3D LGE-MR imaging. We use manually segmented scar volumes from 2D and 3D LGE-MRI. We then reconstruct the 2D scar segmentation boundaries to 3D surfaces using a LogOdds-based interpolation method. We use numerous metrics to quantify and analyze the scar geometry including fractal dimensions, the number-of-connected-components, and mean volume difference. The higher 3D fractal dimension results indicate that the 3D LGE-MRI produces a more complex surface geometry by better capturing the sparse nature of the scar. Finally, 3D LGE-MRI produces a larger scar surface volume (27.49 +- 20.38 cm3) than 2D-reconstructed LGE-MRI (25.07 +- 16.54 cm3).
8

Cooperative Networks with Channel Uncertainty / Réseaux coopératifs avec incertitude du canal

Behboodi, Arash 13 June 2012 (has links)
Dans cette thèse, les réseaux coopératifs sont étudiés sous cette hypothèse que la source est incertain par rapport le canal en opération. Dans le premier chapitre, des stratégies coopératives sont développées pour les canaux à relais simultanés (SRC) lesquelles se composent d'un ensemble de deux canaux à relais parmi lesquels le canal en opération est choisi. Cela est équivalent au canal de diffusion à relais (BRC). Les bornes sur la région de capacité de BRC général sont dérivées. Les résultats de capacité sont obtenus pour les cas particuliers du canal à relais simultané semi-dégradé et dégradé Gaussien. Dans le deuxième chapitre, le canal à relais composite est considéré où le canal est tiré aléatoirement d'un ensemble de la distribution conditionnelle. Le débit est fixé en dépit du canal actuel et la probabilité d'erreur (EP) asymptotique est caractérisée. Une nouvelle stratégie de codage sélectif (SCS) est introduit permettant aux relais de choisir -selon leur mesurage du canal – la meilleur schéma de codage entre Décoder-et-Transmettre (DF) et Comprimer-et-Transmettre (CF). Les théorèmes de codage de réseau bruit généralisées sont démontrés pour le cas de réseau unicast général où les relais utilisent soit DF soit CF. Dans le troisième chapitre, le spectre asymptotique de EP est introduit en tant que nouvelle mesure de performance pour réseaux composites. Il est démontré que chaque code avec le débit hors de la borne cut-set, abouti à EP égal à un et le spectre asymptotique de EP coïncide avec la probabilité d'outage pour les réseaux satisfaisant la converse forte. / In this thesis, cooperative networks are studied under the assumption that the source is uncertain about the channel in operation. In the first chapter, cooperative strategies are developed for simultaneous relay channels (SRC) which consist of a set of two single relay channels out of which the channel in operation is chosen. This is equivalent to the broadcast relay channel (BRC). Bounds on the capacity region of the general BRC with two helper relays are derived. Capacity results are obtained for specific cases of semi-degraded and degraded Gaussian simultaneous relay channels. In the second chapter, the composite relay channel is considered where the channel is randomly drawn from a set of conditional distributions according to a given distribution. The transmission rate is fixed regardless of the current channel and the asymptotic error probability (EP) is characterized. A novel selective coding strategy (SCS) is introduced which enables relays to select –based on their channel measurement– the best coding scheme between Compress-and-Forward (CF) and Decode-and-Forward (DF). Generalized Noisy Network Coding theorems are shown for the case of unicast general networks where the relays use either DF or CF scheme. In the third chapter, the asymptotic behavior of EP is studied for composite multiterminal networks. The asymptotic spectrum of EP is introduced as a novel performance measure for composite networks. It is shown that every code with rate outside cut-set bound, yields EP equal to one and for the networks satisfying strong converse condition, the asymptotic spectrum of EP coincides with the outage probability.
9

Approximating multi-commodity max-flow in practice / Approximera multi-commodity max-flow i praktiken

Emanuelsson, Kristoffer January 2016 (has links)
Garg and Könemann developed a framework for computing multi-commodity maximum flow in a graph, later called a multiplicative weight update framework. Madry used this framework and exchanged Dijkstra’s algorithm to a dynamic graph algorithm for approximating the shortest paths through the graph. With this approachhe developed the fastest algorithm to date for calculating the multi-commodity maximum flow, with a running time of Õ(mnϵ2). This project have implemented the algorithm and compared it with a slightly modified version of the former fastest algorithm by Fleischer with a time complexity of Õ(m2ϵ2). The results show that Madry’s algorithms is slower than Fleischer’s algorithm in practice for graph with less than 100 million edges. This project also computed the space needed for the dynamic algorithms used in Madry’s algorithm and can show a resulting space complexity of O(n(n+m)log2n), compared to the space complexity of Fleischer’s algorithm of O(n). For a graph with 100 million edges, 50 million Gb of space is needed to use Madry’s algorithm, which is more than our test computers had. We can therefore conclude that Madry’s algorithm is not usable in real life today, both in terms of memory usage and time consumption. / Garg and Könemann utvecklade ett framework för att beräkna multi-commodity maximum flöde i en graf sedan kallat ett multiplicative weight update framework. Madry använde detta framework och bytte ut Dijkstra’s algoritm mot en dynamisk grafalgoritm för att kunna approximera kortaste vägen i grafen. Med detta angeppssätt utvecklade han den idag snabbaste algoritmen för beräkning av multicommodity maximum flöde med en tids komplexitet på Õ(mnϵ2). Det här projektet har implementerat hans algoritm och jämfört den med den tidigare snabbaste algoritmen skapad av Fleischer med en tidskomplexitet på Õ(m2ϵ2). Resultatet visar att Madrys algoritm är långsammare än Fleischers algoritm i praktiken för grafer med färre än 100 miljoner kanter. Detta projekt beräknade också minnesåtgången för de dynamiska algoritmerna i Madrys algorithm och kan visa en resulterade minneskomplexitet på O(n(n+m)log2n), jämfört med Fleischers algoritm på O(n). För en graf med 100 miljoner kanter så behövs 50 miljoner Gb av minne för att kunna använda Madrys algoritm, vilket var mer än våra testdatorer hade. Vi kan därför konstatera att Madrys algoritm inte är användbar i praktiken idag, både när det kommer till minnesanvändning och till tidsåtgång.

Page generated in 0.0246 seconds