Genetic Algorithms (GAs) can be easily applied to many different problems since they make few assumptions about the application domain and perform relatively well. They can also be modified with some success for handling a particular problem. The travelling salesperson problem (TSP) is a famous NP-hard problem in combinatorial optimization. As a result it has no known polynomial time solution. The aim of this dissertation will be to investigate the application of a number of GAs to the TSP. These results will be compared with those of traditional solutions to the TSP and with the results of other applications of the GA to the TSP. / Thesis (M.Sc.)-University of Natal, Pietermaritzburg, 1996.

<p>High-performance computing (HPC) and machine learning (ML) have been widely adopted by both academia and industries to address enormous data problems at extreme scales. While research has reported on the interactions of HPC and ML, achieving high performance and scalability for parallel and distributed ML algorithms is still a challenging task. This dissertation first summarizes the major challenges for applying HPC to ML applications: 1) poor performance and scalability, 2) loss of the convergence rate, 3) lower quality of the trained model, and 4) a lack of performance optimization techniques designed for specific applications. Researchers can address the four challenges in new ML applications. This dissertation shows how to solve them for two specific applications: 1) a clustering algorithm and 2) graph optimization algorithms that use reinforcement learning (RL).</p> <p>As to the clustering algorithm, we first propose an algorithm called the simulated-annealing clustering algorithm. By combining a blocked data layout and asynchronous local optimization within each thread, the simulated-annealing enhanced clustering algorithm has a convergence rate that is comparable to the K-means algorithm but with much higher performance. Experiments with synthetic and real-world datasets show that the simulated-annealing enhanced clustering algorithm is significantly faster than the MPI K-means library using up to 1024 cores. However, the optimization costs (Sum of Square Error (SSE)) of the simulated-annealing enhanced clustering algorithm became higher than the original costs. To tackle this problem, we devise a new algorithm called the full-step feel-the-way clustering algorithm. In the full-step feel-the-way algorithm, there are L local steps within each block of data points. We use the first local step’s results to compute accurate global optimization costs. Our results show that the full-step algorithm can significantly reduce the global number of iterations needed to converge while obtaining low SSE costs. However, the time spent on the local steps is greater than the benefits of the saved iterations. To improve this performance, we next optimize the local step time by incorporating a sampling-based method called reassignment-history-aware sampling. Extensive experiments with various synthetic and real world datasets (e.g., MNIST, CIFAR-10, ENRON, and PLACES-2) show that our parallel algorithms can outperform the fastest open-source MPI K-means implementation by up to 110% on 4,096 CPU cores with comparable SSE costs.</p> <p>Our evaluations of the sampling-based feel-the-way algorithm establish the effectiveness of the local optimization strategy, the blocked data layout, and the sampling methods for addressing the challenges of applying HPC to ML applications. To explore more parallel strategies and optimization techniques, we focus on a more complex application: graph optimization problems using reinforcement learning (RL). RL has proved successful for automatically learning good heuristics to solve graph optimization problems. However, the existing RL systems either do not support graph RL environments or do not support multiple or many GPUs in a distributed setting. This has compromised RL’s ability to solve large scale graph optimization problems due to the lack of parallelization and high scalability. To address the challenges of parallelization and scalability, we develop OpenGraphGym-MG, a high performance distributed-GPU RL framework for solving graph optimization problems. OpenGraphGym-MG focuses on a class of computationally demanding RL problems in which both the RL environment and the policy model are highly computation intensive. In this work, we distribute large-scale graphs across distributed GPUs and use spatial parallelism and data parallelism to achieve scalable performance. We compare and analyze the performance of spatial and data parallelism and highlight their differences. To support graph neural network (GNN) layers that take data samples partitioned across distributed GPUs as input, we design new parallel mathematical kernels to perform operations on distributed 3D sparse and 3D dense tensors. To handle costly RL environments, we design new parallel graph environments to scale up all RL-environment-related operations. By combining the scalable GNN layers with the scalable RL environment, we are able to develop high performance OpenGraphGym-MG training and inference algorithms in parallel.</p> <p>To summarize, after proposing the major challenges for applying HPC to ML applications, this thesis explores several parallel strategies and performance optimization techniques using two ML applications. Specifically, we propose a local optimization strategy, a blocked data layout, and sampling methods for accelerating the clustering algorithm, and we create a spatial parallelism strategy, a parallel graph environment, agent, and policy model, and an optimized replay buffer, and multi-node selection strategy for solving large optimization problems over graphs. Our evaluations prove the effectiveness of these strategies and demonstrate that our accelerations can significantly outperform the state-of-the-art ML libraries and frameworks without loss of quality in trained models.</p>

Les distributeurs sont des intermédiaires économiques car ils offrent sur le marché des produits ou services conçus, fabriqués ou exécutés par d’autres (les fournisseurs). Cette recherche démontre que le degré de protection accordé par le droit positif aux différents distributeurs n’est pas corrélé à leur degré de dépendance aux fournisseurs. Les qualifications de la dépendance sont multiples, incertaines, incohérentes et parfois artificiellement restrictives ou extensives, en droit du travail comme en droit économique. Des distributeurs connaissant un même degré de dépendance à leur fournisseur peuvent être traités de façon différente sans justification au regard des fondements des règles considérées. Il existe ainsi des inégalités de traitement. Il est donc proposé un nouveau droit de la dépendance des distributeurs, plus cohérent et égalitaire. Celui-ci se traduirait par un statut légal des distributeurs remplaçant les multiples statuts spéciaux existants, traçant une frontière plus nette avec le droit du travail et conciliant les intérêts des distributeurs avec la liberté des fournisseurs d’organiser la distribution de leurs produits et services. Dépassant la notion trop restrictive de contrat et les conflits de qualification, ce statut viserait la relation de distribution et prévoirait un régime appréhendant la complexité et l’évolutivité de la dépendance des distributeurs grâce à l’information, à une garantie de revenus et différentes indemnités de fin de relation. / Distributors are economic intermediaries because they offer on the market goods and services produced or served by others (the suppliers). This research demonstrates that the degree of protection offered by the Law to the various distributors isn’t correlated with their degree of dependence towards suppliers. The qualifications of dependence are numerous, incoherent and sometimes artificially restrictive or extensive, in labour law as in business law. Distributors experiencing a same degree of dependence towards their supplier can be treated differently without any justification regarding the grounds of the rules in question. Hence, appear inequalities towards the Law. Therefore, new legal rules are suggested. They would take the form of a statute ruling distributors and replacing the numerous statutes in force at the present time. It would draw a clearer line between labour law and business law and conciliate the distributors’ interests with the freedom of suppliers to organize the distribution of their goods and services. Going other the too restrictive notion of contact and the conflicts between qualifications, this statute would rule the relation of distribution and contain rules comprehending the complex and changing nature of dependence. It would ensure sufficient information of distributors and offer them an income guarantee and various compensations when the relation is terminated.