1 |
A virtual machine framework for domain-specific languagesFick, David 19 October 2007 (has links)
Experts in a field regularly apply a defined set of rules or procedures to carry out a problem-solving task or analysis on a given problem. Often the problem can be represented as a computer model, be it mathematical, chemical, or physics based, and so on. It would certainly be advantageous for a domain expert who is not proficient in software development to express solutions to problems in a domain-specific notation that can be executed as a program. Many new ideas aim to make software development easier and shift the development role closer to the end-user. One such means of development is the use of a small, intuitive programming language called a Domain-Specific Language (DSL.) This dissertation examines a generic approach to constructing a Virtual Machine (VM) to provide the runtime semantics for a particular DSL. It proposes a generic, object-oriented framework, called a VM Framework, in which to build a VM by subtyping abstract instruction and environment classes that are part of the VM Framework. The subtyped classes constitute an environment and an interface called an instruction set architecture and the instructions can access and operate on the environment in a deterministic way to provide the runtime semantics of a DSL program. Both instruction classes and environment classes encapsulate functionality of an existing domain, represented programmatically as a namespace construct. The namespace is home to related classes that provide the various concepts inherent of a domain. These are concepts understood by a domain expert and in this dissertation it is shown how they are exposed as DSL constructs. With the use of compiler writing tools, a compiler can be created for a DSL that generates an appropriate instruction sequence that can be executed by the VM. The grammar of the DSL is shown to feature constructs that allow a domain expert to express concepts of the underlying domain in an intuitive manner. The dissertation details how a VM is configured for a specific set of instructions and an environment. Instruction sets and environments can be extended creating VMs with additional semantics for DSLs that are similar, or contain subsets of semantics of other DSLs. The languages are intended to be intuitive and it is shown using examples how a specific DSL program is mapped to an instruction sequence with the instruction set architecture and environment in mind. Comparative performance in relation to other DSL implementations, including a hard-coded approach of a VM and an interpreted approach are also provided. The VM Framework is proven to be most effective in rapidly prototyping a DSL for a particular problem domain. The dissertation also provides examples of DSLs such as a real-valued expression language and a scene description language that uses a ray-tracer for rendering geometric objects onto a canvas. It is shown how the scene description language is an extension to the real-valued expression language in terms of their underlying VMs. All DSL grammars are provided. / Dissertation (MSc (Computer Science))--University of Pretoria, 2007. / Computer Science / MSc / unrestricted
|
2 |
Machine Learning - Managerial Perspective : A Study to define concepts and highlight challenges in a product-based IT OrganizationBangabash, Subhasish, Panda, Srimanta January 2019 (has links)
The purpose of this research is to understand the main managerial challenges that arise in the context of Machine Learning. This research aims to explore the core concepts of Machine Learning and provide the same conceptual foundation to managers to overcome possible obstacles while implementing Machine Learning. Therefore, the main research question is: What are the phases and the main challenges while managing Machine Learning project in a product based IT organization? The focus is on the main concepts of Machine Learning and identifying challenges during each phase through literature review and qualitative data collected from interviews conducted with professionals. The research aims to position itself in the field of research which looks for inputs from consultants and management professionals either associated with Machine Learning or they are planning to start such initiatives. In this research paper we introduce ACDDT (Agile-Customer-Data-Domain-Technology) model framework for managers. This framework is centered on the main challenges in Machine Learning project phases while dealing with customer, data, domain and technology. In addition, the frame work also provides key inputs to managers for managing those challenges and possibly overcome them.
|
3 |
Uma abordagem para construção de sistemas fuzzy baseados em regras integrando conhecimento de especialistas e extraído de dadosLima, Helano Póvoas de 17 September 2015 (has links)
Submitted by Daniele Amaral (daniee_ni@hotmail.com) on 2016-09-15T12:10:51Z
No. of bitstreams: 1
DissHPL.pdf: 5127660 bytes, checksum: 4ffaa3ce20b9eb7adef78d152d5c17d2 (MD5) / Approved for entry into archive by Marina Freitas (marinapf@ufscar.br) on 2016-09-16T19:47:09Z (GMT) No. of bitstreams: 1
DissHPL.pdf: 5127660 bytes, checksum: 4ffaa3ce20b9eb7adef78d152d5c17d2 (MD5) / Approved for entry into archive by Marina Freitas (marinapf@ufscar.br) on 2016-09-16T19:47:15Z (GMT) No. of bitstreams: 1
DissHPL.pdf: 5127660 bytes, checksum: 4ffaa3ce20b9eb7adef78d152d5c17d2 (MD5) / Made available in DSpace on 2016-09-16T19:47:20Z (GMT). No. of bitstreams: 1
DissHPL.pdf: 5127660 bytes, checksum: 4ffaa3ce20b9eb7adef78d152d5c17d2 (MD5)
Previous issue date: 2015-09-17 / Empresa Brasileira de Pesquisa Agropecuária (EMBRAPA) / Historically, since Mamdani proposed his model of fuzzy rule-based system, a lot has changed in the construction process of this type of models. For a long time, the research efforts were directed towards the automatic construction of accurate models starting from data, making fuzzy systems almost mere function approximators. Realizing that this approach escaped from the original concept of fuzzy theory, more recently, researchers attention focused on the automatic construction of more interpretable models. However, such models, although interpretable, might not make sense to the expert. This work proposes an interactive methodology for constructing fuzzy rule-based systems, which aims to integrate the knowledge extracted from experts and induced from data, hoping to contribute to the solution of the mentioned problem. The approach consists of six steps. Feature selection, fuzzy partitions definition, expert rule base definition, genetic learning of rule base, rule bases conciliation and genetic optimization of fuzzy partitions. The optimization and learning steps used multiobjective genetic algorithms with custom operators for each task. A software tool was implemented to support the application of the approach, offering graphical and command line interfaces and a software library. The efficiency of the approach was evaluated by a case study where a fuzzy rule-based system was constructed in order to offer support to the evaluation of reproductive fitness of Nelore bulls. The result was compared to fully manual and fully automatic construction methodologies, the accuracy was also compared to classical algorithms for classification. / Historicamente, desde que Mamdani propôs seu modelo de sistema fuzzy baseado em regras, muita coisa mudou no processo de construção deste tipo de modelo. Durante muito tempo, os esforços de pesquisa foram direcionados à construção automática de sistemas precisos partindo de dados, tornando os sistemas fuzzy quase que meros aproximadores de função. Percebendo que esta abordagem fugia do conceito original da teoria fuzzy, mais recentemente, as atenções dos pesquisadores foram voltadas para a construção automática de modelos mais interpretáveis. Entretanto, tais modelos, embora interpretáveis, podem ainda não fazer sentido para o especialista. Este trabalho propõe uma abordagem interativa
para construção de sistemas fuzzy baseados em regras, que visa ser capaz de integrar o conhecimento extraído de especialistas e induzido de dados, esperando contribuir para a solução do problema mencionado. A abordagem é composta por seis etapas. Seleção de atributos, definição das partições fuzzy das variáveis, definição da base de regras do especialista,
aprendizado genético da base de regras, conciliação da base de regras e otimização genética da base de dados. As etapas de aprendizado e otimização utilizaram algoritmos genéticos multiobjetivo com operadores customizados para cada tarefa. Uma ferramenta de software foi implementada para subsidiar a aplicação da abordagem, oferecendo interfaces gráfica e de linha de comando, bem como uma biblioteca de software. A eficiência da abordagem foi avaliada por meio de um estudo de caso, onde um sistema fuzzy baseado em regras foi construído visando oferecer suporte à avaliação da aptidão reprodutiva de touros Nelore. O resultado foi comparado às metodologias de construção inteiramente manual e inteiramente automática, bem como a acurácia foi comparada a de algoritmos clássicos para classificação.
|
Page generated in 0.0319 seconds