Parallel Monte Carlo tree search in general video game playing
| Nível educacional (dc.audience.educationLevel) | Graduação | pt_BR |
| Orientador(a) (dc.contributor.advisor) | Rigo, Sandro José | |
| Lattes Orientador(a) (dc.contributor.advisorLattes) | http://lattes.cnpq.br/3914159735707328 | pt_BR |
| Autor(a) (dc.contributor.author) | Centeleghe, Luis Gustavo Simioni | |
| Curso (dc.curso) | Ciência da Computação | pt_BR |
| Data de Disponibilização (dc.date.accessioned) | 2022-04-12T19:00:04Z | |
| dc.date.available (dc.date.available) | 2022-04-12T19:00:04Z | |
| Data da defesa / Data do evento (dc.date.issued) | 2019-12-13 | |
| Resumo (dc.description.resumo) | Monte Carlo Tree Search (MCTS) parallelization is one of the many possible enhancements for MCTS algorithms. Since MCTS parallelization methods were first proposed in 2008 by Cazenave and Jouandeau (2008), many researchers have been evaluating them using a variety of testing methodologies and games. However, no work has been done on evaluating these methods in the rather new area of General Video Game Playing (GVGP), an area that challenges the creation of agents that are able to play any video game even without prior knowledge about the video game they are going to play. To address this gap, this paper proposes the implementation and evaluation of the three main MCTS parallelization methods (Leaf, Root, and Tree Parallelization) as agents of the General Video Game AI framework, a popular framework for GVGP agents evaluation. It is important to notice that this paper is not focused on comparing the parallel MCTS agents to other existing GVGP agents, but rather on exploring how the MCTS parallelization methods compare between themselves. This paper also presents a testing methodology for evaluating these agents, which is based on a set of three experiments focused on different aspects of the parallel MCTS algorithms. These experiments were executed using 32 hyper-threads of a computer equipped with two Intel Xeon E5-2620v4 processors. In these experiments, the overall best results were achieved by the root parallelization method using the sum merging technique and the UCT’s sigma value of √ 2. However, it is also discussed in the paper some scenarios where other configurations performed better | en |
| URI (dc.identifier.uri) | http://www.repositorio.jesuita.org.br/handle/UNISINOS/11041 | |
| Nome da instituição (dc.publisher) | Universidade do Vale do Rio dos Sinos | pt_BR |
| Assunto (dc.subject) | Monte Carlo Tree Search | en |
| Assunto (dc.subject) | Parallel Monte Carlo Tree Search | en |
| Assunto (dc.subject) | General Video Game Playing | en |
| Assunto (dc.subject) | General Video Game AI | en |
| Título (dc.title) | Parallel Monte Carlo tree search in general video game playing | en |
| Tipo de arquivo (dc.type) | TCC | pt_BR |
Arquivos
Pacote original
1 - 1 de 1
Carregando...
- Nome:
- Luis Gustavo Simioni Centeleghe.pdf
- Tamanho:
- 581.76 KB
- Formato:
- Adobe Portable Document Format
- Descrição:
- Parallel_monte
Licença do pacote
1 - 1 de 1
Carregando...
- Nome:
- license.txt
- Tamanho:
- 2.12 KB
- Formato:
- Item-specific license agreed upon to submission
- Descrição: