An Experiment of Attention-Based Guided Reinforcement Learning in the Game of Space Invaders = An Experiment of Attention-Based Guided Reinforcement Learning in the Game of Space Invaders


Aziz, Arsyi Syarief (2023) An Experiment of Attention-Based Guided Reinforcement Learning in the Game of Space Invaders = An Experiment of Attention-Based Guided Reinforcement Learning in the Game of Space Invaders. Skripsi thesis, Universitas Hasanuddin.

[thumbnail of Cover]
Preview
Image (Cover)
H071191003_skripsi_14-07-2023 caver1.jpg

Download (265kB) | Preview
[thumbnail of Bab 1-2] Text (Bab 1-2)
H071191003_skripsi_14-07-2023 bab 1-2.pdf

Download (6MB)
[thumbnail of Dapus] Text (Dapus)
H071191003_skripsi_14-07-2023 dp.pdf

Download (3MB)
[thumbnail of Full Text] Text (Full Text)
H071191003_skripsi_14-07-2023.pdf
Restricted to Repository staff only until 6 October 2025.

Download (11MB)

Abstract (Abstrak)

This research aims to imitate three competencies of human-level general intelligence: perception, learning, and attention. The objective of the study was to investigate the effects of visual attention mechanisms on guided-reinforcement learning models. The methodology for the study involved performing two experiments in the Space Invaders game. The first experiment implemented two guided reinforcement learning algorithms and assessed them based on their acquired scores to obtain the most suitable model for the environment. The second experiment then integrated the obtained best model with three types of visual-attention modules and evaluated the modules based on their acquired scores and a visual analysis of their saliency maps. The results demonstrated that the visual attention mechanisms improved the guided-reinforcement learning models’ ability to focus on objects in the environment. However, further research is needed to establish performance improvements over non-attention models.

Keywords : Reinforcement Learning, Imitation Learning, Visual Attention, Generative Adversarial Network, Deep Learning

Item Type: Thesis (Skripsi)
Uncontrolled Keywords: Reinforcement Learning, Imitation Learning, Visual Attention, Generative Adversarial Network, Deep Learning
Subjects: Q Science > QA Mathematics
Divisions (Program Studi): Fakultas Matematika dan Ilmu Peng. Alam > Matematika
Depositing User: S.Sos Rasman -
Date Deposited: 25 Jan 2024 02:00
Last Modified: 25 Jan 2024 02:00
URI: http://repository.unhas.ac.id:443/id/eprint/29440

Actions (login required)

View Item
View Item