Institutional Repository of Vilnius University
Stalo žaidimo dirbtinio intelekto kūrimas naudojant mašininį mokymąsi /
Abstract
dc:descriptionThis paper focuses on the development and optimization of a Convolutional Neural Network (CNN) for international draughts, incorporating the minimax algorithm with alpha-beta pruning. The research primarily investigates identifying the optimal architecture for the CNN by fine-tuning its hyperparameters. This study also compares the skill levels of four models trained with different types of data against an artificial intelligence that has a basic heuristic for an evaluation function. The results indicate that the supposed improvements made to the models had a negative impact on their playing performance. Overall, the Convolutional Neural Network approach lost to the simple heuristic function by a small margin. In terms of speed, the heuristic evaluation function was much better than the CNN. Based on the findings of the study, it is recommended to explore enhancements to the CNN architecture by incorporating elements such as Dropout and MaxPooling layers. Dropout can mitigate overfitting issues by randomly deactivating certain neurons during training, promoting a more robust model. MaxPooling, on the other hand, helps reduce computational complexity via downsampling.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Kraujalis, Liudas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD192063240&prefLang=en_US