Valk, Max (2020) Comparing the performance of Argumentation-Based Learning with tabular and approximation-based QLearning: A quantitative study. Bachelor's Thesis, Artificial Intelligence.
|
Text
AI_BA_2020_MAXVALK.pdf Download (469kB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
Toestemming.pdf Restricted to Registered users only Download (105kB) |
Abstract
Argumentation-Based Learning (ABL) is a newly developed algorithm for on-line incremental learning which has been shown to have outperformed other algorithms in both learning speed and precision. To expand upon the comparisons made to other algorithms in the original paper, this study focused on tabular and Deep Q-Learning. A genetic algorithm was used to explore the parameters for Deep Q-learning. The epsilon-greedy and Boltzman exploration policies for the Deep Q-Learning algorithm were considered. It was found that ABL outperforms both tabular and Deep Q-Learning, both in terms of final precision and learning speed. However, it should be noted that the search for parameters for the Deep Q-Learning neural network and its exploration policy were by no means exhaustive, and further investigations are required for a more definitive conclusion.
| Item Type: | Thesis (Bachelor's Thesis) |
|---|---|
| Supervisor name: | Ayoobi, H. |
| Degree programme: | Artificial Intelligence |
| Thesis type: | Bachelor's Thesis |
| Language: | English |
| Date Deposited: | 12 Aug 2020 06:22 |
| Last Modified: | 12 Aug 2020 06:22 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/23068 |
Actions (login required)
 |
View Item |