Application of Monte Carlo Algorithms with Neural Network-Based Intermediate Area to the Thousand Card Game
Fragment książki (Rozdział monografii pokonferencyjnej)
MNiSW
20
Poziom I
| Status: | |
| Autorzy: | Gałka Łukasz, Karczmarek Paweł, Czerwiński Dariusz |
| Dyscypliny: | |
| Aby zobaczyć szczegóły należy się zalogować. | |
| Wersja dokumentu: | Drukowana | Elektroniczna |
| Język: | angielski |
| Strony: | 68 - 77 |
| Scopus® Cytowania: | 0 |
| Bazy: | Scopus |
| Efekt badań statutowych | NIE |
| Finansowanie: | Finansowanie z funduszu dyscypliny Informatyka techniczna i telekomunikacja: FD-20/IT-3/999 (80% kwoty), FD-20/IT-3/004 (10% kwoty), FD-20/IT-3/008 (10% kwoty). |
| Materiał konferencyjny: | NIE |
| Publikacja OA: | NIE |
| Abstrakty: | angielski |
| Modern techniques of artificial intelligence used in computer games allow to obtain very realistic and effective artificial agents controlling the game on the part of computer players. The main goal of this study is to create modern artificial intelligence algorithms based on neural networks and algorithms based on the Monte Carlo method to control players in the popular card game called Thousand. We propose two approaches based on neural networks trained on the basis of the Monte Carlo algorithm and the recursive Monte Carlo algorithm. Statistical approaches are characterized by high response times. Hence, an attempt is made to implement solutions that maintain high efficiency with a shorter operating time and, consequently, reduced requirements for computing complexity. The research showed no significant differences in the Monte Carlo approaches and the corresponding neural network methods. In the case of recursive method invocation, the effectiveness increased compared to the base methods. |
