0.25
0.5
0.75
1.25
1.5
1.75
2
Kloniranje obnašanja
Published on Mar 29, 20122867 Views
Na projektu EUSAS smo ob predzadnjem mejniku predstavili algoritem in rezultate vedenjskega kloniranja. Algoritem MASDA (avtor Andraž Bežek) se je znal naučiti strateških vzorcev igranja 11 na 11 igra
Related categories
Chapter list
Behaviour analysis and cloning - part 100:00
Presentation outline00:24
Cognitive MASDA - Features01:33
Cognitive MASDA - Algorithm structure02:59
Cognitive MASDA - Action graph construction04:29
Cognitive MASDA / Vignette1.1 - Action graph construction05:31
Taxonomies - action06:40
Taxonomies - agent roles07:30
Taxonomies - cognition07:38
Multi-Agent Strategy Discovering Algorithm - Abstract action graph (1)07:48
Cognitive MASDA - Abstract action graph09:09
Multi-Agent Strategy Discovering Algorithm - Abstract action graph (2)09:42
Language atributes10:11
Multi-Agent Strategy Discovering Algorithm - Machine learning11:02
Multi-Agent Strategy Discovering Algorithm - Civilian pattern (1)11:54
Multi-Agent Strategy Discovering Algorithm - Civilian pattern (2)12:50
Multi-Agent Strategy Discovering Algorithm - Soldier pattern (1)13:20
Multi-Agent Strategy Discovering Algorithm - Soldier pattern (2)14:08
Multi-Agent Strategy Discovering Algorithm - Soldier pattern (3)14:13
Pattern editor14:28
XML output of patterns14:57
Pattern discovery replay / vignette115:18
Conclusion and discussion15:46
Behaviour analysis and cloning - part 216:11
Presentation outline (1)17:53
Presentation outline (2)18:23
Application of CMASDA on Vignette1.120:14
Rule clustering20:22
DTW Algorithm - For measuring similarity between original and cloned behaviour22:12
DTW for comparing 2 games22:52
Influence of no. of clusters23:54
Behaviour cloning (1)25:19
Behaviour cloning (2)26:22
Behaviour cloning (3)27:45
Behaviour cloning - Validation28:04
Behaviour cloning - Simulation, human28:26
Computing similarity by DTW (1)29:44
Computing similarity by DTW (2)32:46
Visual comparison of actual games32:56
Conclusions40:00