Constructing Temporally Extended Actions through Incremental Community Detection

Table 1

Sampled state transition graph changes during learning.

Scenario	Pac-Directional	Pac-Random
Nodes
Initial amount	561	967
Final amount	839.5 (25.4)	1105.4 (11.1)
Increased amount	278.5 (25.4)	138.4 (11.1)
Edges
Initial amount	1504	3406
Final amount	2368.1 (60.8)	4331.6 (16.5)
Increased amount	864.1 (60.8)	925.6 (16.5)