Skip to main content

InfoTech e-conference

32nd International Conference on Information Technologies

Graph Generation Approach for Agent Behaviour

Vanya Markova
Ventseslav Shopov
Institute of Robotics ‚Äď BAS, Plovdiv

Abstract: This paper deals with the problem for automatic graph generation and clustering for construction of curriculum learning. The poposed approach is based on reinforcement learning framework. In addition we use transfer of full transferred reward (Reward FT) knowledge to achieve better reward. Inseparable part of this methods is a graph generation and graph clustering.

Key words:
reinforcement learning
graph clustering
graph generation
knowledge transfer