Unsupervised Learning of Sequential Patterns

In order to allow agents for acting autonomously and making their decisions on a solid basis an interpretation of the current scene has to be done. Scene interpretation can be done by checking if certain patterns match to the current belief of the world. In some cases it is impossible to acquire all situations an agent might have to deal with later at run-time in advance. Here, an automated acquisition of patterns would help an agent to adapt to the environment. Agents in dynamic environments have to deal with world representations that change over time. Additionally, predicates between arbitrary objects can exist in their belief of the world. In this work we present a learning approach that learns temporal patterns from a sequence of predicates.