Evolving Finite State Machine (e-FSM)

Design of evolving methodology: evolving Finite State Machine (e-FSM) for constructing a stochastic model from scratch without human intervensions.

• Instead of designing a model by hand, evolving Finite State Machine constructs a stochastic model from scratch by determining states and identifying state-transitions.
• State Determination: the observations are clustered by the online clustering method (e.g., evolving Takagi-Sugeno) and each cluster-center is considered as a state.
• State-transition Identification: Observed consecutive variation of probability distributions at t-1 and t are used to identify state-transitions in an online way.
• Project is sponsored by Ford Motor Company.
• Publication(s): https://arxiv.org/abs/1908.10823