Abstrait

Forwards-backwards algorithm based on parallel factors

Haiyang Chen, Hongying Nie, Ruirui Mao

The study about highly efficient inference algorithms on dynamic Bayesian networks has become one of the focuses in the area of artificial intelligence. In order to improve the inference efficient of the improved forwards-backwards algorithm, we proposed the forwards-backwards algorithm based on parallel factors. After explaining the basic idea of the parallel factors, we defined the parallel factors, then introduced them to the forwards pass and backwards pass to realize the computation step sharing and reduce the amount of computation further. It is proved by the simulation experiments that the forwardsbackwards algorithm based on parallel factors is correct and efficient.