Biologically inspired learning methods (Back)


Development of mathematical and computational models for interpretation of findings on leaning in animals and human beings is the main focus of this research. The central benefit we expect from this research is devising some biologically inspired computational methods for gradual learning by means of learning to solve simple problems in non-sophisticated sensory system using RL, then abstracting the learned knowledge and transferring it into multimodal sensory space. In this way, we are interested in exploiting ideas enthused by mirror neurons and multi-store model of memory in conjunction with Bayesian Programming and reinforcement learning to develop relational concepts as a mean for knowledge abstraction and imitative learning. Learned concepts might be sensible at one shot of sensing or be temporally extended in the sensory space.


< Back