Challenges of memristor based neuromorphic computing system

摘要

The high efficiency of human brain illmninates thedevelopment of neuromorphic computing system（NCS）. In truman brain, membrane potential thatgoes beyond certain threshold voltage triggers thepropagation of spike signals that carries the infor-mation to proceeding neurons. This process is usu-ally abstracted as the integration and fire model,inspiring circuit design and architecture solutionsfor neuromorphic computing. Different from con-ventional von Neumann architecture, neuromor-phic systems tend to closely integrate the comput-ing units and memories, efficiently reducing thedistance between computing units and memoryand consequently breaking the so-called ＂memorywall＂ .