Event-Based Neuromorphic Systems

Foreword

The motivation for building neuromorphic systems has roots in engineering and neuroscience. On the engineering side, inspiration from how the brain solves complex problems has led to new computing algorithms; however, the goal of reverse engineering the brain is a difficult one because the brain is based on a biological technology that was evolved and not designed by human engineers. On the neuroscience side, the goal is to understand brain function, which is still at an early stage owing the extremely heterogeneous and compact nature of neural circuits. Neuromorphic systems are a bridge between these two ambitious enterprises. The lessons learned from building devices based on neural architectures are providing new engineering capabilities and new biological insights.

Building neuromorphic VLSI chips and perfecting asynchronous event-based communication between them has required a generation of talented engineering scientists. These people were inspired by Carver Mead and his 1989 landmark book on Analog VLSI and Neural Systems. I was a Wiersma Visiting Professor of Neurobiology at the California Institute of Technology in 1987 and attended ‘Carverland’ group meetings. Neuromorphic engineering was still in its infancy, but the strengths and weaknesses of the technology were already apparent. The promise of a new massively parallel, low-power, and inexpensive computing architecture was balanced by the technical challenges of working with the transistor mismatch and noise ...