Oxide-Electronics-Based Artificial Neurons and Synapses

Neuromorphic circuits, that is, electronic circuits that are inspired by the architecture of the brain, are a promising way to overcome the two significant problems of artificial intelligence systems, power consumption and flexibility. In general, neuromorphic systems are based on today’s silicon technology. Our goal is the implementation of neuromorphic circuits using emerging devices. The reason is that some emerging devices have functionalities that are particularly convenient to get further advantages. In this talk, we will present as examples our synapses and neurons based on oxide electronic components. Our synapses have a significant advantage concerning power; we will explain how they work, and how we intend to use them in a system capable of learning. When it comes to our neurons, they present a substantial advantage in size when used in low-frequency systems; in this case, we will discuss the need for arbitrarily slow neuromorphic systems.