Implementing compact, low-power artificial neural processing systems with real-time\non-line learning abilities is still an open challenge. In this paper we present a\nfull-custom mixed-signal VLSI device with neuromorphic learning circuits that emulate the\nbiophysics of real spiking neurons and dynamic synapses for exploring the properties of\ncomputational neuroscience models and for building brain-inspired computing systems.\nThe proposed architecture allows the on-chip configuration of a wide range of network\nconnectivities, including recurrent and deep networks, with short-term and long-term\nplasticity. The device comprises 128 K analog synapse and 256 neuron circuits with\nbiologically plausible dynamics and bi-stable spike-based plasticity mechanisms that\nendow it with on-line learning abilities. In addition to the analog circuits, the device\ncomprises also asynchronous digital logic circuits for setting different synapse and neuron\nproperties as well as different network configurations. This prototype device, fabricated\nusing a 180nm 1P6M CMOS process, occupies an area of 51.4mm2, and consumes\napproximately 4mW for typical experiments, for example involving attractor networks.\nHere we describe the details of the overall architecture and of the individual circuits and\npresent experimental results that showcase its potential. By supporting a wide range\nof cortical-like computational modules comprising plasticity mechanisms, this device will\nenable the realization of intelligent autonomous systems with on-line learning capabilities
Loading....