Invited - Spiking neuron circuits in ULSIC vs TFT technologies

Conference Dates

May 15-18, 2023


Recent advances on computing systems have enabled increasing success of algorithms using artificial intelligence. Researchers are now exploring new computational paradigms and materials to enable computing at the level of the device, allowing increased privacy and also reduction in energy. One of the most promising techniques is to realize circuits that imitate how neurons in biological brains function. Spike-based neural networks have been shown to hold more computational power than other neuromorphic architectures and their integration into mainstream computing is projected to herald a new age of computational power. Integrating neuron circuits with the functionality of materials used in flexible electronics is likely to open up a large field of applications, most notably for sensors for continuous health monitoring. In traditional MOFET technologies, spiking neuron circuits are typically operated in the deep subthreshold in order to take advantage of the exponential dependence of Vg to achieve the spiking action and also to optimize energy consumption. Nevertheless, this gives rise to some challenging problems when implemented in flexible technologies where the desire for using low cost and low temperature processes leads to lower mobility and much greater variability in device processing.

Please click Download on the upper right corner to see the full abstract.

35.pdf (95 kB)

This document is currently not available here.