This paper introduces the hardware and the ASIC implementations of the four most popular biologically inspired neuron models. The models are quartic, Izhikevich, Hindmarsh Rose and Fitzhugh-Nagumo. Moreover, some approximate computing techniques are applied on these models to reduce the area and power consumption. In addition, ASIC implementations of these models and their approximate versions are carried out. Also, spiking behavior error between these models and the Hodgkin Huxley model, the reference accurate model, is presented. Finally, a fair comparative analysis is discussed to help the Spiking Neural Networks designers to select the best neuron model hardware implementation from the power, area and accuracy perspectives.
|Original language||English (US)|
|Title of host publication||2018 IEEE 61st International Midwest Symposium on Circuits and Systems (MWSCAS)|
|Publisher||Institute of Electrical and Electronics Engineers (IEEE)|
|Number of pages||4|
|State||Published - Feb 28 2019|
Bibliographical noteKAUST Repository Item: Exported on 2020-10-01
Acknowledgements: This research was partially funded by ONE Lab at Cairo University, Zewail City of Science and Technology, and KAUST.