Homeostatic Plasticity in a Leaky Integrate and Fire Neuron Using Tunable Leak

In this paper, in an effort to implement an unsupervised learning algorithm for silicon neurons, we present a mixed-signal Leaky Integrate-And-Fire (LIF) neuron with two different integrated homeostasis circuits using programmable leak. The homeostasis mechanism is realized by controlling the charge accumulation rate on the neuron integrator by varying the leakage rate using external signals. The proposed homeostasis circuits have been simulated using a 65nm CMOS process and their performances have been compared with existing homeostasis implementations. Results show that our designs achieve 12.8%-18.1 % power improvements and 25.1 %-48.2% area improvements over similar prior implementation. Also, power consumption can be reduced in the circuits by adjusting the leakage through bias currents.