Realization of neural coding by stochastic switching of magnetic tunnel junction

In this paper, we present the realization of neural rate coding with probabilistic spike trains by using two back-to-back magnetic tunnel junctions (MTJs) for the bio-inspired computing applications. By exploiting the intrinsic stochastic switching of the MTJ device between two different resistance states, an analog stimulus can be converted into a probabilistic spike train and its spike rate can be modulated by the switching probability, which depends on the magnitude of the stimulus. To implement such conversion, we propose a hybrid CMOS/MTJ circuit. By using a physics-based MTJ compact model and a commercial CMOS 40nm design kit, its functionality has been validated. Additionally, we also investigate the relationship between switching probability and stimulus by Monte Carlo simulations.