Fig. 10: Computing with stochastic network dynamics.

a MTJ-based probabilistic synapses are used for hardware encoding synaptic plasticity, which is tuned according to the spike timing-dependent plasticity rule. b Demonstration of clustering images using unsupervised learning in simulation. c MTJ-based spintronic memristor system for binary stochastic neuron or probabilistic bit (P-bit) in the absence of significant excitation. d Random telegraph output signals under different input voltages, where more “0 V” and “5 V” are observed for lower and higher input voltages, respectively. e Network of P-bits is configured according to the nature of a problem to solve the problem. f Network of six P-bits is used to solve a simple integer factorization problem, 161 = 23 × 7. Parts (a, b) reprinted with permission from ref. ¹³¹, Copyright © 2015, IEEE. Parts (d, f) reprinted with permission from ref. ²⁶, Springer Nature Limited.