Fig. 1: The opportunities of neuromorphic computing and possible applications are illustrated.
From: Leveraging stochasticity in memristive synapses for efficient and reliable neuromorphic systems
a Shows the dynamic adaptation between conventional and stochastic synapses with proper data distribution. It also shows the possibility of utilizing our proposed synapses with a regular SNN and reservoir computing with optimized operation costs. Finally, with stochastic synapses, the energy of our proposed architecture gets optimized at run-time. b Performance of various applications is observed with regular and stochastic synapses. Static and temporal datasets are considered for the performance evaluation2,29,31,32. In addition, control applications are considered to observe the perforation at different setups2. Reservoir computing is also checked with stochastic synapses, which can be a design option with optimized power10.
