Stochastic sampling via synaptic delay in spiking RBMs using integrated resistive and threshold switching devices

Key Points

• The study aims to explore the role of synaptic delay in facilitating stochastic sampling for enhanced learning in spiking restricted Boltzmann machines.
• Implemented a tunable RRAM–TS device for inducing synaptic delay.
• Focused on sampling-based learning strategies within spiking restricted Boltzmann machines.
• Analyzed the impact of temporal stochasticity on learning efficiency.
• The integration of synaptic delay significantly improved sampling efficiency in spiking RBMs.
• Demonstrated enhanced learning performance due to increased temporal stochasticity from the device.
• Indicated a notable reduction in training time required for effective learning.