Demonstration of Stochastic Resonance, Population Coding, and Population Voting Using Artificial MoS₂ Based Synapses

Area of Science:

• Neuroscience
• Materials Science
• Electrical Engineering

Background:

• Animals require fast, low-energy signal detection for survival in challenging environments.
• Neural systems utilize stochastic resonance (SR), population coding (PC), and population voting (PV) for efficient signal detection.
• Existing artificial systems often struggle with low-energy, high-sensitivity signal detection.

Purpose of the Study:

• To experimentally demonstrate a bio-inspired artificial neural system for efficient weak signal detection.
• To leverage stochastic resonance and population voting in artificial neurons for enhanced sensitivity and reduced latency.
• To achieve frugal energy expenditure for signal detection, mimicking biological systems.

Main Methods:

• Developed a population of stochastic artificial neurons using monolayer MoS2 field-effect transistors (FETs).
• Implemented optimal white Gaussian noise to enhance signal detection via stochastic resonance.
• Utilized population voting mechanisms for robust and unambiguous signal identification.
• Quantified energy expenditure at approximately 10s of nano-Joules per detection event.

Main Results:

• Successfully detected weak signals that would otherwise be invisible.
• Achieved significant reduction in detection latency through population coding principles.
• Demonstrated unambiguous signal detection even in the presence of substantial noise.
• Confirmed frugal energy expenditure, in the order of 10s of nano-Joules.

Conclusions:

• The proposed artificial neural system effectively mimics biological strategies for low-energy, high-sensitivity signal detection.
• This approach offers a promising pathway for developing energy-efficient sensors for the Internet of Things (IoT).
• Findings have implications for remote sensing and other applications demanding minimal power consumption.