Flash Memory for Synaptic Plasticity in Neuromorphic Computing: A Review.

Area of Science:

• Neuromorphic Engineering
• Computer Science
• Materials Science

Background:

• The exponential growth of data necessitates energy-efficient solutions for storage and processing.
• Neuromorphic electronics, mimicking biological neural systems, offer a path beyond the von Neumann bottleneck.
• In-memory computing with synaptic devices promises reduced energy consumption for computational tasks.

Purpose of the Study:

• To review recent advancements in neuromorphic computing utilizing NOR, AND, and NAND flash memory.
• To provide a comprehensive overview of flash memory-based neuromorphic computing.
• To highlight the potential of flash memory in optimizing energy efficiency for neural networks.

Main Methods:

• Analysis of multi-bit non-volatility and biologically inspired features in synaptic devices.
• Exploration of Ohm's law for energy reduction in multiplication and accumulation operations.
• Review of array architectures, operational methods, and electrical properties of NOR, AND, and NAND flash memory.

Main Results:

• Flash memory, particularly NOR, AND, and NAND types, demonstrates significant potential for large-scale data storage in neuromorphic systems.
• Flash memory-based synaptic devices can effectively reduce energy consumption in essential computing operations.
• Integration of flash memory enables efficient implementation of various neural network designs.

Conclusions:

• Flash memory is a competitive technology for energy-efficient neuromorphic computing and in-memory processing.
• Understanding flash memory characteristics is crucial for optimizing its application in diverse neural network architectures.
• This review provides insights for advancing flash memory-based neuromorphic computing across various applications.