Memristive and CMOS Devices for Neuromorphic Computing

Area of Science:

• Computer Science
• Materials Science
• Neuroscience

Background:

• Conventional von Neumann architecture faces limitations in energy efficiency and processing power.
• The human brain offers a highly efficient and compact model for computation.
• Neuromorphic computing aims to replicate brain-like processing for enhanced performance.

Purpose of the Study:

• To provide an overview of promising device concepts for neuromorphic computing.
• To discuss the potential of complementary metal-oxide semiconductor (CMOS) and memristive technologies.
• To highlight challenges and future perspectives in neuromorphic device development.

Main Methods:

• Review of CMOS-based floating-gate memory devices for artificial neural networks.
• Discussion of various memristive device concepts for deep and spiking neural networks.
• Analysis of technological challenges and future outlooks in neuromorphic computing.

Main Results:

• CMOS and memristive technologies show significant promise for neuromorphic applications.
• Floating-gate devices are suitable for artificial neural networks.
• Memristive devices offer potential for deep and spiking neural network architectures.

Conclusions:

• Novel device concepts are crucial for achieving brain-like scalability and low-power operation.
• Both CMOS and memristive technologies are key enablers for future neuromorphic systems.
• Overcoming technological challenges will pave the way for advanced neuromorphic computing.