Distance Computation Based on Coupled Spin-Torque Oscillators: Application to Image Processing

Area of Science:

• Spintronics
• Nano-oscillators
• Non-Boolean Computing

Background:

• Coupled-oscillator networks show potential for non-Boolean computation.
• Spin-torque oscillators (STOs) are CMOS compatible, integrable, scalable, and tunable.

Purpose of the Study:

• To propose and demonstrate a hybrid spintronic and CMOS architecture for computing multidimensional norms.
• To evaluate the system's performance as a distance metric and convolution primitive for image processing.

Main Methods:

• Experimental demonstration of a hybrid system with four injection-locked STOs and a CMOS detector.
• Simulations using experimental data to assess computational capabilities.
• Energy and scaling analysis comparing STO-based and CMOS-based systems.

Main Results:

• The hybrid system successfully computes multidimensional norms.
• Demonstrated utility as a distance metric and convolution primitive for image processing.
• STO-based system shows higher efficiency and an order of magnitude faster computation speed for high-dimensional data.

Conclusions:

• The proposed STO-based coupled-oscillator architecture offers a promising approach for efficient, high-speed non-Boolean computation.
• This hybrid system presents a viable alternative to traditional CMOS for specific computational tasks, particularly in image processing.