28 November 2024 Thu 4 pm

Chaos computing is one of the unconventional computing paradigms which utilizes sensitive dependence on initial conditions (SDIC) yet the deterministic behavior of nonlinear systems [1, 2]. The key advantages of this computing system are 1) reconfigurability and 2) fault-tolerance within a single computing unit. In other words, it can dynamically morph into different logic or arithmetic operations by adjusting its initial conditions, while maintaining output stability against errors during physical computation. Additionally, 3) parallel computation facilitates it in meeting the demands of current computing tasks, such as machine learning and optimization problems.

However, the disadvantages include 1) reduced reliability (ironically caused by SDIC), 2) the complexity of operational conditions, and 3) synchronization issues between nodes. These challenges would hinder the development of chaos computing as a candidate for fast, scalable, and energy-efficient computing. Therefore, innovative solutions are needed to prevent it from evolving into bulky devices and overly complex algorithms.

Recent advancements in spintronics offer promising solutions to these obstacles and have the potential to enhance chaos computing. For instance, chaos in the conversion of topological charge in magnetic skyrmions demonstrates the feasibility of applying spintronics to advanced chaos computing [3]. This seminar will provide an overview of chaos computing, propose improvements to the computing algorithm, and present examples of physical implementations using spintronics.

[1] Dynamics Based Computation, S. Sinha and W. L. Ditto, Phys. Rev. Lett. 81, 2156 (1998)

[2] Computing with distributed chaos, S. Sinha and W. L. Ditto, Phys. Rev. E 60, 363 (1999)

[3] Reconfigurable all-in-one chaotic computing with skyrmions: Leveraging periodic modulations of perpendicular magnetic anisotropy, G. Park and S. -K. Kim, Phys. Rev. B 109, 174420 (2024)

< Previous