Exploring the landscapes of "computing": digital, neuromorphic, unconventional -- and beyond

Abstract

The acceleration race of digital computing technologies seems to be steering toward impasses -- technological, economical and environmental -- a condition that has spurred research efforts in alternative, "neuromorphic" (brain-like) computing technologies. Furthermore, since decades the idea of exploiting nonlinear physical phenomena "directly" for non-digital computing has been explored under names like "unconventional computing", "natural computing", "physical computing", or "in-materio computing". This has been taking place in niches which are small compared to other sectors of computer science. In this paper I stake out the grounds of how a general concept of "computing" can be developed which comprises digital, neuromorphic, unconventional and possible future "computing" paradigms. The main contribution of this paper is a wide-scope survey of existing formal conceptualizations of "computing". The survey inspects approaches rooted in three different kinds of background mathematics: discrete-symbolic formalisms, probabilistic modeling, and dynamical-systems oriented views. It turns out that different choices of background mathematics lead to decisively different understandings of what "computing" is. Across all of this diversity, a unifying coordinate system for theorizing about "computing" can be distilled. Within these coordinates I locate anchor points for a foundational formal theory of a future computing-engineering discipline that includes, but will reach beyond, digital and neuromorphic computing.

arXiv commentAn extended and carefully revised version of this manuscript has now (March 2021) been published as "Toward a generalized theory comprising digital, neuromorphic, and unconventional computing" in the new open-access journal Neuromorphic Computing and Engineering

References in corpus (19)

• Self-Attention Generative Adversarial Networks
• Event-based Vision: A Survey
• Recent Advances in Physical Reservoir Computing: A Review
• Memory and information processing in neuromorphic systems
• Dota 2 with Large Scale Deep Reinforcement Learning
• A scalable multi-core architecture with heterogeneous memory structures for Dynamic Neuromorphic Asynchronous Processors (DYNAPs)
• Multiply periodic states and isolated skyrmions in an anisotropic frustrated magnet
• A magnetic skyrmion as a non-linear resistive element - a potential building block for reservoir computing
• Neural-Symbolic Learning and Reasoning: A Survey and Interpretation
• Boosting computational power through spatial multiplexing in quantum reservoir computing
• Computing with Networks of Nonlinear Mechanical Oscillators
• Pattern theory: the mathematics of perception
• Task Agnostic Continual Learning via Meta Learning
• Periodicity doubling cascades: direct observation in ferroelastic materials
• Extending Landauer's Bound from Bit Erasure to Arbitrary Computation
• Dimensions of Timescales in Neuromorphic Computing Systems
• Attractor metadynamics in terms of target points in slow-fast systems: adiabatic vs. symmetry protected flow in a recurrent neural network
• The logic of quantum mechanics - Take II
• A Survey on Analog Models of Computation

Cited by in corpus (12)

• Opportunities in Quantum Reservoir Computing and Extreme Learning Machines
• Dynamical phase transitions in quantum reservoir computing
• Vector Symbolic Architectures as a Computing Framework for Emerging Hardware
• Integration of Neuromorphic AI in Event-Driven Distributed Digitized Systems: Concepts and Research Directions
• Molecular dynamics simulation of nanofilament breakage in neuromorphic nanoparticle networks
• Dimensions of Timescales in Neuromorphic Computing Systems
• Optical neuromorphic computing via temporal up-sampling and trainable encoding on a telecom device platform
• Integer Factorization with Compositional Distributed Representations
• The sounds of science a symphony for many instruments and voices part II
• Neuromorphic nanocluster networks: Critical role of the substrate in nano-link formation
• Improved Tomographic Estimates by Specialised Neural Networks
• Quantum process matrices as images: new tools to design novel denoising methods