Factor Machine: Mixed-signal Architecture for Fine-Grained Graph-Based Computing

This paper proposes the design and implementation strategy of a novel computing architecture, the Factor Machine. The work is a step towards a general-purpose parallel system operating in a non-sequential manner, exploiting processing/memory co-integration and replacing the traditional Turing/von Neumann model of a computer system with a framework based on "factorised computation". This architecture is inspired by neural information processing principles and aims to progress the development of brain-like machine intelligence systems, through providing a computing substrate designed from the ground up to enable efficient implementations of algorithms based on relational networks. The paper provides a rationale for such machine, in the context of the history of computing, and more recent developments in neuromorphic hardware, reviews its general features, and proposes a mixed-signal hardware implementation, based on using analogue circuits to carry out computation and localised and sparse communication between the compute units.