Cognitive Computing Programming Paradigm: A Corelet Language for Composing Networks of Neurosynaptic Cores
📜 Abstract
Marching along the DARPA SyNAPSE roadmap, IBM unveils a trilogy of innovations towards the TrueNorth cognitive computing system inspired by the brain’s function and efficiency. The sequential programming paradigm of the von Neumann architecture is wholly unsuited for TrueNorth. Therefore, as our main contribution, we develop a new programming paradigm that permits construction of complex cognitive algorithms and applications while being efficient for TrueNorth and effective for programmer productivity. The programming paradigm consists of (a) an abstraction for a TrueNorth program, named Corelet, for representing a network of neurosynaptic cores that encapsulates all details except external inputs and outputs; (b) an object-oriented Corelet Language for creating, composing, and decomposing corelets; (c) a Corelet Library that acts as an ever-growing repository of reusable corelets from which programmers compose new corelets; and (d) an end-to-end Corelet Laboratory that is a programming environment which integrates with the TrueNorth architectural simulator, Compass, to support all aspects of the programming cycle from design, through development, debugging, and up to deployment. The new paradigm seamlessly scales from a handful of synapses and neurons to networks of neurosynaptic cores of progressively increasing size and complexity. The utility of the new programming paradigm is underscored by the fact that we have designed and implemented more than 100 algorithms as corelets for TrueNorth in a very short time span.
✨ Summary
The paper introduces Corelets as encapsulated, hierarchically composable abstractions for programming IBM’s TrueNorth neuromorphic architecture. A Corelet exposes only external input and output connectors while hiding internal neurons, synaptic weights, cores, and connectivity. The associated object-oriented language provides neuron, neurosynaptic-core, connector, and Corelet abstractions; composition builds larger systems from smaller reusable components, while decomposition removes encapsulation and produces a directly implementable TrueNorth network. The paper also presents a Corelet Library of reusable functional components and a Corelet Laboratory integrated with the Compass simulator for verification, model generation, input/output mapping, simulation, and visualization. A music-composer-recognition system demonstrates the approach using a liquid-state-machine module and hierarchical classifier.
Subsequent work cited the Corelet framework as the programming layer for TrueNorth applications. Research on energy-efficient convolutional networks used Corelets to compile learned network parameters into TrueNorth hardware configuration files, demonstrating a practical bridge from platform-independent model parameters to hardware-specific deployment. (pmc.ncbi.nlm.nih.gov) Other TrueNorth research identified Corelet as the hierarchical, compositional language used to develop applications, including sparse-coding experiments and event-based image-sensor processing. (pmc.ncbi.nlm.nih.gov) The framework also motivated tooling work, including a documented transcompiler from the Fugu language to IBM’s TrueNorth Corelet Programming Environment. (doi.org) Later surveys and system papers continued to classify Corelet as an early platform-specific programming approach for neuromorphic hardware, while noting that it remained closely tied to TrueNorth rather than becoming a broadly adopted cross-platform standard. (researchgate.net)