Hybrid Digital-Neuromorphic Computing Services: Interoperability, Interfacing and Programming

Purpose and goal:

The project focused on the challenges associated with integrating neuromorphic information technology into the existing digital infrastructure and the edge-to-cloud continuum. Concepts for digital interfaces and abstractions were developed and examined through prototype implementations based on SNN and pub/sub/query models. The objective is to enable energy-efficient and massively distributed artificial intelligence (AI) based on the efficient and flexible use of various types of hardware from edge to cloud.

Expected results and effects:

The project has developed a concept for digital interfaces to neuromorphic information technology, enabling integration with digital infrastructure in an edge-to-cloud continuum. By developing three prototypes, the project has explored the modularization of spiking neural networks, programming of neuromorphic systems, and the implementation of digital abstractions.

The results contribute to skills supply and industrial renewal, with the potential for innovations that strengthen Sweden´s position in energy-efficient AI. Approach and implementation:

The project employed an iterative, agile methodology supported by two workshops organized by Vinnova. In the initial phase, the concept for digital interfaces and programming was developed based on a survey of the field and discussions with key representatives in industry and academia. Subsequently, concepts for modularization and communication were developed.

Prototypes were implemented and results evaluated in project meetings. Questions and concepts have been presented at conferences and seminars.