Steven Abreu (RUG)

Steven Abreu (ESR1) has co-authored a widely cited tutorial article that introduces physical reservoir computing as a way of “programming” physical computers of a certain class, in collaboration with researchers at TU Dresden. This review was published in the Neuromorphic Computing and Engineering journal as “Matteo Cucchi*, Steven Abreu*, Giuseppe Ciccone, Daniel Brunner, Hans Kleemann (2022): Hands-on reservoir computing: a tutorial for practical implementation” (open access at 10.1088/2634-4386/ac7db7). The article was featured as a highlight of 2022 ” (2nd most read article of the journal). ESR1 further co-led the effort to develop an intermediate representation that is supported by 7 neuromorphic simulators and 4 neuromorphic hardware platforms. This is a significant advance towards a computational abstraction for neuromorphic systems which currently fully support digital neuromorphic platforms. Importantly, digital neuromorphic platforms are represented as continuous-time dynamical systems, which provides the basis for future support of analog neuromorphic platforms as well. This work is currently in review at a journal and available as a pre-print: “Jens E. Pedersen*, Steven Abreu*, et al. (2023): Neuromorphic Intermediate Representation: A Unified Instruction Set for Interoperable Brain-Inspired Computing” (arxiv:2311.14641). ESR1 is further working on a review of programming paradigms that are suitable for neuromorphic computing which is available as a work-in-progress preprint in “Steven Abreu (2023): Concepts and Paradigms for Neuromorphic Programming” (arxiv:2310.18260). This review makes neuromorphic computing accessible to computer scientists and programming language researchers in an effort to develop strategies, languages and paradigms for programming neuromorphic systems. An extension of this work was also presented as a poster at the NNPC conference (Hannover, Germany) in October 2024.