Neurosymbolic Computing: Hardware–Software Co-design
The rapid evolution of artificial intelligence (AI) has given rise to neuro-symbolic (NeSy) computing, an emerging paradigm that unites the pattern-recognition power of neural networks with the logical reasoning of symbolic methods. While neural models are highly effective for data-driven learning, symbolic systems excel at structured inference. When combined, they offer complementary strengths – producing interpretable, explainable systems that can operate over complex knowledge sources, adapt to small or noisy datasets, and generalize across tasks. However, investigations of NeSy AI from a systems perspective remain relatively sparse. In this study, we survey current hardware–software co-design strategies tailored for NeSy computing and highlight the challenges in boosting performance and efficiency within heterogeneous platforms. By exploring the synergy between NeSy algorithms and system-level design, our work seeks to close this gap and accelerate progress in the field.
