arXiv:2601.22300v3 Announce Type: replace-cross Abstract: We propose a deep photonic neuromorphic network (PNN) architecture based on phase-change material (PCM) synapses and local optical feedback for online, unsupervised Hebbian learning. The proposed architecture combines optical vector-matrix multiplication, non-volatile PCM synaptic weighting, and local coincidence-driven synaptic adaptation within a multilayer photonic crossbar framework compatible with photonic integrated circuits. Unlike conventional PNNs that rely on externally computed gradients, repeated optical-electrical-optical c
Advances in photonic integrated circuits and phase-change materials are converging to enable practical implementations of all-optical AI hardware, moving beyond theoretical proposals.
This development indicates a potential paradigm shift in AI hardware by enabling faster, more energy-efficient, and potentially scalable AI processing without constant optical-electrical conversions.
The reliance on external gradient computation and electrical-optical conversions in neuromorphic networks could be significantly reduced, leading to more autonomous and efficient on-chip learning.
- · Photonic integrated circuit manufacturers
- · AI hardware developers
- · Data centers
- · Edge AI applications
- · Traditional electronic AI accelerator developers (if disruptive)
- · Companies heavily invested in O-E-O conversion technologies
Increased research and development into all-optical AI hardware architectures and materials.
Reduced energy consumption and increased processing speed for specific AI tasks, leading to a competitive advantage for early adopters.
The proliferation of more powerful and ubiquitous AI at the edge due to lower power requirements and higher speeds, decentralizing AI compute.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG