A study in Nature Computational Science describes INSPIRE, a method for training photonic circuits directly in the physical hardware rather than relying entirely on computer simulations. Experiments reported single-shot photonic meta-learning, with 251-fold model compression and 136-fold task-specific training acceleration in the demonstrated setup. These are study-specific results, not universal performance guarantees.
Read more https://www.nature.com/articles/s43588-026-01057-y
