Photonic neural networks gain faster calibration from shared population data
LCAP: Population-Informed Latent Chip Adaptation from Few Output Probes for Photonic Neural Networks
Machine Learning
Summary
Photonic neural networks can perform fast AI calculations but often work less well after being physically made, because of tiny differences in each chip. The authors created a method called LCAP that learns from many previous chips to correct these differences quickly. Instead of adjusting each chip separately, LCAP uses some fixed measurements to personalize corrections using shared knowledge, improving accuracy without lengthy tuning. This approach means faster, more reliable chip use in practical AI tasks.
What this means in practice
- •For hardware engineers: Calibrate multiple photonic neural network chips rapidly using shared adaptation data and fixed output probes.
- •For optical computing developers: Improve inference accuracy on photonic chips without per-device optimization by inferring personalization from population corrections.
Authors
Tianyu Gao, Guantian Zheng
Abstract
Photonic neural networks (PNNs) offer efficient analog inference, but parameters optimized under ideal device models can degrade after fabrication, creating a persistent simulation-to-hardware (sim-to-real) gap. When many identically designed chips are deployed, calibrating each device from scratch compounds this cost. We propose Latent Chip Adaptation from Probes (LCAP), a population-informed framework that decomposes hardware adaptation into a transferable population correction and probe-inferred latent personalization. LCAP first learns a shared correction from 80 historical chips, then extracts a low-dimensional correction space from device-specific refinements. At deployment, 32 fixed unlabeled output probes infer an unseen chip's latent correction coordinates, enabling feed-forward personalization without target-device optimization. On a three-layer 64-mode MZI simulator with phase variation, beam-splitter errors, quantization, and crosstalk, accuracy improves from 80.4147% under direct deployment to 92.6860% after shared calibration and 93.3617% with LCAP. LCAP improves 27/30 unseen chips and raises worst-device accuracy from 89.18% to 90.54%.