Pruning Coherent Integrated Photonic Neural Networks Using the Lottery Ticket Hypothesis
This work addresses efficiency challenges in photonic neural networks for hardware applications, but it is incremental as it applies an existing hypothesis to a new domain.
The paper tackled the problem of large footprint and high static power consumption in singular-value-decomposition-based coherent integrated photonic neural networks (SC-IPNNs) by proposing a hardware-aware pruning method based on the lottery ticket hypothesis, achieving up to 89% pruning of phase angles with less than 5% accuracy loss and up to 86% reduction in static power consumption.
Singular-value-decomposition-based coherent integrated photonic neural networks (SC-IPNNs) have a large footprint, suffer from high static power consumption for training and inference, and cannot be pruned using conventional DNN pruning techniques. We leverage the lottery ticket hypothesis to propose the first hardware-aware pruning method for SC-IPNNs that alleviates these challenges by minimizing the number of weight parameters. We prune a multi-layer perceptron-based SC-IPNN and show that up to 89% of the phase angles, which correspond to weight parameters in SC-IPNNs, can be pruned with a negligible accuracy loss (smaller than 5%) while reducing the static power consumption by up to 86%.