Photonic reservoir computing overcomes readout size limits with compression
Photonic reservoir computing with dimensionally compressed readout
Neural and Evolutionary Computing
Summary
Some hardware-based computing systems struggle because their output parts can't handle very large data sizes. The authors study a way to compress data from the system into smaller chunks using random projections, which keeps important information intact. They compare a small system with a large one that compresses its outputs and show that the compressed approach can sometimes work better. This method could help build bigger, more powerful photonic computing devices without being limited by output size.
What this means in practice
- •For hardware engineers: Implement photonic reservoir computers with smaller physical readout layers while maintaining performance through output compression.
- •For signal processing teams: Use dimensionality compressed outputs to improve system scalability without sacrificing information processing for time series prediction tasks.
Authors
Gerald Kobi, Mohab Abdalla, Miguel C. Soriano, Damien Rontani
Abstract
This work addresses a hardware constraint in reservoir computing: the limited size of the readout layer imposed by systems with a physical readout. We investigate a strategy to accommodate this constraint based on random projection, which compresses high-dimensional reservoir states into a lower-dimensional subspace while preserving key properties of the source space and information- processing capabilities. To evaluate this approach, we compare a small, standalone time delay reservoir against a larger configuration whose output is projected down to match the same restricted readout dimension. Using task-independent metrics, we demonstrate that the distribution of information-processing capacities may differ between the two configurations, even at identical readout sizes. Furthermore, we perform a comprehensive hyperparameter scan to assess how both systems behave under varying physical regimes. Finally, we benchmark this approach on the standard NARMA10 task, showing that the random projection framework can yield superior performance compared to a standalone constrained reservoir, within specific compression range. These results provide a scalable pathway to bypass physical readout bottlenecks in hardware-based reservoir computing.