Polarization-Coded Bulk-Optical OR Modules for Finite-Length Stochastic Binary Sequences: State Selection, Intensity Fluctuations, and Robustness
We investigate the statistical performance of finite-length stochastic binary sequences processed by three polarization-coded bulk-optical OR architectures: threshold/ratiometric readout, electro-optically controlled polarization transformation, and thresholded all-optical switching. The architectures implement the same bitwise Boolean operation through different physical decision mechanisms. A common numerical framework combines state-conditioned optical response, Bernoulli sequence generation, input dependence, source-intensity mismatch, pulse-to-pulse fluctuations, and architecture-specific parameter uncertainties. For Bernoulli inputs with factorized joint probabilities, simulations reproduce the expected decrease in probability-estimation error with increasing sequence length. At input probabilities of 0.30 and 0.40, the ideal-output root-mean-square error decreases from 0.0608 for 64-bit sequences to 0.00767 for 4096-bit sequences. Correlation analysis demonstrates the dependence of the probability-domain OR response on joint input statistics. Intensity-noise simulations identify architecture-dependent classification errors and operating limitations under specified parameter distributions. A synthetic two-feature event-aggregation example illustrates the application of the OR operation to probabilistic feature combination. The results characterize the computational behavior and physical constraints of the three proposed architectures within a reproducible numerical framework.
Authors
- Myroslav Strynadko (ORCID: https://orcid.org/0009-0004-9549-2198)
Publication Details
- Published
- 2026-11-18
- DOI
- https://doi.org/10.1364/opticaopen.34948705.v1
- Primary Topic
- Neural Networks and Reservoir Computing
- Type
- preprint