We have developed an algorithm that constructs a model of a reconfigurable optical interferometer, independent of specific architectural constraints. The programming of unitary transformations on the interferometer's optical modes relies on either an analytical method for deriving the unitary matrix from a set of phase shifts or an optimization routine when such decomposition is not available. Our algorithm employs a supervised learning approach, aligning the interferometer model with a training set derived from the device being studied. A straightforward optimization procedure leverages this trained model to determine the phase shifts of the interferometer with a specific architecture, obtaining the required unitary transformation. This approach enables the effective tuning of interferometers without requiring a precise analytical solution, paving the way for the exploration of new interferometric circuit architectures.
Integrated photonic technologies have recently shown significant advances, enabling the possibility to implement reconfigurable interferometers with increasing size. One of the main tasks to fully exploit the capabilities of reconfigurable integrated interferometers is the possibility to precisely program their operation to perform a desired target unitary. While recipes are known for circuit layouts based on a cascade of beam-splitter and phase-shifter operations, a methodology applicable for reconfigurable continuously-coupled waveguide arrays is currently missing. Here, we devise a machine learning based approach for this task, using a black box methodology that does not rely on precise a-priori modeling of the circuit internal architectures. We verify the effectiveness and the robustness of this approach via numerical simulations on different continuously-coupled waveguides layouts, either with planar or 3D structures. The proposed method makes use of a limited number of single- and two-photon measurements, making it suitable for optical quantum information processing. The obtained results open the perspective of employing this methodology as an effective tool to program the operation of integrated interferometers designed via different architectures.
We developed an algorithm that constructs a model of a reconfigurable optical interferometer, independent of specific architectural constraints. The programming of unitary transformations on the interferometer's optical modes relies on either an analytical method for deriving the unitary matrix from a set of phase shifts or an optimization routine when such decomposition is not available. Our algorithm employs a supervised learning approach, aligning the interferometer model with a training set derived from the device being studied. A straightforward optimization procedure leverages this trained model to determine the phase shifts of the interferometer with a specific architecture, obtaining the required unitary transformation. This approach enables the effective tuning of interferometers without requiring a precise analytical solution, paving the way for the exploration of new interferometric circuit architectures.
Reconfigurable photonics have rapidly become an invaluable tool for information processing. Light-based computing accelerators are promising for boosting neural network learning and inference and optical interconnects are foreseen as a solution to the information transfer bottleneck in high-performance computing. In this study, we demonstrate the successful programming of a transformation implemented using a reconfigurable photonic circuit with a non-conventional architecture. The core of most photonic processors is an MZI-based architecture that establishes an analytical connection between the controllable parameters and circuit transformation. However, several architectures that are substantially more difficult to program have improved robustness to fabrication defects. We use two algorithms that rely on different initial datasets to reconstruct the circuit model of a complex interferometer, and then program the required unitary transformation. Both methods performed accurate circuit programming with an average fidelity greater than 99% and 97%, respectively. Our results provide a strong foundation for the introduction of non-conventional interferometric architectures for photonic information processing.
Coherent multiport interferometers are a promising approach to realize matrix multiplication in integrated photonics. However, most known architectures - such as MZI and beamsplitter meshes, as well as more general interferometers - suffer from complicated procedures for mapping the matrix elements of the desired transformation to specific phaseshifts in the device. We point out that the high programming complexity is intrinsic, rather than accidental. At the same time, we argue that interferometers admitting efficient programming algorithms in general yield a much lower useful output energy, which ultimately limits their accuracy and energy efficiency.
We demonstrate that AI-driven program synthesis can autonomously discover fundamental strategies for decomposing unitary matrices in photonic networks. By extending DreamCoder to complex-valued linear algebra, the system generates decomposition programs achieving the minimal
N(N−1)/2 Mach-Zehnder interferometers, distinct from both Reck and Clements architectures. Learned programs encode dimension-agnostic invariants: strategies discovered for
5×5 matrices generalize to higher dimensions such as
64×64. The discovered programs encode interpretable, dimension-agnostic construction rules. These rules generalize across matrix sizes without retraining, demonstrating that autonomous program synthesis can serve as a scalable paradigm for algorithm discovery and the automated design of universal unitary operators. Beyond universal decompositions, the system automatically exploits matrix structure to reduce the interferometer count below the universal theoretical bound. For instance, for Householder matrices, it discovers a dimension-independent rule that requires only
2N−3 MZIs. This achieves linear, rather than quadratic, scaling and generalizes to arbitrary
N without retraining. For matrices obtained from the singular value decomposition of sparse matrices, reductions generally increase with sparsity, reaching up to 38% fewer MZIs than the universal theoretical bound
N(N−1)/2 at 95% sparsity. These MZI reductions translate directly into practical hardware benefits for scalable photonic implementations. Taken together, the system functions as a single unified engine that discovers both universal decomposition rules and matrix-specific optimizations, without being provided with the structural or analytical properties of the input matrices.