arXiv subjects

Mikhail Saygin

Publications and source records attributed to Mikhail Saygin.

4 recordsLinked to original sources

arXiv2026

LLM-Guided Evolutionary Search for Algebraic T-Count Optimization

T-count minimization is an NP-hard problem that arises in fault-tolerant quantum compilation. In the parity-matrix representation, which captures the non-Clifford part of a quantum circuit, algebraic optimizers such as TODD can achieve state-of-the-art results. However, heuristics fixed in advance determine which transformation is applied, limiting the exploration of alternative trajectories that may lead to better solutions. We show how LLM-guided evolutionary search can help explore these degrees of freedom, which VarTODD exposes through a policy that determines how to allocate the available evaluation budget and how to guide the search. Execution diagnostics guide LLM-generated revisions to both numerical search parameters and program logic, with the possibility of exploiting earlier results by starting from intermediate matrices saved during previous runs. This formulation turns heuristic design into an automated search problem and, across all evaluated instances, matches or improves on the lowest listed reference T-count for every evaluated instance; for example, on the GF(2^n) multiplier benchmarks, it yields a mean relative reduction of 7.0%.

quant-ph↗

arXiv2025

Low-rank surrogate modeling and stochastic zero-order optimization for training of neural networks with black-box layers

The growing demand for energy-efficient, high-performance AI systems has led to increased attention on alternative computing platforms (e.g., photonic, neuromorphic) due to their potential to accelerate learning and inference. However, integrating such physical components into deep learning pipelines remains challenging, as physical devices often offer limited expressiveness, and their non-differentiable nature renders on-device backpropagation difficult or infeasible. This motivates the development of hybrid architectures that combine digital neural networks with reconfigurable physical layers, which effectively behave as black boxes. In this work, we present a framework for the end-to-end training of such hybrid networks. This framework integrates stochastic zeroth-order optimization for updating the physical layer's internal parameters with a dynamic low-rank surrogate model that enables gradient propagation through the physical layer. A key component of our approach is the implicit projector-splitting integrator algorithm, which updates the lightweight surrogate model after each forward pass with minimal hardware queries, thereby avoiding costly full matrix reconstruction. We demonstrate our method across diverse deep learning tasks, including: computer vision, audio classification, and language modeling. Notably, across all modalities, the proposed approach achieves near-digital baseline accuracy and consistently enables effective end-to-end training of hybrid models incorporating various non-differentiable physical components (spatial light modulators, microring resonators, and Mach-Zehnder interferometers). This work bridges hardware-aware deep learning and gradient-free optimization, thereby offering a practical pathway for integrating non-differentiable physical components into scalable, end-to-end trainable AI systems.

cs.LG↗

arXiv2019

Robust architecture for programmable universal unitaries

The decomposition of large unitary matrices into smaller ones is important, because it provides ways to realization of classical and quantum information processing schemes. Today, most of the methods use planar meshes of tunable two-channel blocks, however, the schemes turn out to be sensitive to fabrication errors. We study a novel decomposition method based on multi-channel blocks. We have shown that the scheme is universal even when the block`s transfer matrices are chosen at random, making it virtually insensitive to errors. Moreover, the placement of the variable elements can be arbitrary, so that the scheme is not bound to specific topologies. Our method can be beneficial for large-scale implementations of unitary transformations by techniques, which are not of wide proliferation today or yet to be developed.

quant-ph↗

arXiv2016

Low-loss single-mode integrated waveguides in soda-lime glass

Low-loss single-mode optical waveguide fabrication process in extra-white soda-lime glass is demonstrated. Waveguiding structures are formed in bulk substrates employing femtosecond laser writing technology. The combination of a slit beam-shaping method and a multiscan fabrication process enables printing of waveguides with a well-defined symmetric cross-section profile. Fabricated waveguides exhibit 0.86 dB/cm propagation loss for 800~nm wavelength. Bending loss in the waveguides are addressed experimentally and compared with a model for radiation loss.

physics.optics↗