arXiv subjects

Tigran Sedrakyan

Publications and source records attributed to Tigran Sedrakyan.

11 recordsLinked to original sources

arXiv2026

Private training in quantum machine learning

With the emergence of machine learning (ML) models trained on large datasets containing potentially sensitive data, a major question in AI safety is how to make learning private with respect to the training data. Similar to classical machine learning, quantum machine learning (QML) models are not devoid of privacy vulnerabilities. Differential privacy (DP) is a standard tool for training ML models on sensitive data, but its impact in QML remains poorly understood. In this work we study private training in hybrid variational QML models using a classical private DP-SGD optimizer applied to pipelines with classical inputs and outputs. We analyze the interplay between gradient clipping and calibrated noise addition in DP-SGD, and its impact on optimization and accuracy for noisy and noiseless quantum models. We first explain why quantum noise does not provide a satisfactory replacement for the calibrated noise in DP-SGD for ensuring privacy. We then show how the deterministic bounds on gradient norms for a wide class of quantum models translate into explicit control of the detrimental clipping bias introduced by DP-SGD. Finally, we formulate a numerical comparison protocol under fixed clipping threshold and privacy budget and evaluate it on synthetic and image-classification tasks for equivalent quantum and classical models. Our results suggest that quantum models can retain higher accuracy in private-training regimes where the formal privacy guarantee is ensured by a classical DP-SGD mechanism.

quant-ph↗

arXiv2025

Quantum Image Loading: Hierarchical Learning and Block-Amplitude Encoding

Given the excitement for the potential of quantum computing for machine learning methods, a natural subproblem is how to load classical data into a quantum state. Leveraging insights from [GST24] where certain qubits play an outsized role in the amplitude encoding, we extend the hierarchical learning framework to encode images into quantum states. We successfully load digits from the MNIST dataset as well as road scenes from the Honda Scenes dataset. Additionally, we consider the use of block amplitude encoding, where different parts of the image are encoded in a tensor product of smaller states. The simulations and overall orchestration of workflows was done on the BlueQubit platform. Finally, we deploy our learned circuits on both IBM and Quantinuum hardware and find that these loading circuits are sufficiently shallow to fit within existing noise rates.

quant-ph↗

arXiv2025

Quantum Image Classification: Experiments on Utility-Scale Quantum Computers

We perform image classification on the Honda Scenes Dataset on Quantinuum's H-2 and IBM's Heron chips utilizing up to 72 qubits and thousands of two-qubit gates. For data loading, we extend the hierarchical learning to the task of approximate amplitude encoding and block amplitude encoding for commercially relevant images up to 2 million pixels. Hierarchical learning enables the training of variational circuits with shallow enough resources to fit within the classification pipeline. For comparison, we also study how classifier performance is affected by using piecewise angle encoding. At the end of the VQC, we employ a fully-connected layer between measured qubits and the output classes. Some deployed models are able to achieve above 90\% accuracy even on test images. In comparing with classical models, we find we are able to achieve close to state of the art accuracy with relatively few parameters. These results constitute the largest quantum experiment for image classification to date.

quant-ph↗

arXiv2025

Error-mitigated photonic quantum circuit Born machine

In this article, we study quantum circuit Born machines (QCBMs) in the context of photonic quantum computing. QCBMs are a popular choice of quantum generative machine learning models, and we present a QCBM designed for linear optics. We show that a recently developed error mitigation technique called recycling mitigation greatly improves the training of QCBMs in realistic scenarios with photon loss, which is the primary source of noise in photonic systems. We demonstrate this through numerical simulations and through an experiment on a quantum photonic integrated processor. We expect our work to pave the way towards more demonstrations of error mitigation techniques tailored to photonic devices which can enhance the performance of a quantum algorithm.

quant-ph↗

arXiv2025

Photonic quantum generative adversarial networks for classical data

In generative learning, models are trained to produce new samples that follow the distribution of the target data. These models were historically difficult to train, until proposals such as Generative Adversarial Networks (GANs) emerged, where a generative and a discriminative model compete against each other in a minimax game. Quantum versions of the algorithm were since designed, both for the generation of classical and quantum data. While most work so far has focused on qubit-based architectures, in this article we present a quantum GAN based on linear optical circuits and Fock-space encoding, which makes it compatible with near-term photonic quantum computing. We demonstrate that the model can learn to generate images by training the model end-to-end experimentally on a single-photon quantum processor.

quant-ph↗

arXiv2024

Trapping Hard-Core Bosons in Flatband Lattices

We investigate 1D and 2D cross-stitch lattices with hard-core bosons and analytically construct exact groundstates that feature macroscopic degeneracy. The construction relies on the presence of a flatband in the single particle spectrum and the orthogonality of the associated compact localized states (CLS). Up to filling fraction $ν=1/2$, the groundstate is constructed by occupying the CLS. Exactly at $ν=1/2$, the groundstate becomes a Wigner crystal. For higher filling fractions, the groundstate is constructed by filling the CLS sites completely one by one. Macroscopic degeneracy arises from the multiple choices available when occupying or filling the CLS sites. An occupied CLS acts as an impenetrable barrier for bosons both in 1D and 2D, leading to Hilbert space fragmentation. A similar phenomenology also holds for hard-core bosons on the diamond chain and its higher dimensional generalizations. We also discuss the mapping of these hard-core models onto spin models with quantum many-body scars.

cond-mat.str-el↗

arXiv2023

Isolated flat bands in 2D lattices based on a novel path-exchange symmetry

The increased ability to engineer two-dimensional (2D) systems, either using materials, photonic lattices, or cold atoms, has led to the search for 2D structures with interesting properties. One such property is the presence of flat bands. Typically, the presence of these requires long-ranged hoppings, fine-tuning of nearest neighbor hoppings, or breaking time-reversal symmetry by using a staggered flux distribution in the unit cell. We provide a prescription based on carrying out projections from a parent system to generate different flat band systems. We identify the conditions for maintaining the flatness and identify a path-exchange symmetry in such systems that cause the flat band to be degenerate with the other dispersive ones. Breaking this symmetry leads to lifting the degeneracy while still preserving the flatness of the band. This technique does not require changing the topology nor breaking time-reversal symmetry as was suggested earlier in the literature. The prescription also eliminates the need for any fine-tuning. Moreover, it is shown that the subsequent projected systems inherit the precise fine-tuning conditions that were discussed in the literature for similar systems, in order to have and isolate a flat band. As examples, we demonstrate the use of our prescription to arrive at the flat band conditions for popular systems like the Kagome, the Lieb, and the Dice lattices. Finally, we are also able to show that a flat band exists in a recently proposed chiral spin-liquid state of the Kagome lattice only if it is associated with a gauge field that produces a flux modulation of the Chern-Simons type.

cond-mat.str-el↗

arXiv2021

Theory of competing Chern-Simons orders and emergent phase transitions

Based on the Chern-Simons fermionization of spin-1/2 operators, we propose a systematic framework to investigate the competition between emergent phases in frustrated two-dimensional XY quantum magnets. Application of the method to the antiferromagnetic honeycomb spin-1/2 $J_1$-$J_2$ XY model reveals an unconventional phase transition between two Chern-Simons orders: the Chern-Simons superconductor and the exciton insulator of Chern-Simons fermions. We show that in the spin language, this transition translates to the transition from the planar Néel state to the non-uniform chiral spin-liquid that was proposed earlier in the literature. Namely, the Chern-Simons superconductor describes the planar Néel state, while the Chern-Simons exciton insulator corresponds to the non-uniform chiral spin-liquid. These results are further confirmed by our high-precision tensor network calculation, which provides the first numerical evidence for the transition from Néel order to a non-uniform chiral spin-liquid. We argue that the developed method can be applied to other frustrated quantum magnets of XXZ type and can detect unconventional phase transitions.

cond-mat.str-el↗

arXiv2019

Is the composite fermion state of Graphene a doped Chern insulator?

Graphene in the presence of a strong external magnetic field is a unique attraction for investigations of the fractional quantum Hall (fQH) states with odd and even denominators of the fraction. Most of the attempts to understand Graphene in the strong-field regime were made through exploiting the universal low-energy effective description of Dirac fermions emerging from the nearest neighbor hopping model of electrons on a honeycomb lattice. We highlight that accounting for the next-nearest-neighbor hopping terms in doped Graphene can lead to a unique redistribution of magnetic fluxes within the unit cell of the lattice. While this affects all the fQH states, it has a striking effect at a half-filled Landau-level state: it leads to a composite fermion state that is equivalent to the doped topological Chern insulator on a honeycomb lattice. At energies comparable to the Fermi energy, this state possesses a Haldane gap in the bulk proportional to the next-nearest-neighbor hopping and density of dopants. We argue that this microscopically derived energy gap survives the projection to the lowest band. We also conjecture that the gap should be present in a microscopic theory giving the recently proposed particle-hole symmetric Dirac composite fermion scenario of the half-filled Landau-level. The proposed gap is lower than the chemical potential, and is predicted to be parametrically separated from the Dirac point in the latter description. Finally we conclude by proposing experiments to detect this gap; the associated boundary mode; and encourage cold-atom setups to test other predictions of the theory.

cond-mat.str-el↗

arXiv2018

D-wave superconductivity induced by proximity to the non-uniform chiral spin liquid on a square lattice

We use the tensor network algorithm to show evidences of a non-uniform chiral spin-liquid (CSL) ground state in a frustrated spin-1/2 model on a square lattice, in the regime of moat-like band structure of the lattice, i.e., a band with infinitely degenerate energy minima attained along a closed contour in the Brillouin zone. The analytical description of the state is given by the effective field theory of a topological square-lattice fermionic Chern insulator coupled to the Chern-Simons gauge field. The observed non-uniform CSL has a substantial effect on a nearby free-fermion environment. We show that, in the presence of arbitrarily small spin exchange interaction, the CSL can endow a gauge-field-modulated effective interaction between the environmental fermions. The induced effective interaction can be attractive within a significant parameter region, leading to an instability towards d-wave superconductivity in the fermionic bath.

cond-mat.str-el↗

arXiv2011

Vortices in spin-orbit-coupled Bose-Einstein condensates

Realistic methods to create vortices in spin-orbit-coupled Bose-Einstein condensates are discussed. It is shown that, contrary to common intuition, rotation of the trap containing a spin-orbit condensate does not lead to an equilibrium state with static vortex structures, but gives rise instead to non-equilibrium behavior described by an intrinsically time-dependent Hamiltonian. We propose here the following alternative methods to induce thermodynamically stable static vortex configurations: (1) to rotate both the lasers and the anisotropic trap; and (2) to impose a synthetic Abelian field on top of synthetic spin-orbit interactions. Effective Hamiltonians for spin-orbit condensates under such perturbations are derived for most currently known realistic laser schemes that induce synthetic spin-orbit couplings. The Gross-Pitaevskii equation is solved for several experimentally relevant regimes. The new interesting effects include spatial separation of left- and right-moving spin-orbit condensates, the appearance of unusual vortex arrangements, and parity effects in vortex nucleation where the topological excitations are predicted to appear in pairs. All these phenomena are shown to be highly non-universal and depend strongly on a specific laser scheme and system parameters.

cond-mat.quant-gas↗