Publications

Order by: publication date number of citations

• From complex to simple : hierarchical free-energy landscape renormalized in deep neural networks

  Hajime Yoshino

  SciPost Phys. Core 2, 005 (2020) · published 15 April 2020 | Toggle abstract · pdf

  We develop a statistical mechanical approach based on the replica method to study the design space of deep and wide neural networks constrained to meet a large number of training data. Specifically, we analyze the configuration space of the synaptic weights and neurons in the hidden layers in a simple feed-forward perceptron network for two scenarios: a setting with random inputs/outputs and a teacher-student setting. By increasing the strength of constraints,~i.e. increasing the number of training data, successive 2nd order glass transition (random inputs/outputs) or 2nd order crystalline transition (teacher-student setting) take place layer-by-layer starting next to the inputs/outputs boundaries going deeper into the bulk with the thickness of the solid phase growing logarithmically with the data size. This implies the typical storage capacity of the network grows exponentially fast with the depth. In a deep enough network, the central part remains in the liquid phase. We argue that in systems of finite width N, the weak bias field can remain in the center and plays the role of a symmetry-breaking field that connects the opposite sides of the system. The successive glass transitions bring about a hierarchical free-energy landscape with ultrametricity, which evolves in space: it is most complex close to the boundaries but becomes renormalized into progressively simpler ones in deeper layers. These observations provide clues to understand why deep neural networks operate efficiently. Finally, we present some numerical simulations of learning which reveal spatially heterogeneous glassy dynamics truncated by a finite width $N$ effect.

  5 citations

• On the generalizability of artificial neural networks in spin models

  Hon Man Yau, Nan Su

  SciPost Phys. Core 5, 032 (2022) · published 8 June 2022 | Toggle abstract · pdf

  The applicability of artificial neural networks (ANNs) is typically limited to the models they are trained with and little is known about their generalizability, which is a pressing issue in the practical application of trained ANNs to unseen problems. Here, by using the task of identifying phase transitions in spin models, we establish a systematic generalizability such that simple ANNs trained with the two-dimensional ferromagnetic Ising model can be applied to the ferromagnetic q-state Potts model in different dimensions for q>=2. The same scheme can be applied to the highly nontrivial antiferromagnetic q-state Potts model. We demonstrate that similar results can be obtained by reducing the exponentially large state space spanned by the training data to one that comprises only three representative configurations artificially constructed through symmetry considerations. We expect our findings to simplify and accelerate the development of machine learning-assisted tasks in spin-model related disciplines in physics and materials science.

  5 citations

• Holonomic implementation of CNOT gate on topological Majorana qubits

  Alessio Calzona, Nicolas P. Bauer, Björn Trauzettel

  SciPost Phys. Core 3, 014 (2020) · published 3 December 2020 | Toggle abstract · pdf

  The CNOT gate is a two-qubit gate which is essential for universal quantum computation. A well-established approach to implement it within Majorana-based qubits relies on subsequent measurement of (joint) Majorana parities. We propose an alternative scheme which operates a protected CNOT gate via the holonomic control of a handful of system parameters, without requiring any measurement. We show how the adiabatic tuning of pair-wise couplings between Majoranas can robustly lead to the full entanglement of two qubits, insensitive with respect to small variations in the control of the parameters.

  5 citations

• Probing Chern number by opacity and topological phase transition by a nonlocal Chern marker

  Paolo Molignini, Bastien Lapierre, Ramasubramanian Chitra, Wei Chen

  SciPost Phys. Core 6, 059 (2023) · published 29 August 2023 | Toggle abstract · pdf

  In 2D semiconductors and insulators, the Chern number of the valence band Bloch state is an important quantity that has been linked to various material properties, such as the topological order. We elaborate that the opacity of 2D materials to circularly polarized light over a wide range of frequencies, measured in units of the fine structure constant, can be used to extract a spectral function that frequency-integrates to the Chern number, offering a simple optical experiment to measure it. This method is subsequently generalized to finite temperature and locally on every lattice site by a linear response theory, which helps to extract the Chern marker that maps the Chern number to lattice sites. The long range response in our theory corresponds to a Chern correlator that acts like the internal fluctuation of the Chern marker, and is found to be enhanced in the topologically nontrivial phase. Finally, from the Fourier transform of the valence band Berry curvature, a nonlocal Chern marker is further introduced, whose decay length diverges at topological phase transitions and therefore serves as a faithful indicator of the transitions, and moreover can be interpreted as a Wannier state correlation function. The concepts discussed in this work explore multi-faceted aspects of topology and should help address the impact of system inhomogeneities.

  5 citations

• Nonlocal correlations in noisy multiqubit systems simulated using matrix product operators

  Haggai Landa, Grégoire Misguich

  SciPost Phys. Core 6, 037 (2023) · published 1 May 2023 | Toggle abstract · pdf

  We introduce an open-source solver for the Lindblad master equation, based on matrix product states and matrix product operators. Using this solver we study the dynamics of tens of interacting qubits with different connectivities, focusing on a problem where an edge qubit is being continuously driven on resonance, which is a fundamental operation in quantum devices. Because of the driving, induced excitations propagate through the qubits until the system reaches a steady state due to the incoherent terms. We find that with alternating-frequency qubits whose interactions with their off-resonant neighbors appear weak, the tunneling excitations lead to large correlations between distant qubits in the system. Some two-qubit correlation functions are found to increase as a function of distance in the system (in contrast to the typical decay with distance), peaking on the two edge qubits farthest apart from each other.

  5 citations

• Loop amplitudes from precision networks

  Simon Badger, Anja Butter, Michel Luchmann, Sebastian Pitz, Tilman Plehn

  SciPost Phys. Core 6, 034 (2023) · published 25 April 2023 | Toggle abstract · pdf

  Evaluating loop amplitudes is a time-consuming part of LHC event generation. For di-photon production with jets we show that simple, Bayesian networks can learn such amplitudes and model their uncertainties reliably. A boosted training of the Bayesian network further improves the uncertainty estimate and the network precision in critical phase space regions. In general, boosted network training of Bayesian networks allows us to move between fit-like and interpolation-like regimes of network training.

  5 citations

• Optically detected magnetic resonance with an open source platform

  Hossein Babashah, Hoda Shirzad, Elena Losero, Valentin Goblot, Christophe Galland, Mayeul Chipaux

  SciPost Phys. Core 6, 065 (2023) · published 9 October 2023 | Toggle abstract · pdf

  Localized electronic spins in solid-state environments form versatile and robust platforms for quantum sensing, metrology and quantum information processing. With optically detected magnetic resonance (ODMR), it is possible to prepare and readout highly coherent spin systems, up to room temperature, with orders of magnitude enhanced sensitivities and spatial resolutions compared to induction-based techniques, allowing for single spin manipulations. While ODMR was first observed in organic molecules, many other systems have since then been identified. Among them is the nitrogen-vacancy (NV) center in diamond, which is used both as a nanoscale quantum sensor for external fields and as a spin qubit. Other systems permitting ODMR are rare earth ions used as quantum memories and many other color centers trapped in bulk or 2-dimensional host materials. In order to allow the broadest possible community of researchers and engineers to investigate and develop novel ODMR-based materials and applications, we review here the setting up of ODMR experiments using commercially available hardware. We also present in detail the dedicated collaborative open-source interface named Qudi and describe the features we added to speed-up data acquisition, relax instrument requirements and extend its applicability to ensemble measurements. Covering both hardware and software development, this article aims to overview the setting of ODMR experiments and provide an efficient, portable and collaborative interface to implement innovative experiments to optimize the development time of ODMR experiments for scientists of any backgrounds.

  4 citations

• Equilibration of multitime quantum processes in finite time intervals

  Neil Dowling, Pedro Figueroa-Romero, Felix A. Pollock, Philipp Strasberg, Kavan Modi

  SciPost Phys. Core 6, 043 (2023) · published 14 June 2023 | Toggle abstract · pdf

  A generic non-integrable (unitary) out-of-equilibrium quantum process, when interrogated across many times, is shown to yield the same statistics as an (non-unitary) equilibrated process. In particular, using the tools of quantum stochastic processes, we prove that under loose assumptions, quantum processes equilibrate within finite time intervals. Sufficient conditions for this to occur are that multitime observables are coarse grained in both space and time, and that the initial state overlaps with many different energy eigenstates. These results help bridge the gap between (unitary) quantum and (non-unitary) statistical physics, i.e., when all multitime properties and correlations are well approximated by stationary quantities, which includes non-Markovianity and temporal entanglement. We discuss implications of this result for the emergence of classical stochastic processes from multitime measurements of an underlying genuinely quantum system.

  4 citations

• Dissipative dynamics in the free massive boson limit of the sine-Gordon model

  Á. Bácsi, C. P. Moca, G. Zaránd, B. Dóra

  SciPost Phys. Core 5, 004 (2022) · published 27 January 2022 | Toggle abstract · pdf

  We study the dissipative dynamics of one-dimensional fermions, described in terms of the sine-Gordon model in its free massive boson or semi-classical limit, while keeping track of forward scattering processes. The system is prepared in the gapped ground state, and then coupled to environment through local currents within the Lindblad formalism. The heating dynamics of the system is followed using bosonization. The single particle density matrix exhibits correlations between the left and right moving particles. While the density matrix of right movers and left movers is translationally invariant, the left-right sector is not, corresponding to a translational symmetry breaking charge density wave state. Asymptotically, the single particle density matrix decays exponentially with exponent proportional to $-\gamma t|x|\Delta^2$ where $\gamma$ and $\Delta$ are the dissipative coupling and the gap, respectively. The charge density wave order parameter decays exponentially in time with an interaction independent decay rate. The second R\'enyi entropy grows linearly with time and is essentially insensitive to the presence of the gap.

  4 citations

• Classical and quantum harmonic mean-field models coupled intensively and extensively with external baths

  Francesco Andreucci, Stefano Lepri, Stefano Ruffo, Andrea Trombettoni

  SciPost Phys. Core 5, 036 (2022) · published 15 July 2022 | Toggle abstract · pdf

  We study the nonequilibrium steady-state of a fully-coupled network of $N$ quantum harmonic oscillators, interacting with two thermal reservoirs. Given the long-range nature of the couplings, we consider two setups: one in which the number of particles coupled to the baths is fixed (intensive coupling) and one in which it is proportional to the size $N$ (extensive coupling). In both cases, we compute analytically the heat fluxes and the kinetic temperature distributions using the nonequilibrium Green's function approach, both in the classical and quantum regimes. In the large $N$ limit, we derive the asymptotic expressions of both quantities as a function of $N$ and the temperature difference between the baths. We discuss a peculiar feature of the model, namely that the bulk temperature vanishes in the thermodynamic limit, due to a decoupling of the dynamics of the inner part of the system from the baths. At variance with usual cases, this implies that the steady state depends on the initial state of the particles in the bulk. We also show that quantum effects are relevant only below a characteristic temperature that vanishes as $1/N$. In the quantum low-temperature regime the energy flux is proportional to the universal quantum of thermal conductance.

  4 citations