Featured Researches

Disordered Systems And Neural Networks

Emulating Quantum Interference with Generalized Ising Machines

The recent groundbreaking demonstration of quantum supremacy in the noisy intermediate scale quantum (NISQ) era has led to an intense activity in establishing finer boundaries between classical and quantum computing. In this paper, we use quantum Monte Carlo (QMC) techniques to formulate a systematic procedure for translating any sequence of d quantum gates acting on n q-bits into a Boltzmann machine (BM) having n+g(d) classical spins or p-bits with two values "0" and "1", but with a complex energy function E . Using this procedure we emulate Shor's algorithm with up to 36 q-bits using 90 p-bits, on an ordinary laptop computer in less than a day, while a naive Schrödinger implementation would require multiplying matrices with ≈ 10 21 elements. Even larger problems should be accessible on dedicated Ising Machines. However, we also identify clear limitations of the probabilistic approach by introducing a quantitative metric S Total for its inefficiency relative to a quantum computer. For example, a straightforward probabilistic implementation of Shor's algorithm with n q-bits leads to an S Total ∼exp(−n/2) , making the computation time for the probabilistic Shor's algorithm scale exponentially as 2 n/2 instead of the polynomial scaling expected for true quantum computers. This is a manifestation of the well-known sign problem in QMC and it may be possible to "tame" it with appropriate transformations. Finally, we present an example featuring a standard optimization algorithm based on a purely real energy function to which we add an imaginary part I(E) , thereby augmenting the statistical suppression of Feynman paths with quantum-like phase cancellation. This example illustrates how the sign problem encountered in classical annealers can potentially be turned into a computational resource for quantum annealers.

Read more
Disordered Systems And Neural Networks

Enhancing Scalability of a Matrix-Free Eigensolver for Studying Many-Body Localization

In [Van Beeumen, et. al, HPC Asia 2020, this https URL] a scalable and matrix-free eigensolver was proposed for studying the many-body localization (MBL) transition of two-level quantum spin chain models with nearest-neighbor XX+YY interactions plus Z terms. This type of problem is computationally challenging because the vector space dimension grows exponentially with the physical system size, and averaging over different configurations of the random disorder is needed to obtain relevant statistical behavior. For each eigenvalue problem, eigenvalues from different regions of the spectrum and their corresponding eigenvectors need to be computed. Traditionally, the interior eigenstates for a single eigenvalue problem are computed via the shift-and-invert Lanczos algorithm. Due to the extremely high memory footprint of the LU factorizations, this technique is not well suited for large number of spins L , e.g., one needs thousands of compute nodes on modern high performance computing infrastructures to go beyond L=24 . The matrix-free approach does not suffer from this memory bottleneck, however, its scalability is limited by a computation and communication imbalance. We present a few strategies to reduce this imbalance and to significantly enhance the scalability of the matrix-free eigensolver. To optimize the communication performance, we leverage the consistent space runtime, CSPACER, and show its efficiency in accelerating the MBL irregular communication patterns at scale compared to optimized MPI non-blocking two-sided and one-sided RMA implementation variants. The efficiency and effectiveness of the proposed algorithm is demonstrated by computing eigenstates on a massively parallel many-core high performance computer.

Read more
Disordered Systems And Neural Networks

Ensemble perspective for understanding temporal credit assignment

Recurrent neural networks are widely used for modeling spatio-temporal sequences in both nature language processing and neural population dynamics. However, understanding the temporal credit assignment is hard. Here, we propose that each individual connection in the recurrent computation is modeled by a spike and slab distribution, rather than a precise weight value. We then derive the mean-field algorithm to train the network at the ensemble level. The method is then applied to classify handwritten digits when pixels are read in sequence, and to the multisensory integration task that is a fundamental cognitive function of animals. Our model reveals important connections that determine the overall performance of the network. The model also shows how spatio-temporal information is processed through the hyperparameters of the distribution, and moreover reveals distinct types of emergent neural selectivity. It is thus promising to study the temporal credit assignment in recurrent neural networks from the ensemble perspective.

Read more
Disordered Systems And Neural Networks

Entanglement Clustering for ground-stateable quantum many-body states

Despite their fundamental importance in dictating the quantum mechanical properties of a system, ground states of many-body local quantum Hamiltonians form a set of measure zero in the many-body Hilbert space. Hence determining whether a given many-body quantum state is ground-stateable is a challenging task. Here we propose an unsupervised machine learning approach, dubbed the Entanglement Clustering ("EntanCl"), to separate out ground-stateable wave functions from those that must be excited state wave functions using entanglement structure information. EntanCl uses snapshots of an ensemble of swap operators as input and projects this high dimensional data to two-dimensions, preserving important topological features of the data associated with distinct entanglement structure using the uniform manifold approximation and projection (UMAP). The projected data is then clustered using K-means clustering with k=2 . By applying EntanCl to two examples, a one-dimensional band insulator and the two-dimensional toric code, we demonstrate that EntanCl can successfully separate ground states from excited states with high computational efficiency. Being independent of a Hamiltonian and associated energy estimates, EntanCl offers a new paradigm for addressing quantum many-body wave functions in a computationally efficient manner.

Read more
Disordered Systems And Neural Networks

Entanglement Properties of Disordered Quantum Spin Chains with Long-Range Antiferromagnetic Interactions

We examine the concurrence and entanglement entropy in quantum spin chains with random long-range couplings, spatially decaying with a power-law exponent α . Using the strong disorder renormalization group (SDRG) technique, we find by analytical solution of the master equation a strong disorder fixed point, characterized by a fixed point distribution of the couplings with a finite dynamical exponent, which describes the system consistently in the regime α>1/2 . A numerical implementation of the SDRG method yields a power law spatial decay of the average concurrence, which is also confirmed by exact numerical diagonalization. However, we find that the lowest-order SDRG approach is not sufficient to obtain the typical value of the concurrence. We therefore implement a correction scheme which allows us to obtain the leading order corrections to the random singlet state. This approach yields a power-law spatial decay of the typical value of the concurrence, which we derive both by a numerical implementation of the corrections and by analytics. Next, using numerical SDRG, the entanglement entropy (EE) is found to be logarithmically enhanced for all α , corresponding to a critical behavior with an effective central charge c=ln2 , independent of α . This is confirmed by an analytical derivation. Using numerical exact diagonalization (ED), we confirm the logarithmic enhancement of the EE and a weak dependence on α . For a wide range of distances l , the EE fits a critical behavior with a central charge close to c=1 , which is the same as for the clean Haldane-Shastry model with a power-la-decaying interaction with α=2 . Consistent with this observation, we find using ED that the concurrence shows power law decay, albeit with smaller power exponents than obtained by SDRG.

Read more
Disordered Systems And Neural Networks

Entanglement Transition in the Projective Transverse Field Ising Model

Discrete quantum trajectories of systems under random unitary gates and projective measurements have been shown to feature transitions in the entanglement scaling that are not encoded in the density matrix. In this paper, we study the projective transverse field Ising model, a stochastic model with two noncommuting projective measurements and no unitary dynamics. We numerically demonstrate that their competition drives an entanglement transition between two distinct steady states that both exhibit area law entanglement, and introduce a classical but nonlocal model that captures the entanglement dynamics completely. Exploiting a map to bond percolation, we argue that the critical system in one dimension is described by a conformal field theory, and derive the universal scaling of the entanglement entropy and the critical exponent for the scaling of the mutual information of two spins exactly. We conclude with an interpretation of the entanglement transition in the context of quantum error correction.

Read more
Disordered Systems And Neural Networks

Entanglement scaling in fermion chains with a localization-delocalization transition and inhomogeneous modulations

We study the scaling of logarithmic negativity between adjacent subsystems in critical fermion chains with various inhomogeneous modulations through numerically calculating its recently established lower and upper bounds. For random couplings, as well as for a relevant aperiodic modulation of the couplings, which induces an aperiodic singlet state, the bounds are found to increase logarithmically with the subsystem size, and both prefactors agree with the predicted values characterizing the corresponding asymptotic singlet state. For the marginal Fibonacci modulation, the prefactors in front of the logarithm are different for the lower and the upper bound, and vary smoothly with the strength of the modulation. In the delocalized phase of the quasi-periodic Harper model, the scaling of the bounds of the logarithmic negativity as well as that of the entanglement entropy are compatible with the logarithmic scaling of the homogeneous chain. At the localization transition, the scaling of the above entanglement characteristics holds to be logarithmic, but the prefactors are significantly reduced compared to those of the translationally invariant case, roughly by the same factor.

Read more
Disordered Systems And Neural Networks

Entropic barriers as a reason for hardness in both classical and quantum algorithms

We study both classical and quantum algorithms to solve a hard optimization problem, namely 3-XORSAT on 3-regular random graphs. By introducing a new quasi-greedy algorithm that is not allowed to jump over large energy barriers, we show that the problem hardness is mainly due to entropic barriers. We study, both analytically and numerically, several optimization algorithms, finding that entropic barriers affect in a similar way classical local algorithms and quantum annealing. For the adiabatic algorithm, the difficulty we identify is distinct from that of tunnelling under large barriers, but does, nonetheless, give rise to exponential running (annealing) times.

Read more
Disordered Systems And Neural Networks

Entropy, Free Energy, and Work of Restricted Boltzmann Machines

A restricted Boltzmann machine is a generative probabilistic graphic network. A probability of finding the network in a certain configuration is given by the Boltzmann distribution. Given training data, its learning is done by optimizing parameters of the energy function of the network. In this paper, we analyze the training process of the restricted Boltzmann machine in the context of statistical physics. As an illustration, for small size Bar-and-Stripe patterns, we calculate thermodynamic quantities such as entropy, free energy, and internal energy as a function of training epoch. We demonstrate the growth of the correlation between the visible and hidden layers via the subadditivity of entropies as the training proceeds. Using the Monte-Carlo simulation of trajectories of the visible and hidden vectors in configuration space, we also calculate the distribution of the work done on the restricted Boltzmann machine by switching the parameters of the energy function. We discuss the Jarzynski equality which connects the path average of the exponential function of the work and the difference in free energies before and after training.

Read more
Disordered Systems And Neural Networks

Equilibration time in many-body quantum systems

Isolated many-body quantum systems quenched far from equilibrium can eventually equilibrate, but it is not yet clear how long they take to do so. To answer this question, we use exact numerical methods and analyze the entire evolution, from perturbation to equilibration, of a paradigmatic disordered many-body quantum system in the chaotic regime. We investigate how the equilibration time depends on the system size and observables. We show that if dynamical manifestations of spectral correlations in the form of the correlation hole ("ramp") are taken into account, the time for equilibration scales exponentially with system size, while if they are neglected, the scaling is better described by a power law with system size, though with an exponent larger than what is expected for diffusive transport.

Read more

Ready to get started?

Join us today