Stochastic Variational Method Research Articles

Uncertainty relation in viscous hydrodynamics and its effects in collective flow observables

AbstractThe uncertainty relation is considered to be one of key features of quantum mechanics which distinguishes quantum and classical systems. Recently, we developed a new formulation of the uncertainty relation based on the generalized scheme of variational principle, the stochastic variational method (SVM). In this method, the uncertainty relation is related to the nondifferentiability of observables and thus can be obtained even in classical stochastic systems. This new formulation resolves the famous paradox in quantum mechanics, the angular uncertainty relation without introducing artificial assumptions. In this paper, we show that the fluctuations of position and momentum for a nonrelativistic viscous fluid element satisfies the uncertainty relation analogous to the corresponding quantum mechanical one. Such a fluctuation is sensitive to the temperature gradient at the freezeout surface and can affect the collective flow anisotropy in relativistic heavy‐ion collisions.

Calculation of Dynamical Response Functions Using a Bound-State Method

We investigate a method to extract response functions (dynamical polarisabilities) directly from a bound-state approach applied to calculations of perturbation-induced reactions. The use of a square-integrable basis leads to a response in the form of a sum of delta functions. We integrate this over energy and fit a smooth function to the resulting stepwise-continuous one. Its derivative gives the final approximation to the physical response function. We show that the method reproduces analytical results where known, and analyse the details for a variety of models. We apply it to some simple models, using the stochastic variational method as the numerical method. Albeit we find that this approach, and other numerical techniques, have some difficulties with the threshold behavior in coupled-channel problems with multiple thresholds, its stochastic nature allows us to extract robust results even for such cases.

Impurity-induced pairing in two-dimensional Fermi gases

We study induced pairing between two identical fermions mediated by an attractively interacting quantum impurity in two-dimensional systems. Based on a stochastic variational method (SVM), we investigate the influence of confinement and finite interaction range effects on the mass ratio beyond which the ground state of the quantum three-body problem undergoes a transition from a composite bosonic trimer to an unbound dimer-fermion state. We find that confinement as well as a finite interaction range can greatly enhance trimer stability, bringing it within reach of experimental implementations such as found in ultracold atom systems. In the context of solid-state physics, our solution of the confined three-body problem shows that exciton-mediated interactions can become so dominant that they can even overcome detrimental Coulomb repulsion between electrons in atomically-thin semiconductors. Our paper thus paves the way towards a universal understanding of boson-induced pairing across various fermionic systems at finite density, and opens perspectives towards realizing unexplored forms of electron pairing beyond the conventional paradigm of Cooper pair formation.

Three-$$\alpha $$ configurations of the second $$J^\pi =2^+$$ state in $$^{12}$$C

We investigate geometric configurations of $\alpha$ ($^4$He nucleus) clusters in the second $J^\pi=2^+$ state of $^{12}$C, which has been discussed as a rotational band member of the second $0^+$ state, the Hoyle state. The ground and excited $0^+$ and $2^+$ states are described by a three-$\alpha$ cluster model. The three-body Schr\"odinger equation with orthogonality conditions is accurately solved by the stochastic variational method with correlated Gaussian basis functions. To analyse the structure of these resonant states in a convenient form, we introduce a confining potential. The two-body density distributions together with the spectroscopic information clarify the structure of these states. We find that main configurations of both the second $0^+$ and $2^+$ states are acute-angled triangle shapes originating from the $^8$Be($0^+$)$+\alpha$ configuration. However, the $^8$Be$+\alpha$ components in the second $2^+$ state become approximately 2/3 because the 8Be subsystem is hard to excite, indicating that the state is not an ideal rigid rotational band member of the Hoyle state.

Deep kernel learning of dynamical models from high-dimensional noisy data

This work proposes a stochastic variational deep kernel learning method for the data-driven discovery of low-dimensional dynamical models from high-dimensional noisy data. The framework is composed of an encoder that compresses high-dimensional measurements into low-dimensional state variables, and a latent dynamical model for the state variables that predicts the system evolution over time. The training of the proposed model is carried out in an unsupervised manner, i.e., not relying on labeled data. Our learning method is evaluated on the motion of a pendulum—a well studied baseline for nonlinear model identification and control with continuous states and control inputs—measured via high-dimensional noisy RGB images. Results show that the method can effectively denoise measurements, learn compact state representations and latent dynamical models, as well as identify and quantify modeling uncertainties.

Stochastic Variational Method for Viscous Hydrodynamics

In this short review, we focus on some of the subjects, related to J. Cleymans’ pioneering contribution of statistical approaches to the particle production process in heavy-ion collisions. We discuss these perspectives from the effects of stochastic processes in collective variables of hydrodynamic description, which is described by a stochastic variational method. In this connection, we stress also the necessity of the inclusion of surface and quantum effects in the study of relativistic heavy-ion reactions.

Spectrum of light nuclei in a finite volume

Lattice quantum chromodynamics calculations of multi-baryon systems with physical quark masses would start a new age of ab initio predictions in nuclear physics. Performed on a finite grid, such calculations demand extrapolation of their finite volume numerical results to free-space physical quantities. Such extraction of the physical information can be carried out fitting effective field theories (EFTs) directly to the finite-volume results or utilizing the L\"uscher free-space formula or its generalizations for extrapolating the lattice data to infinite volume. To understand better the effect of periodic boundary conditions on the binding energy of few nucleon systems we explore here light nuclei with physical masses in a finite box and in free space. The stochastic variational method is used to solve the few-body systems. Substantial optimizations of the method are introduced to enable efficient calculations in a periodic box. With the optimized code, we perform accurate calculations of light nuclei $A \le 4$ within leading order pionless EFT. Using L\"uscher formula for the two-body system, and its generalization for 3- and 4-body systems, we examine the box effect and explore possible limitations of these formulas for the considered nuclear systems.

Consequences of increased hypertriton binding for s -shell Λ -hypernuclear systems

Consequences of increasing the binding energy of the hypertriton ground state ${_{\Lambda}^3}{\rm H}(J^P={\frac{1}{2}}^+)$ from the emulsion value $B^{\rm EMUL}_{\Lambda}({_{\Lambda}^3}{\rm H}_{\rm g.s.})$=0.13$\pm$0.05 MeV to the STAR value $B^{\rm STAR}_{\Lambda}(^{3}_{\Lambda}{\rm H}) = (0.41\pm 0.12 \pm 0.11)$ MeV are studied for $s$-shell hypernuclei within a pionless EFT approach at leading order, constrained by the binding energies of the $0^+$ and $1^+$ ${_{\Lambda}^4} {\rm H}$ states. The stochastic variational method is used in bound-state calculations, whereas the inverse analytic continuation in the coupling constant method is used to locate $S$-matrix poles of continuum states. It is found that the $\Lambda nn({\frac{1}{2}}^+)$ resonance becomes broader and less likely to be observed experimentally, whereas the ${_{\Lambda}^3}{\rm H}({\frac{3}{2}}^+)$ spin-flip virtual state moves closer to the $\Lambda d$ threshold to become a shallow bound state for specific $\Lambda N$ interaction strengths. The effect of such a near-threshold ${_{\Lambda}^3}{\rm H}({\frac{3}{2}}^+)$ state on femtoscopic studies of $\Lambda$-deuteron correlations, and its lifetime if bound, are discussed. Increasing $B_{\Lambda}({_{\Lambda}^3} {\rm H}_{\rm g.s.})$ moderately, up to $\sim$0.5 MeV, hardly affects calculated values of $B_{\Lambda}({_{\Lambda}^5}{\rm He})$.

Strongly coupled matrix theory and stochastic quantization: A new approach to holographic dualities

Stochastic quantization provides an alternate approach to the computation of quantum observables, by stochastically sampling phase space in a path integral. Furthermore, the stochastic variational method can provide analytical control over the strong coupling regime of a quantum field theory -- provided one has a decent qualitative guess at the form of certain observables at strong coupling. In the context of the holographic duality, the strong coupling regime of a Yang-Mills theory can capture gravitational dynamics. This can provide enough insight to guide a stochastic variational ansatz. We demonstrate this in the bosonic Banks-Fischler-Shenker-Susskind Matrix theory. We compute a two-point function at all values of coupling using the variational method showing agreement with lattice numerical computations and capturing the confinement-deconfinement phase transition at strong coupling. This opens up a new realm of possibilities for exploring the holographic duality and emergent geometry.

Spectrum of few-body systems in a finite volume

Lattice quantum chromodynamics calculations of multi-baryon systems with physical quark masses could bridge the gap between nuclear physics and its fundamental theory. Such calculations are performed on a finite grid, and thus extrapolation of finite-volume numerical results to the physical-relevant free-space quantities is needed. A common approach for the extrapolation utilizes the two-body Lüscher formula and its generalization to larger systems. To understand better the accuracy of this approach we study here two model systems, bosonic and fermionic. The few-body systems are solved with the stochastic variational method in a periodic box, and the extrapolated values are compared to those calculated directly in free space.

Infinite Switching Dynamic Probabilistic Network With Bayesian Nonparametric Learning

To model sequentially observed multivariate nonstationary count data, we propose a switching Poisson-gamma dynamical systems (SPGDS), a dynamic probabilistic network with switching mechanism. Different from previous models, SPGDS assigns its latent variables into mixture of gamma distributed parameters to model complex sequences and describe the nonlinear dynamics, meanwhile, capture various temporal dependencies. Moreover, SPGDS can model all discrete and nonnegative real data by linking them to latent counts. To take advantage of Bayesian nonparametrics in handling the unknown number of mixture components, we integrate Dirichlet process (DP) mixture into SPGDS and develop an infinite switching Poisson-gamma dynamical systems (iSPGDS). For efficient and nonparametric inference, we develop a infinite switching recurrent variational inference network, combined with a scalable hybrid stochastic gradient-MCMC and variational inference method, which is scalable to large scale sequences and fast in out-of-sample prediction. Besides, to handle the time-series categorization task, we further propose an supervised attention iSPGDS (attn-iSPGDS), which combines the representation power of iSPGDS, discriminative power of deep neural networks, and selection power of the attention mechanism under a principled probabilistic framework. Experiments on both unsupervised and supervised tasks demonstrate that the proposed model not only has excellent fitting and prediction performance on complex sequences, but also separates different dynamical patterns within them.

Tracking Disease Outbreaks from Sparse Data with Bayesian Inference

The COVID-19 pandemic provides new motivation for a classic problem in epidemiology: estimating the empirical rate of transmission during an outbreak (formally, the time-varying reproduction number) from case counts. While standard methods exist, they work best at coarse-grained national or state scales with abundant data, and struggle to accommodate the partial observability and sparse data common at finer scales (e.g., individual schools or towns). For example, case counts may be sparse when only a small fraction of infections are caught by a testing program. Or, whether an infected individual tests positive may depend on the kind of test and the point in time when they are tested. We propose a Bayesian framework which accommodates partial observability in a principled manner. Our model places a Gaussian process prior over the unknown reproduction number at each time step and models observations sampled from the distribution of a specific testing program. For example, our framework can accommodate a variety of kinds of tests (viral RNA, antibody, antigen, etc.) and sampling schemes (e.g., longitudinal or cross-sectional screening). Inference in this framework is complicated by the presence of tens or hundreds of thousands of discrete latent variables. To address this challenge, we propose an efficient stochastic variational inference method which relies on a novel gradient estimator for the variational objective. Experimental results for an example motivated by COVID-19 show that our method produces an accurate and well-calibrated posterior, while standard methods for estimating the reproduction number can fail badly.

Few-nucleon matrix elements in pionless effective field theory in a finite volume

Pionless effective field theory in a finite volume (FVEFT$_{\pi\!/}$) is investigated as a framework for the analysis of multi-nucleon spectra and matrix elements calculated in lattice QCD (LQCD). By combining FVEFT$_{\pi\!/}$ with the stochastic variational method, the spectra of nuclei with atomic number $A\in\{2,3\}$ are matched to existing finite-volume LQCD calculations at heavier-than-physical quark masses corresponding to a pion mass $m_\pi=806$ MeV, thereby enabling infinite-volume binding energies to be determined using infinite-volume variational calculations. Based on the variational wavefunctions that are constructed in this approach, the finite-volume matrix elements of various local operators are computed in FVEFT$_{\pi\!/}$ and matched to LQCD calculations of the corresponding QCD operators in the same volume, thereby determining the relevant one and two-body EFT counterterms and enabling an extrapolation of the LQCD matrix elements to infinite volume. As examples, the scalar, tensor, and axial matrix elements are considered, as well as the magnetic moments and the isovector longitudinal momentum fraction.

Nature of the Λnn(Jπ=1/2+, I=1) and Λ3H*(Jπ=3/2+, I=0) states

The nature of the $\Lambda nn$ and ${\rm ^3_\Lambda H^*} (J^\pi=3/2^+,~I=0)$ states is investigated within a pionless effective field theory at leading order, constrained by the low energy $\Lambda N$ scattering data and hypernuclear 3- and 4-body data. Bound state solutions are obtained using the stochastic variational method, the continuum region is studied by employing two independent methods - the inverse analytic continuation in the coupling constant method and the complex scaling method. Our calculations yield both the $\Lambda nn$ and ${\rm ^3_\Lambda H^*}$ states unbound. We conclude that the excited state ${\rm ^3_\Lambda H^*}$ is a virtual state and the $\Lambda nn$ pole located close to the three-body threshold in a complex energy plane could convert to a true resonance with Re$(E)>0$ for some considered $\Lambda N$ interactions. Finally, the stability of resonance solutions is discussed and limits of the accuracy of performed calculations are assessed.

Uncertainty Relations in Hydrodynamics

The qualitative behaviors of uncertainty relations in hydrodynamics are numerically studied for fluids with low Reynolds numbers in 1+1 dimensional system. We first give a review for the formulation of the generalized uncertainty relations in the stochastic variational method (SVM), following the work by two of the present authors [Phys. Lett. A 382, 1472 (2018)]. In this approach, the origin of the finite minimum value of uncertainty is attributed to the non-differentiable (virtual) trajectory of a quantum particle and then both of the Kennard and Robertson-Schrödinger inequalities in quantum mechanics are reproduced. The same non-differentiable trajectory is applied to the motion of fluid elements in the Navier-Stokes-Fourier equation or the Navier-Stokes-Korteweg equation. By introducing the standard deviations of position and momentum for fluid elements, the uncertainty relations in hydrodynamics are derived. These are applicable even to the Gross-Pitaevskii equation and then the field-theoretical uncertainty relation is reproduced. We further investigate numerically the derived relations and find that the behaviors of the uncertainty relations for liquid and gas are qualitatively different. This suggests that the uncertainty relations in hydrodynamics are used as a criterion to classify liquid and gas in fluid.

Extrapolating lattice QCD results using effective field theory

Lattice simulations are the only viable way to obtain ab-initio Quantum Chromodynamics (QCD) predictions for low energy nuclear physics. These calculations are done, however, in a finite box and therefore extrapolation is needed to get the free space results. Here we use nuclear Effective Field Theory (EFT), designed to provide a low energy description of QCD using baryonic degrees of freedom, to extrapolate the lattice results from finite to infinite volumes. To this end, we fit the EFT to the results calculated with nonphysical high quark masses and solve it with the stochastic variational method in both finite and infinite volumes. Moreover, we perform similar EFT calculations of the physical point and predict the finite-volume effects to be found in future Lattice QCD calculations for atomic nuclei with mass number $A\le4$.

Coulomb screening correction to the Q value of the triple-alpha process in thermal plasmas

Abstract The triple-alpha reaction is key to $^{12}$C production and is expected to occur in weakly coupled thermal plasmas as encountered in normal stars. We investigate how Coulomb screening affects the structure of a system of three alpha particles in such a plasma environment by precise three-body calculations within the Debye–Hückel approximation. A three-alpha model that has the Coulomb interaction modified in the Yukawa form is employed. Precise three-body wave functions are obtained by a superposition of correlated Gaussian basis with the aid of the stochastic variational method. The energy shifts of the Hoyle state due to the Coulomb screening are obtained as a function of the Debye screening length. The results, which automatically incorporate the finite-size effect of the Hoyle state, are consistent with the conventional result based on the Coulomb correction to the chemical potentials of ions that are regarded as point charges in a weakly coupled thermal plasma. We have given a theoretical basis to the conventional point-charge approach to the Coulomb screening problem relevant for nuclear reactions in normal stars by providing the first evaluation of the Coulomb corrections to the $Q$ value of the triple-alpha process that produces a finite-size Hoyle state.

An Efficient Image Categorization Method With Insufficient Training Samples.

Image classification is an important part of pattern recognition. With the development of convolutional neural networks (CNNs), many CNN methods are proposed, which have a large number of samples for training, which can have high performance. However, there may exist limited samples in some real-world applications. In order to improve the performance of CNN learning with insufficient samples, this article proposes a new method called the classifier method based on a variational autoencoder (CFVAE), which is comprised of two parts: 1) a standard CNN as a prior classifier and 2) a CNN based on variational autoencoder (VAE) as a posterior classifier. First, the prior classifier is utilized to generate the prior label and information about distributions of latent variables; and the posterior classifier is trained to augment some latent variables from regularized distributions to improve the performance. Second, we also present the uniform objective function of CFVAE and put forward an optimization method based on the stochastic gradient variational Bayes method to solve the objective model. Third, we analyze the feasibility of CFVAE based on Hoeffding's inequality and Chernoff's bounding method. This analysis indicates that the latent variables augmentation method based on regularized latent variables distributions can generate samples fitting well with the distribution of data such that the proposed method can improve the performance of CNN with insufficient samples. Finally, the experiments manifest that our proposed CFVAE can provide more accurate performance than state-of-the-art methods.

Variational formulation of compressible hydrodynamics in curved spacetime and symmetry of stress tensor

Hydrodynamics of the non-relativistic compressible fluid in the curved spacetime is derived using the generalized framework of the stochastic variational method (SVM) for continuum medium. The fluid-stress tensor of the resultant equation becomes asymmetric for the exchange of the indices, different from the standard Euclidean one. Its incompressible limit suggests that the viscous term should be represented with the Bochner Laplacian. Moreover the modified Navier–Stokes–Fourier equation proposed by Brenner can be considered even in the curved spacetime. To confirm the compatibility with the symmetry principle, SVM is applied to the gauge-invariant Lagrangian of a charged compressible fluid and then the Lorentz force is reproduced as the interaction between the Abelian gauge fields and the viscous charged fluid.

Variational Metric Scaling for Metric-Based Meta-Learning

Metric-based meta-learning has attracted a lot of attention due to its effectiveness and efficiency in few-shot learning. Recent studies show that metric scaling plays a crucial role in the performance of metric-based meta-learning algorithms. However, there still lacks a principled method for learning the metric scaling parameter automatically. In this paper, we recast metric-based meta-learning from a Bayesian perspective and develop a variational metric scaling framework for learning a proper metric scaling parameter. Firstly, we propose a stochastic variational method to learn a single global scaling parameter. To better fit the embedding space to a given data distribution, we extend our method to learn a dimensional scaling vector to transform the embedding space. Furthermore, to learn task-specific embeddings, we generate task-dependent dimensional scaling vectors with amortized variational inference. Our method is end-to-end without any pre-training and can be used as a simple plug-and-play module for existing metric-based meta-algorithms. Experiments on miniImageNet show that our methods can be used to consistently improve the performance of existing metric-based meta-algorithms including prototypical networks and TADAM.