Class Of Statistical Problems Research Articles

Magic in generalized Rokhsar-Kivelson wavefunctions

Magic is a property of a quantum state that characterizes its deviation from a stabilizer state, serving as a useful resource for achieving universal quantum computation e.g., within schemes that use Clifford operations. In this work, we study magic, as quantified by the stabilizer Renyi entropy, in a class of models known as generalized Rokhsar-Kivelson systems, i.e., Hamiltonians that allow a stochastic matrix form (SMF) decomposition. The ground state wavefunctions of these systems can be written explicitly throughout their phase diagram, and their properties can be related to associated classical statistical mechanics problems, thereby allowing powerful analytical and numerical approaches that are not usually available in conventional quantum many body settings. As a result, we are able to express the SRE in terms of wave function coefficients that can be understood as a free energy difference of related classical problems. We apply this insight to a range of quantum many body SMF Hamiltonians, which affords us to study numerically the SRE of large high-dimensional systems, and in some cases to obtain analytical results. We observe that the behaviour of the SRE is relatively featureless across quantum phase transitions in these systems, although it is indeed singular (in its first or higher order derivative, depending on the nature of the transition). On the contrary, we find that the maximum of the SRE generically occurs at a cusp away from the quantum critical point, where the derivative suddenly changes sign. Furthermore, we compare the SRE and the logarithm of overlaps with specific stabilizer states, asymptotically realised in the ground state phase diagrams of these systems. We find that they display strikingly similar behaviors, which in turn establish rigorous bounds on the min-relative entropy of magic.

A Data-Driven Framework for Power System Event Type Identification via Safe Semi-Supervised Techniques

This paper investigates the use of phasor measurement unit (PMU) data with deep learning techniques to construct real-time event identification models for transmission networks. Increasing penetration of distributed energy resources represents a great opportunity to achieve decarbonization, as well as challenges in systematic situational awareness. When high-resolution PMU data and sufficient manually recorded event labels are available, the power event identification problem is defined as a statistical classification problem that can be solved by numerous cutting-edge classifiers. However, in real grids, collecting tremendous high-quality event labels is quite expensive. Utilities frequently have a large number of event records without in-depth details (i.e., unlabeled events). To bridge this gap, we propose a novel semi-supervised learning-based method to improve the performance of event classifiers trained with a limited number of labeled events by exploiting the information from massive unlabeled events. In other words, compared to existing data-driven methods, our method requires only a small portion of labeled data to achieve a similar level of accuracy. Meanwhile, this work discusses and addresses the performance degradation caused by class distribution mismatch between the training set and the real applications. Specifically, this method utilizes pseudo-labeling technique to investigate the value of unlabeled events and incrementally expands the training dataset. Moreover, a safe learning mechanism is developed to mitigate the impacts of class distribution mismatch and prevent performance degradation. Based on the proposed safe learning mechanism, our model does not directly use all unlabeled events during model training, but selectively uses them through a comprehensive evaluation procedure. Numerical studies on a sizable PMU dataset have been used to validate the performance of the proposed method.

Information-Theoretic Approach to the Problem of Automatic Image Recognition

The problem of automatic half-tone image recognition from the large database is formulated. Its decision using data on a problem of statistical classification of discrete objects on the basis of the minimum information discrimination principle is proposed. The new probabilistic model of a half-tone image is studied. On the basis of this model, a new image recognition algorithm using the minimum Kullback–Leibler information discrimination criterion is synthesized. It is proved that this criterion is optimal in Bayesian terms for the classification problem of discrete random objects, including digital images. The program and results of experimental research on the face recognition problem are presented. It is shown that the accuracy of image recognition with various light conditions on the basis of the offered criterion is essentially higher than the accuracy of traditional methods.

Pointwise comparison of two multivariate density functions

AbstractTesting the equality of two density functions based on independent samples is a classical problem in statistics. While the focus is often on global equality, it is also of interest to conduct local comparisons of density functions. Typically a type of Wald statistic is employed, where the local difference in densities is standardized by an estimate of the asymptotic standard error of that difference. We study the null distribution of this test statistic. The literature has suggested that this will be asymptotically standard normal, but we show that this is by no means always the case. In particular, when using bandwidth matrices of optimal order (for estimation), we prove that the asymptotic mean of this null distribution is nonzero when either the sample sizes differ, or when the Hessian matrices of the densities differ at the point where the densities are equal. In numerical studies we find the erroneous use of the standard normal null distribution in such cases can severely corrupt the test size. We show that these problems can be managed effectively by using common bandwidths when the Hessian matrices are equal, and applying adjusted undersmoothing bandwidth matrices when they are not.

Hilbert space fragmentation produces an effective attraction between fractons

Fracton systems exhibit restricted mobility of their excitations due to the presence of higher-order conservation laws. Here we study the time evolution of a one-dimensional fracton system with charge and dipole moment conservation using a random unitary circuit description. Previous work has shown that when the random unitary operators act on four or more sites, an arbitrary initial state eventually thermalizes via a universal subdiffusive dynamics. In contrast, a system evolving under three-site gates fails to thermalize due to strong ``fragmentation'' of the Hilbert space. Here we show that three-site gate dynamics causes a given initial state to evolve toward a highly nonthermal state on a timescale consistent with Brownian diffusion. Strikingly, the dynamics produces an effective attraction between isolated fractons or between a single fracton and the boundaries of the system, as in the Casimir effect of quantum electrodynamics. We show how this attraction can be understood by exact mapping to a simple classical statistical mechanics problem, which we solve exactly for the case of an initial state with either one or two fractons.

Probabilistic machine learning for breast cancer classification.

A probabilistic neural network has been implemented to predict the malignancy of breast cancer cells, based on a data set, the features of which are used for the formulation and training of a model for a binary classification problem. The focus is placed on considerations when building the model, in order to achieve not only accuracy but also a safe quantification of the expected uncertainty of the calculated network parameters and the medical prognosis. The source code is included to make the results reproducible, also in accordance with the latest trending in machine learning research, named Papers with Code. The various steps taken for the code development are introduced in detail but also the results are visually displayed and critically analyzed also in the sense of explainable artificial intelligence. In statistical-classification problems, the decision boundary is the region of the problem space in which the classification label of the classifier is ambiguous. Problem aspects and model parameters which influence the decision boundary are a special aspect of practical investigation considered in this work. Classification results issued by technically transparent machine learning software can inspire more confidence, as regards their trustworthiness which is very important, especially in the case of medical prognosis. Furthermore, transparency allows the user to adapt models and learning processes to the specific needs of a problem and has a boosting influence on the development of new methods in relevant machine learning fields (transfer learning).

Relative Inference for Paired Data: More than Meets the Eye

Relative Inference for Paired Data: More than Meets the Eye

Scalable kernel two-sample tests via empirical likelihood and jackknife

Two-sample hypothesis testing is a classical problem in statistics. Given i.i.d. samples and the two-sample testing is to test whether P is significantly different from Q or not. Among the various statistical tests, the maximum mean discrepancy (MMD) has been proven to have excellent performance compared to traditional methods. In application of MMD, we are confronted with two challenges: (I) For large data set, MMD is computationally demanding with computational cost (II) The limiting distribution under the null hypothesis usually contains infinitely many unknown parameters. In this paper, we propose two testing methods that solve (I) and (II) respectively. To address (I), the data are independently and evenly divided into k groups, calculate the MMD on each group and then combine them by empirical likelihood to yield a test statistic It converges in distribution to the chi-square distribution with degree of freedom one ( ) under the null hypothesis as k goes to infinity. To address (II), we propose to compute the two-sample jackknife pseudo-values on each group and combine them by empirical likelihood to get a new test statistic Under the null hypothesis, converges in distribution to if The simulation and real data application indicates that the proposed tests performs well.

Entanglement of local operators and the butterfly effect

We study the robustness of quantum and classical information to perturbations implemented by local operator insertions. We do this by computing multipartite entanglement measures in the Hilbert space of local operators in the Heisenberg picture. The sensitivity to initial conditions that we explore is an illuminating manifestation of the butterfly effect in quantum many-body systems. We present a ``membrane theory'' in Haar random unitary circuits to compute the mutual information, logarithmic negativity, and reflected entropy in the local operator state by mapping to a classical statistical mechanics problem and find that any local operator insertion delocalizes information as fast as is allowed by causality after taking the large local Hilbert space dimension limit. Identical behavior is found for conformal field theories admitting holographic duals where the bulk geometry is described by the eternal black hole with a local object situated at the horizon. In contrast to these maximal scramblers, only an $O(1)$ amount of information is found to be delocalized by local operators in free fermionic systems and random Clifford circuits.

Change point estimation under a copula-based Markov chain model for binomial time series

Estimation of a change point is a classical statistical problem in sequential analysis and process control. For binomial time series, the existing maximum likelihood estimators (MLEs) for a change point are limited to independent observations. If the independence assumption is violated, the MLEs substantially lose their efficiency, and a likelihood function provides a poor fit to the data. A novel change point estimator is proposed under a copula-based Markov chain model for serially dependent observations. The main novelty is the adaptation of a three-state copula model, consisting of the in-control state, out-of-control state, and transition state. Under this model, a MLE is proposed with the aid of profile likelihood. A parametric bootstrap method is adopted to compute a confidence set for the unknown change point. The simulation studies show that the proposed MLE is more efficient than the existing estimators when serial dependence in observations are specified by the model. The proposed method is illustrated by the jewelry manufacturing data, where the proposed model gives an improved fit.

On parameters for inner products under imprecision and their applications to certain geometric and statistical problems

This work aims to study the imprecision on the inner product of a real vector space. It starts defining the notion of imprecise inner product. This notion can also model the uncertainty about the variability of a multivariate random variable. The primary goal is to introduce parameters in order to provide information about the shape and the size of this kind of inner products. Imprecision is inherent to any quantity which depends on the inner product (or the covariance matrix). Useful techniques are developed by means of the introduced parameters to obtain fast and easy approximations for that kind of quantities. Throughout several examples, the convenience of the techniques is proven. These techniques can be applied to a wider class of statistical problems dealing with imprecision.

Estimating location parameters in sample-heterogeneous distributions

Abstract Estimating the mean of a probability distribution using i.i.d. samples is a classical problem in statistics, wherein finite-sample optimal estimators are sought under various distributional assumptions. In this paper, we consider the problem of mean estimation when independent samples are drawn from $d$-dimensional non-identical distributions possessing a common mean. When the distributions are radially symmetric and unimodal, we propose a novel estimator, which is a hybrid of the modal interval, shorth and median estimators and whose performance adapts to the level of heterogeneity in the data. We show that our estimator is near optimal when data are i.i.d. and when the fraction of ‘low-noise’ distributions is as small as $\varOmega \left (\frac{d \log n}{n}\right )$, where $n$ is the number of samples. We also derive minimax lower bounds on the expected error of any estimator that is agnostic to the scales of individual data points. Finally, we extend our theory to linear regression. In both the mean estimation and regression settings, we present computationally feasible versions of our estimators that run in time polynomial in the number of data points.

Robustness Guarantees for Mode Estimation with an Application to Bandits

Mode estimation is a classical problem in statistics with a wide range of applications in machine learning. Despite this, there is little understanding in its robustness properties under possibly adversarial data contamination. In this paper, we give precise robustness guarantees as well as privacy guarantees under simple randomization. We then introduce a theory for multi-armed bandits where the values are the modes of the reward distributions instead of the mean. We prove regret guarantees for the problems of top arm identification, top m-arms identification, contextual modal bandits, and infinite continuous arms top arm recovery. We show in simulations that our algorithms are robust to perturbation of the arms by adversarial noise sequences, thus rendering modal bandits an attractive choice in situations where the rewards may have outliers or adversarial corruptions.

Preliminary Multiple-Test Estimation, With Applications to k-Sample Covariance Estimation

Multisample covariance estimation—that is, estimation of the covariance matrices associated with k distinct populations—is a classical problem in multivariate statistics. A common solution is to base estimation on the outcome of a test that these covariance matrices show some given pattern. Such a preliminary test may, for example, investigate whether or not the various covariance matrices are equal to each other (test of homogeneity), or whether or not they have common eigenvectors (test of common principal components), etc. Since it is usually unclear what the possible pattern might be, it is natural to consider a collection of such patterns, leading to a collection of preliminary tests, and to base estimation on the outcome of such a multiple testing rule. In the present work, we therefore study preliminary test estimation based on multiple tests. Since this is of interest also outside k-sample covariance estimation, we do so in a very general framework where it is only assumed that the sequence of models at hand is locally asymptotically normal. In this general setup, we define the proposed estimators and derive their asymptotic properties. We come back to k-sample covariance estimation to illustrate the asymptotic and finite-sample behaviors of our estimators. Finally, we treat a real data example that allows us to show their practical relevance in a supervised classification framework.

A Comparison Of Regularized Linear Discriminant Functions For Poorly-Posed Classification Problems

For statistical classification problems where the total sample size is slightly greater than the feature dimension, regularized statistical discriminant rules may reduce classification error rates. We review ten dispersion-matrix regularization approaches, four for the pooled sample covariance matrix, four for the inverse pooled sample covariance matrix, and two for a diagonal covariance matrix, for use in Anderson’s (1951) linear discriminant function (LDF). We compare these regularized classifiers against the traditional LDF for a variety of parameter configurations, and use the estimated expected error rate (EER) to assess performance. We also apply the regularized LDFs to a well-known real-data example on colon cancer. We found that no regularized classifier uniformly outperformed the others. However, we found that the more contemporary classifiers (e.g., Thomaz and Gillies, 2005; Tong et al., 2012; and Xu et al., 2009) tended to outperform the older classifiers, and that certain simple methods (e.g., Pang et al., 2009; Thomaz and Gillies, 2005; and Tong et al., 2012) performed very well, questioning the need for involved cross-validation in estimating regularization parameters. Nonetheless, an older regularized classifier proposed by Smidt and McDonald (1976) yielded consistently low misclassification rates across all scenarios, despite the shape of the true covariance matrix. Finally, our simulations showed that regularized classifiers that relied primarily on asymptotic approximations with respect to the training sample size rarely outperformed the traditional LDF, and are thus not recommended. We discuss our results as they pertain to the effect of high dimension, and offer general guidelines for choosing a regularization method for poorly-posed problems.

Asymptotic optimality in stochastic optimization

We study local complexity measures for stochastic convex optimization problems, providing a local minimax theory analogous to that of Hájek and Le Cam for classical statistical problems. We give complementary optimality results, developing fully online methods that adaptively achieve optimal convergence guarantees. Our results provide function-specific lower bounds and convergence results that make precise a correspondence between statistical difficulty and the geometric notion of tilt-stability from optimization. As part of this development, we show how variants of Nesterov’s dual averaging—a stochastic gradient-based procedure—guarantee finite time identification of constraints in optimization problems, while stochastic gradient procedures fail. Additionally, we highlight a gap between problems with linear and nonlinear constraints: standard stochastic-gradient-based procedures are suboptimal even for the simplest nonlinear constraints, necessitating the development of asymptotically optimal Riemannian stochastic gradient methods.

Statistical treatment of nuclear clusters in the continuum

The evaluation of the sub-saturation nuclear equation of state at finite temperature requires a proper state counting of the internal partition sum of nuclei which are immersed in the background of their continuum states. This classical statistical problem is addressed within the self-consistent mean-field approximation, which naturally accounts for isospin and effective mass effects in the nuclear density of states. The nuclear free energy is decomposed into bulk and surface terms, allowing a simple analytical prescription for the subtraction of gas states from the nuclear partition sum, that avoids double counting of unbound single particle states. We show that this correction leads to a sizeable effect in the composition of matter at high temperature and low proton fractions, such as it is formed in supernova collapse, early proto-neutron star evolution, as well as laboratory experiments. Specifically, the energy stored in the internal nuclear degrees of freedom is reduced, as well as the mass fraction of heavy clusters in the statistical equilibrium. The gas subtraction prescription is compared to different phenomenological methods proposed in the literature, based on a high energy truncation of the partition sum. We show that none of these methods satisfactorily reproduces the gas subtracted level density, if the temperature overcomes ~4 MeV.

Simultaneous estimation of normal means with side information

The integrative analysis of multiple datasets is an important strategy in data analysis. It is increasingly popular in genomics, which enjoys a wealth of publicly available datasets that can be compared, contrasted, and combined in order to extract novel scientific insights. This paper studies a stylized example of data integration for a classical statistical problem: leveraging side information to estimate a vector of normal means. This task is formulated as a compound decision problem, an oracle integrative decision rule is derived, and a data-driven estimate of this rule based on minimizing an unbiased estimate of its risk is proposed. The data-driven rule is shown to asymptotically achieve the minimum possible risk among all separable decision rules, and it can outperform existing methods in numerical properties. The proposed procedure leads naturally to an integrative high-dimensional classification procedure, which is illustrated by combining data from two independent gene expression profiling studies.

Statistical Classification via Robust Hypothesis Testing: Non-Asymptotic and Simple Bounds

We consider Bayesian multiple statistical classification problem in the case where the unknown source distributions are estimated from the labeled training sequences, then the estimates are used as nominal distributions in a robust hypothesis test. Specifically, we employ the DGL test due to Devroye et al. and provide non-asymptotic, exponential upper bounds on the error probability of classification. The proposed upper bounds are simple to evaluate and reveal the effects of the length of the training sequences, the alphabet size and the numbers of hypothesis on the error exponent. The proposed method can also be used for large alphabet sources when the alphabet grows sub-quadratically in the length of the test sequence. The simulations indicate that the performance of the proposed method gets close to that of optimal hypothesis testing as the length of the training sequences increases.

Penalized Fieller's confidence interval for the ratio of bivariate normal means.

Constructing a confidence interval for the ratio of bivariate normal means is a classical problem in statistics. Several methods have been proposed in the literature. The Fieller method is known as an exact method, but can produce an unbounded confidence interval if the denominator of the ratio is not significantly deviated from 0; while the delta and some numeric methods are all bounded, they are only first-order correct. Motivated by a real-world problem, we propose the penalized Fieller method, which employs the same principle as the Fieller method, but adopts a penalized likelihood approach to estimate the denominator. The proposed method has a simple closed form, and can always produce a bounded confidence interval by selecting a suitable penalty parameter. Moreover, the new method is shown to be second-order correct under the bivariate normality assumption, that is, its coverage probability will converge to the nominal level faster than other bounded methods. Simulation results show that our proposed method generally outperforms the existing methods in terms of controlling the coverage probability and the confidence width and is particularly useful when the denominator does not have adequate power to reject being 0. Finally, we apply the proposed approach to the interval estimation of the median response dose in pharmacology studies to show its practicalusefulness.