We introduce the cross-match test - an exact, distribution free, high-dimensional hypothesis test as an intrinsic evaluation metric for word embeddings. We show that cross-match is an effective means of measuring distributional similarity between different vector representations and of evaluating the statistical significance of different vector embedding models. Additionally, we find that cross-match can be used to provide a quantitative measure of linguistic similarity for selecting bridge languages for machine translation. We demonstrate that the results of the hypothesis test align with our expectations and note that the framework of two sample hypothesis testing is not limited to word embeddings and can be extended to all vector representations.
Approximate Markov chain Monte Carlo (MCMC) offers the promise of more rapid sampling at the cost of more biased inference. Since standard MCMC diagnostics fail to detect these biases, researchers have developed computable Stein discrepancy measures that provably determine the convergence of a sample to its target distribution. This approach was recently combined with the theory of reproducing kernels to define a closed-form kernel Stein discrepancy (KSD) computable by summing kernel evaluations across pairs of sample points. We develop a theory of weak convergence for KSDs based on Stein's method, demonstrate that commonly used KSDs fail to detect non-convergence even for Gaussian targets, and show that kernels with slowly decaying tails provably determine convergence for a large class of target distributions. The resulting convergence-determining KSDs are suitable for comparing biased, exact, and deterministic sample sequences and simpler to compute and parallelize than alternative Stein discrepancies. We use our tools to compare biased samplers, select sampler hyperparameters, and improve upon existing KSD approaches to one-sample hypothesis testing and sample quality improvement.
While many multiple graph inference methodologies operate under the implicit assumption that an explicit vertex correspondence is known across the vertex sets of the graphs, in practice these correspondences may only be partially or errorfully known. Herein, we provide an information theoretic foundation for understanding the practical impact that errorfully observed vertex correspondences can have on subsequent inference, and the capacity of graph matching methods to recover the lost vertex alignment and inferential performance. Working in the correlated stochastic blockmodel setting, we establish a duality between the loss of mutual information due to an errorfully observed vertex correspondence and the ability of graph matching algorithms to recover the true correspondence across graphs. In the process, we establish a phase transition for graph matchability in terms of the correlation across graphs, and we conjecture the analogous phase transition for the relative information loss due to shuffling vertex labels. We demonstrate the practical effect that graph shuffling---and matching---can have on subsequent inference, with examples from two sample graph hypothesis testing and joint spectral graph clustering.
Kernel embeddings of distributions and the Maximum Mean Discrepancy (MMD), the resulting distance between distributions, are useful tools for fully nonparametric two-sample testing and learning on distributions. However, it is rarely that all possible differences between samples are of interest -- discovered differences can be due to different types of measurement noise, data collection artefacts or other irrelevant sources of variability. We propose distances between distributions which encode invariance to additive symmetric noise, aimed at testing whether the assumed true underlying processes differ. Moreover, we construct invariant features of distributions, leading to learning algorithms robust to the impairment of the input distributions with symmetric additive noise.
Discovering a correlation from one variable to another variable is of fundamental scientific and practical interest. While existing correlation measures are suitable for discovering average correlation, they fail to discover hidden or potential correlations. To bridge this gap, (i) we postulate a set of natural axioms that we expect a measure of potential correlation to satisfy; (ii) we show that the rate of information bottleneck, i.e., the hypercontractivity coefficient, satisfies all the proposed axioms; (iii) we provide a novel estimator to estimate the hypercontractivity coefficient from samples; and (iv) we provide numerical experiments demonstrating that this proposed estimator discovers potential correlations among various indicators of WHO datasets, is robust in discovering gene interactions from gene expression time series data, and is statistically more powerful than the estimators for other correlation measures in binary hypothesis testing of canonical examples of potential correlations.
The two-sample hypothesis testing problem is studied for the challenging scenario of high dimensional data sets with small sample sizes. We show that the two-sample hypothesis testing problem can be posed as a one-class set classification problem. In the set classification problem the goal is to classify a set of data points that are assumed to have a common class. We prove that the average probability of error given a set is less than or equal to the Bayes error and decreases as a power of $n$ number of sample data points in the set. We use the positive definite Set Kernel for directly mapping sets of data to an associated Reproducing Kernel Hilbert Space, without the need to learn a probability distribution. We specifically solve the two-sample hypothesis testing problem using a one-class SVM in conjunction with the proposed Set Kernel. We compare the proposed method with the Maximum Mean Discrepancy, F-Test and T-Test methods on a number of challenging simulated high dimensional and small sample size data. We also perform two-sample hypothesis testing experiments on six cancer gene expression data sets and achieve zero type-I and type-II error results on all data sets.
We study a novel multi-terminal source coding setup motivated by the biclustering problem. Two separate encoders observe two i.i.d. sequences $X^n$ and $Y^n$, respectively. The goal is to find rate-limited encodings $f(x^n)$ and $g(z^n)$ that maximize the mutual information $I(f(X^n); g(Y^n))/n$. We discuss connections of this problem with hypothesis testing against independence, pattern recognition, and the information bottleneck method. Improving previous cardinality bounds for the inner and outer bounds allows us to thoroughly study the special case of a binary symmetric source and to quantify the gap between the inner and the outer bound in this special case. Furthermore, we investigate a multiple description (MD) extension of the Chief Operating Officer (CEO) problem with mutual information constraint. Surprisingly, this MD-CEO problem permits a tight single-letter characterization of the achievable region.
Sequential hypothesis testing is a desirable decision making strategy in any time sensitive scenario. Compared with fixed sample-size testing, sequential testing is capable of achieving identical probability of error requirements using less samples in average. For a binary detection problem, it is well known that for known density functions accumulating the likelihood ratio statistics is time optimal under a fixed error rate constraint. This paper considers the problem of learning a binary sequential detector from training samples when density functions are unavailable. We formulate the problem as a constrained likelihood ratio estimation which can be solved efficiently through convex optimization by imposing Reproducing Kernel Hilbert Space (RKHS) structure on the log-likelihood ratio function. In addition, we provide a computationally efficient approximated solution for large scale data set. The proposed algorithm, namely Wald-Kernel, is tested on a synthetic data set and two real world data sets, together with previous approaches for likelihood ratio estimation. Our empirical results show that the classifier trained through the proposed technique achieves smaller average sampling cost than previous approaches proposed in the literature for the same error rate.
A fundamental issue for statistical classification models in a streaming environment is that the joint distribution between predictor and response variables changes over time (a phenomenon also known as concept drifts), such that their classification performance deteriorates dramatically. In this paper, we first present a hierarchical hypothesis testing (HHT) framework that can detect and also adapt to various concept drift types (e.g., recurrent or irregular, gradual or abrupt), even in the presence of imbalanced data labels. A novel concept drift detector, namely Hierarchical Linear Four Rates (HLFR), is implemented under the HHT framework thereafter. By substituting a widely-acknowledged retraining scheme with an adaptive training strategy, we further demonstrate that the concept drift adaptation capability of HLFR can be significantly boosted. The theoretical analysis on the Type-I and Type-II errors of HLFR is also performed. Experiments on both simulated and real-world datasets illustrate that our methods outperform state-of-the-art methods in terms of detection precision, detection delay as well as the adaptability across different concept drift types.
Markov models lie at the interface between statistical independence in a probability distribution and graph separation properties. We review model selection and estimation in directed and undirected Markov models with Gaussian parametrization, emphasizing the main similarities and differences. These two model classes are similar but not equivalent, although they share a common intersection. We present the existing results from a historical perspective, taking into account the amount of literature existing from both the artificial intelligence and statistics research communities, where these models were originated. We cover classical topics such as maximum likelihood estimation and model selection via hypothesis testing, but also more modern approaches like regularization and Bayesian methods. We also discuss how the Markov models reviewed fit in the rich hierarchy of other, higher level Markov model classes. Finally, we close the paper overviewing relaxations of the Gaussian assumption and pointing out the main areas of application where these Markov models are nowadays used.
The statistical analysis of discrete data has been the subject of extensive statistical research dating back to the work of Pearson. In this survey we review some recently developed methods for testing hypotheses about high-dimensional multinomials. Traditional tests like the $\chi^2$ test and the likelihood ratio test can have poor power in the high-dimensional setting. Much of the research in this area has focused on finding tests with asymptotically Normal limits and developing (stringent) conditions under which tests have Normal limits. We argue that this perspective suffers from a significant deficiency: it can exclude many high-dimensional cases when - despite having non Normal null distributions - carefully designed tests can have high power. Finally, we illustrate that taking a minimax perspective and considering refinements of this perspective can lead naturally to powerful and practical tests.
In sequential hypothesis testing, Generalized Binary Search (GBS) greedily chooses the test with the highest information gain at each step. It is known that GBS obtains the gold standard query cost of $O(\log n)$ for problems satisfying the $k$-neighborly condition, which requires any two tests to be connected by a sequence of tests where neighboring tests disagree on at most $k$ hypotheses. In this paper, we introduce a weaker condition, split-neighborly, which requires that for the set of hypotheses two neighbors disagree on, any subset is splittable by some test. For four problems that are not $k$-neighborly for any constant $k$, we prove that they are split-neighborly, which allows us to obtain the optimal $O(\log n)$ worst-case query cost.
The amount of digital imagery recorded has recently grown exponentially, and with the advancement of software, such as Photoshop or Gimp, it has become easier to manipulate images. However, most images on the internet have not been manipulated and any automated manipulation detection algorithm must carefully control the false alarm rate. In this paper we discuss a method to automatically detect local resampling using deep learning while controlling the false alarm rate using a-contrario analysis. The automated procedure consists of three primary steps. First, resampling features are calculated for image blocks. A deep learning classifier is then used to generate a heatmap that indicates if the image block has been resampled. We expect some of these blocks to be falsely identified as resampled. We use a-contrario hypothesis testing to both identify if the patterns of the manipulated blocks indicate if the image has been tampered with and to localize the manipulation. We demonstrate that this strategy is effective in indicating if an image has been manipulated and localizing the manipulations.
AI researchers employ not only the scientific method, but also methodology from mathematics and engineering. However, the use of the scientific method - specifically hypothesis testing - in AI is typically conducted in service of engineering objectives. Growing interest in topics such as fairness and algorithmic bias show that engineering-focused questions only comprise a subset of the important questions about AI systems. This results in the AI Knowledge Gap: the number of unique AI systems grows faster than the number of studies that characterize these systems' behavior. To close this gap, we argue that the study of AI could benefit from the greater inclusion of researchers who are well positioned to formulate and test hypotheses about the behavior of AI systems. We examine the barriers preventing social and behavioral scientists from conducting such studies. Our diagnosis suggests that accelerating the scientific study of AI systems requires new incentives for academia and industry, mediated by new tools and institutions. To address these needs, we propose a two-sided marketplace called TuringBox. On one side, AI contributors upload existing and novel algorithms to be studied scientifically by others. On the other side, AI examiners develop and post machine intelligence tasks designed to evaluate and characterize algorithmic behavior. We discuss this market's potential to democratize the scientific study of AI behavior, and thus narrow the AI Knowledge Gap.
A group of transition probability functions form a Shannon's channel whereas a group of truth functions form a semantic channel. By the third kind of Bayes' theorem, we can directly convert a Shannon's channel into an optimized semantic channel. When a sample is not big enough, we can use a truth function with parameters to produce the likelihood function, then train the truth function by the conditional sampling distribution. The third kind of Bayes' theorem is proved. A semantic information theory is simply introduced. The semantic information measure reflects Popper's hypothesis-testing thought. The Semantic Information Method (SIM) adheres to maximum semantic information criterion which is compatible with maximum likelihood criterion and Regularized Least Squares criterion. It supports Wittgenstein's view: the meaning of a word lies in its use. Letting the two channels mutually match, we obtain the Channels' Matching (CM) algorithm for machine learning. The CM algorithm is used to explain the evolution of the semantic meaning of natural language, such as "Old age". The semantic channel for medical tests and the confirmation measures of test-positive and test-negative are discussed. The applications of the CM algorithm to semi-supervised learning and non-supervised learning are simply introduced. As a predictive model, the semantic channel fits variable sources and hence can overcome class-imbalance problem. The SIM strictly distinguishes statistical probability and logical probability and uses both at the same time. This method is compatible with the thoughts of Bayes, Fisher, Shannon, Zadeh, Tarski, Davidson, Wittgenstein, and Popper.It is a competitive alternative to Bayesian inference.
Textual grounding, i.e., linking words to objects in images, is a challenging but important task for robotics and human-computer interaction. Existing techniques benefit from recent progress in deep learning and generally formulate the task as a supervised learning problem, selecting a bounding box from a set of possible options. To train these deep net based approaches, access to a large-scale datasets is required, however, constructing such a dataset is time-consuming and expensive. Therefore, we develop a completely unsupervised mechanism for textual grounding using hypothesis testing as a mechanism to link words to detected image concepts. We demonstrate our approach on the ReferIt Game dataset and the Flickr30k data, outperforming baselines by 7.98% and 6.96% respectively.
In statistical inference problems, we wish to obtain lower bounds on the minimax risk, that is to bound the performance of any possible estimator. A standard technique to obtain risk lower bounds involves the use of Fano's inequality. In an information-theoretic setting, it is known that Fano's inequality typically does not give a sharp converse result (error lower bound) for channel coding problems. Moreover, recent work has shown that an argument based on binary hypothesis testing gives tighter results. We adapt this technique to the statistical setting, and argue that Fano's inequality can always be replaced by this approach to obtain tighter lower bounds that can be easily computed and are asymptotically sharp. We illustrate our technique in three applications: density estimation, active learning of a binary classifier, and compressed sensing, obtaining tighter risk lower bounds in each case.
We investigate the problem of identity testing for multidimensional histogram distributions. A distribution $p: D \rightarrow \mathbb{R}_+$, where $D \subseteq \mathbb{R}^d$, is called a $k$-histogram if there exists a partition of the domain into $k$ axis-aligned rectangles such that $p$ is constant within each such rectangle. Histograms are one of the most fundamental nonparametric families of distributions and have been extensively studied in computer science and statistics. We give the first identity tester for this problem with {\em sub-learning} sample complexity in any fixed dimension and a nearly-matching sample complexity lower bound. In more detail, let $q$ be an unknown $d$-dimensional $k$-histogram distribution in fixed dimension $d$, and $p$ be an explicitly given $d$-dimensional $k$-histogram. We want to correctly distinguish, with probability at least $2/3$, between the case that $p = q$ versus $\|p-q\|_1 \geq \epsilon$. We design an algorithm for this hypothesis testing problem with sample complexity $O((\sqrt{k}/\epsilon^2) 2^{d/2} \log^{2.5 d}(k/\epsilon))$ that runs in sample-polynomial time. Our algorithm is robust to model misspecification, i.e., succeeds even if $q$ is only promised to be {\em close} to a $k$-histogram. Moreover, for $k = 2^{\Omega(d)}$, we show a sample complexity lower bound of $(\sqrt{k}/\epsilon^2) \cdot \Omega(\log(k)/d)^{d-1}$ when $d\geq 2$. That is, for any fixed dimension $d$, our upper and lower bounds are nearly matching. Prior to our work, the sample complexity of the $d=1$ case was well-understood, but no algorithm with sub-learning sample complexity was known, even for $d=2$. Our new upper and lower bounds have interesting conceptual implications regarding the relation between learning and testing in this setting.
Generative adversarial networks (GANs) are innovative techniques for learning generative models of complex data distributions from samples. Despite remarkable recent improvements in generating realistic images, one of their major shortcomings is the fact that in practice, they tend to produce samples with little diversity, even when trained on diverse datasets. This phenomenon, known as mode collapse, has been the main focus of several recent advances in GANs. Yet there is little understanding of why mode collapse happens and why existing approaches are able to mitigate mode collapse. We propose a principled approach to handling mode collapse, which we call packing. The main idea is to modify the discriminator to make decisions based on multiple samples from the same class, either real or artificially generated. We borrow analysis tools from binary hypothesis testing---in particular the seminal result of Blackwell [Bla53]---to prove a fundamental connection between packing and mode collapse. We show that packing naturally penalizes generators with mode collapse, thereby favoring generator distributions with less mode collapse during the training process. Numerical experiments on benchmark datasets suggests that packing provides significant improvements in practice as well.
Cellular Electron CryoTomography (CECT) is a 3D imaging technique that captures information about the structure and spatial organization of macromolecular complexes within single cells, in near-native state and at sub-molecular resolution. Although template matching is often used to locate macromolecules in a CECT image, it is insufficient as it only measures the relative structural similarity. Therefore, it is preferable to assess the statistical credibility of the decision through hypothesis testing, requiring many templates derived from a diverse population of macromolecular structures. Due to the very limited number of known structures, we need a generative model to efficiently and reliably sample pseudo-structures from the complex distribution of macromolecular structures. To address this challenge, we propose a novel image-derived approach for performing hypothesis testing for template matching by constructing generative models using the generative adversarial network. Finally, we conducted hypothesis testing experiments for template matching on both simulated and experimental subtomograms, allowing us to conclude the identity of subtomograms with high statistical credibility and significantly reducing false positives.