Literatura científica selecionada sobre o tema "Randomized sketches"

Quantiles are fundamental statistics in various data science tasks, but costly to compute, e.g., by loading the entire data in memory for ranking. With limited memory space, prevalent in end devices or databases with heavy loads, it needs to scan the data in multiple passes. The idea is to gradually shrink the range of the queried quantile till it is small enough to fit in memory for ranking the result. Existing methods use deterministic sketches to determine the exact range of quantile, known as deterministic filter, which could be inefficient in range shrinking. In this study, we propose to shrink the ranges more aggressively, using randomized summaries such as KLL sketch. That is, with a high probability the quantile lies in a smaller range, namely probabilistic filter, determined by the randomized sketch. Specifically, we estimate the expected passes for determining the exact quantiles with probabilistic filters, and select a proper probability that can minimize the expected passes. Analyses show that our exact quantile determination method can terminate in P passes with 1-δ confidence, storing O(N 1/P logP-1/2P (1/δ)) items, close to the lower bound Ømega(N1/P) for a fixed δ. The approach has been deployed as a function in an LSM-tree based time-series database Apache IoTDB. Remarkably, the randomized sketches can be pre-computed for the immutable SSTables in LSM-tree. Moreover, multiple quantile queries could share the data passes for probabilistic filters in range estimation. Extensive experiments on real and synthetic datasets demonstrate the superiority of our proposal compared to the existing methods with deterministic filters. On average, our method takes 0.48 fewer passes and 18% of the time compared with the state-of-the-art deterministic sketch (GK sketch).

Image segmentation process for high quality visual saliency map is very dependent on the existing visual saliency metrics. It is mostly only get sketchy effect of saliency map, and roughly based visual saliency map will affect the image segmentation results. The paper had presented the randomized visual saliency detection algorithm. The randomized visual saliency detection method can quickly generate the same size as the original input image and detailed results of the saliency map. The randomized saliency detection method can be applied to real-time requirements for image content-based scaling saliency results map. The randomization method for fast randomized video saliency area detection, the algorithm only requires a small amount of memory space can be detected detailed oriented visual saliency map, the presented results are shown that the method of visual saliency map used in image after the segmentation process can be an ideal segmentation results.

Measuring flow spread in real time from large, high-rate data streams has numerous practical applications, where a data stream is modeled as a sequence of data items from different flows and the spread of a flow is the number of distinct items in the flow. Past decades have witnessed tremendous performance improvement for single-flow spread estimation. However, when dealing with numerous flows in a data stream, it remains a significant challenge to measure per-flow spread accurately while reducing memory footprint. The goal of this paper is to introduce new multi-flow spread estimation designs that incur much smaller processing overhead and query overhead than the state of the art, yet achieves significant accuracy improvement in spread estimation. We formally analyze the performance of these new designs. We implement them in both hardware and software, and use real-world data traces to evaluate their performance in comparison with the state of the art. The experimental results show that our best sketch significantly improves over the best existing work in terms of estimation accuracy, data item processing throughput, and online query throughput.

CountSketch and Feature Hashing (the ``hashing trick'') are popular randomized dimensionality reduction methods that support recovery of l2 -heavy hitters and approximate inner products. When the inputs are not adaptive (do not depend on prior outputs), classic estimators applied to a sketch of size O(l / epsilon) are accurate for a number of queries that is exponential in l. When inputs are adaptive, however, an adversarial input can be constructed after O(l) queries with the classic estimator and the best known robust estimator only supports ~O(l^2) queries. In this work we show that this quadratic dependence is in a sense inherent: We design an attack that after O(l^2) queries produces an adversarial input vector whose sketch is highly biased. Our attack uses ``natural'' non-adaptive inputs (only the final adversarial input is chosen adaptively) and universally applies with any correct estimator, including one that is unknown to the attacker. In that, we expose inherent vulnerability of this fundamental method.

Les noyaux de produit scalaire, tels que les noyaux polynomiaux et exponentiels (softmax), sont parmi les noyaux les plus utilisés en apprentissage automatique, car ils permettent de modéliser les interactions entre les composantes des vecteurs d'entrée, ce qui est crucial dans des applications telles que la vision par ordinateur, le traitement du langage naturel et les systèmes de recommandation. Cependant, un inconvénient fondamental des modèles statistiques basés sur les noyaux est leur évolutivité limitée à un grand nombre de données d'entrée, ce qui nécessite de recourir à des approximations. Dans cette thèse, nous étudions des techniques pour linéariser les méthodes à base de noyaux de produit scalaire au moyen d'approximations de caractéristiques aléatoires. En particulier, nous nous concentrons sur une analyse de variance pour étudier et améliorer leur efficacité statistique
Dot product kernels, such as polynomial and exponential (softmax) kernels, are among the most widely used kernels in machine learning, as they enable modeling the interactions between input features, which is crucial in applications like computer vision, natural language processing, and recommender systems. However, a fundamental drawback of kernel-based statistical models is their limited scalability to a large number of inputs, which requires resorting to approximations. In this thesis, we study techniques to linearize kernel-based methods by means of random feature approximations and we focus on the approximation of polynomial kernels and more general dot product kernels to make these kernels more useful in large scale learning. In particular, we focus on a variance analysis as a main tool to study and improve the statistical efficiency of such sketches

Probabilistic ideas and tools have recently begun to permeate into several fields where they had traditionally not played a major role, including fields such as numerical linear algebra and optimization. One of the key ways in which these ideas influence these fields is via the development and analysis of randomized algorithms for solving standard and new problems of these fields. Such methods are typically easier to analyze, and often lead to faster and/or more scalable and versatile methods in practice. This thesis explores the design and analysis of new randomized iterative methods for solving linear systems and inverting matrices. The methods are based on a novel sketch-and-project framework. By sketching we mean, to start with a difficult problem and then randomly generate a simple problem that contains all the solutions of the original problem. After sketching the problem, we calculate the next iterate by projecting our current iterate onto the solution space of the sketched problem. The starting point for this thesis is the development of an archetype randomized method for solving linear systems. Our method has six different but equivalent interpretations: sketch-and-project, constrain-and-approximate, random intersect, random linear solve, random update and random fixed point. By varying its two parameters – a positive definite matrix (defining geometry), and a random matrix (sampled in an i.i.d. fashion in each iteration) – we recover a comprehensive array of well known algorithms as special cases, including the randomized Kaczmarz method, randomized Newton method, randomized coordinate descent method and random Gaussian pursuit. We also naturally obtain variants of all these methods using blocks and importance sampling. However, our method allows for a much wider selection of these two parameters, which leads to a number of new specific methods. We prove exponential convergence of the expected norm of the error in a single theorem, from which existing complexity results for known variants can be obtained. However, we also give an exact formula for the evolution of the expected iterates, which allows us to give lower bounds on the convergence rate. We then extend our problem to that of finding the projection of given vector onto the solution space of a linear system. For this we develop a new randomized iterative algorithm: stochastic dual ascent (SDA). The method is dual in nature, and iteratively solves the dual of the projection problem. The dual problem is a non-strongly concave quadratic maximization problem without constraints. In each iteration of SDA, a dual variable is updated by a carefully chosen point in a subspace spanned by the columns of a random matrix drawn independently from a fixed distribution. The distribution plays the role of a parameter of the method. Our complexity results hold for a wide family of distributions of random matrices, which opens the possibility to fine-tune the stochasticity of the method to particular applications. We prove that primal iterates associated with the dual process converge to the projection exponentially fast in expectation, and give a formula and an insightful lower bound for the convergence rate. We also prove that the same rate applies to dual function values, primal function values and the duality gap. Unlike traditional iterative methods, SDA converges under virtually no additional assumptions on the system (e.g., rank, diagonal dominance) beyond consistency. In fact, our lower bound improves as the rank of the system matrix drops. By mapping our dual algorithm to a primal process, we uncover that the SDA method is the dual method with respect to the sketch-and-project method from the previous chapter. Thus our new more general convergence results for SDA carry over to the sketch-and-project method and all its specializations (randomized Kaczmarz, randomized coordinate descent ... etc.). When our method specializes to a known algorithm, we either recover the best known rates, or improve upon them. Finally, we show that the framework can be applied to the distributed average consensus problem to obtain an array of new algorithms. The randomized gossip algorithm arises as a special case. In the final chapter, we extend our method for solving linear system to inverting matrices, and develop a family of methods with specialized variants that maintain symmetry or positive definiteness of the iterates. All the methods in the family converge globally and exponentially, with explicit rates. In special cases, we obtain stochastic block variants of several quasi-Newton updates, including bad Broyden (BB), good Broyden (GB), Powell-symmetric-Broyden (PSB), Davidon-Fletcher-Powell (DFP) and Broyden-Fletcher-Goldfarb-Shanno (BFGS). Ours are the first stochastic versions of these updates shown to converge to an inverse of a fixed matrix. Through a dual viewpoint we uncover a fundamental link between quasi-Newton updates and approximate inverse preconditioning. Further, we develop an adaptive variant of the randomized block BFGS (AdaRBFGS), where we modify the distribution underlying the stochasticity of the method throughout the iterative process to achieve faster convergence. By inverting several matrices from varied applications, we demonstrate that AdaRBFGS is highly competitive when compared to the well established Newton-Schulz and approximate preconditioning methods. In particular, on large-scale problems our method outperforms the standard methods by orders of magnitude. The development of efficient methods for estimating the inverse of very large matrices is a much needed tool for preconditioning and variable metric methods in the big data era.

AbstractThis paper presents PAYNT, a tool to automatically synthesise probabilistic programs. PAYNT enables the synthesis of finite-state probabilistic programs from a program sketch representing a finite family of program candidates. A tight interaction between inductive oracle-guided methods with state-of-the-art probabilistic model checking is at the heart of PAYNT. These oracle-guided methods effectively reason about all possible candidates and synthesise programs that meet a given specification formulated as a conjunction of temporal logic constraints and possibly including an optimising objective. We demonstrate the performance and usefulness of PAYNT using several case studies from different application domains; e.g., we find the optimal randomized protocol for network stabilisation among 3M potential programs within minutes, whereas alternative approaches would need days to do so.

Abstract This chapter introduces the three components of a generalized linear model (GLM): the linear predictor, the link function, and the probability function. It discusses the exponential dispersion family as a generator model for GLMs in a large sense. It sketches the fitting of a GLM with the iteratively weighted least squares algorithm for maximum likelihood in the frequentist framework. It introduces the main methods for assessing the effects of explanatory variables in frequentist GLMs (the Wald and likelihood ratio tests), the use of deviance as a measure of lack of model fit in GLMs, and the main types of residuals (Pearson, deviance, and randomized quantile) used in GLM model validation. It also discusses Bayesian fitting of GLMs, and some issues involved in defining priors for the GLM parameters.

The evidence for AMBIT is described, together with our intentions regarding future development and dissemination. Existing evidence, and its limits, is sketched out. Challenges for future data collection/analysis are highlighted in relation to a “model” that is, by design, broad and adaptable, rendering “fidelity” difficult to define in conventional terms. Opportunities for meaningful randomized trials are limited, but possible. A description of AMBIT’s strategy for future model development (referencing open-source computing, learning organizations, and quality improvement models) concludes with a summary of AMBIT’s “evidence-informed” approach. AMBIT’s role as a learning system is supported by innovative web-based collaborations (for instance POD, a pooled online data system). Diverse training methods and reflective self-audit questions for teams considering training are described. Skepticism about how conventional trainings could ever create the necessary “step change” in services is balanced by the opportunities for dissemination offered by web-based learning and devolved engagement of a community of practice.