Inspired by empirical studies of networked systems such as the Internet, social networks, and biological networks, researchers have in recent years developed a variety of techniques and models to help us understand or predict the behavior of these systems. Here we review developments in this field, including such concepts as the small-world effect, degree distributions, clustering, network correlations, random graph models, models of network growth and preferential attachment, and dynamical processes taking place on networks.

/pdf/the-structure-and-function-of-complex-networks-34deompc9l.pdf

The Structure and Function of Complex Networks

Large sample properties of generalized method of moments estimators

This paper provides a theoretical framework for analysis of consensus algorithms for multi-agent networked systems with an emphasis on the role of directed information flow, robustness to changes in network topology due to link/node failures, time-delays, and performance guarantees. An overview of basic concepts of information consensus in networks and methods of convergence and performance analysis for the algorithms are provided. Our analysis framework is based on tools from matrix theory, algebraic graph theory, and control theory. We discuss the connections between consensus problems in networked dynamic systems and diverse applications including synchronization of coupled oscillators, flocking, formation control, fast consensus in small-world networks, Markov processes and gossip-based algorithms, load balancing in networks, rendezvous in space, distributed sensor fusion in sensor networks, and belief propagation. We establish direct connections between spectral and structural properties of complex networks and the speed of information diffusion of consensus algorithms. A brief introduction is provided on networked systems with nonlocal information flow that are considerably faster than distributed systems with lattice-type nearest neighbor interactions. Simulation results are presented that demonstrate the role of small-world effects on the speed of consensus algorithms and cooperative control of multivehicle formations

/pdf/consensus-and-cooperation-in-networked-multi-agent-systems-5ame7z72ny.pdf

Consensus and Cooperation in Networked Multi-Agent Systems

Today's Web-enabled deluge of electronic data calls for automated methods of data analysis. Machine learning provides these, developing methods that can automatically detect patterns in data and then use the uncovered patterns to predict future data. This textbook offers a comprehensive and self-contained introduction to the field of machine learning, based on a unified, probabilistic approach. The coverage combines breadth and depth, offering necessary background material on such topics as probability, optimization, and linear algebra as well as discussion of recent developments in the field, including conditional random fields, L1 regularization, and deep learning. The book is written in an informal, accessible style, complete with pseudo-code for the most important algorithms. All topics are copiously illustrated with color images and worked examples drawn from such application domains as biology, text processing, computer vision, and robotics. Rather than providing a cookbook of different heuristic methods, the book stresses a principled model-based approach, often using the language of graphical models to specify models in a concise and intuitive way. Almost all the models described have been implemented in a MATLAB software package--PMTK (probabilistic modeling toolkit)--that is freely available online. The book is suitable for upper-level undergraduates with an introductory-level college math background and beginning graduate students.

Machine Learning : A Probabilistic Perspective

The role of imperfect information in a principal-agent relationship subject to moral hazard is considered. A necessary and sufficient condition for imperfect information to improve on contracts based on the payoff alone is derived, and a characterization of the optimal use of such information is given.

/pdf/moral-hazard-and-observability-719lvi5me9.pdf

Moral Hazard and Observability

Foreword.Preface.PART ONE. SURVEY OF PROBABILITY THEORY.Chapter 1. Introduction.Chapter 2. Experiments, Sample Spaces, and Probability.2.1 Experiments and Sample Spaces.2.2 Set Theory.2.3 Events and Probability.2.4 Conditional Probability.2.5 Binomial Coefficients.Exercises.Chapter 3. Random Variables, Random Vectors, and Distributions Functions.3.1 Random Variables and Their Distributions.3.2 Multivariate Distributions.3.3 Sums and Integrals.3.4 Marginal Distributions and Independence.3.5 Vectors and Matrices.3.6 Expectations, Moments, and Characteristic Functions.3.7 Transformations of Random Variables.3.8 Conditional Distributions.Exercises.Chapter 4. Some Special Univariate Distributions.4.1 Introduction.4.2 The Bernoulli Distributions.4.3 The Binomial Distribution.4.4 The Poisson Distribution.4.5 The Negative Binomial Distribution.4.6 The Hypergeometric Distribution.4.7 The Normal Distribution.4.8 The Gamma Distribution.4.9 The Beta Distribution.4.10 The Uniform Distribution.4.11 The Pareto Distribution.4.12 The t Distribution.4.13 The F Distribution.Exercises.Chapter 5. Some Special Multivariate Distributions.5.1 Introduction.5.2 The Multinomial Distribution.5.3 The Dirichlet Distribution.5.4 The Multivariate Normal Distribution.5.5 The Wishart Distribution.5.6 The Multivariate t Distribution.5.7 The Bilateral Bivariate Pareto Distribution.Exercises.PART TWO. SUBJECTIVE PROBABILITY AND UTILITY.Chapter 6. Subjective Probability.6.1 Introduction.6.2 Relative Likelihood.6.3 The Auxiliary Experiment.6.4 Construction of the Probability Distribution.6.5 Verification of the Properties of a Probability Distribution.6.6 Conditional Likelihoods.Exercises.Chapter 7. Utility.7.1 Preferences Among Rewards.7.2 Preferences Among Probability Distributions.7.3 The Definitions of a Utility Function.7.4 Some Properties of Utility Functions.7.5 The Utility of Monetary Rewards.7.6 Convex and Concave Utility Functions.7.7 The Anxiomatic Development of Utility.7.8 Construction of the Utility Function.7.9 Verification of the Properties of a Utility Function.7.10 Extension of the Properties of a Utility Function to the Class ?E.Exercises.PART THREE. STATISTICAL DECISION PROBLEMS.Chapter 8. Decision Problems.8.1 Elements of a Decision Problem.8.2 Bayes Risk and Bayes Decisions.8.3 Nonnegative Loss Functions.8.4 Concavity of the Bayes Risk.8.5 Randomization and Mixed Decisions.8.6 Convex Sets.8.7 Decision Problems in Which ~2 and D Are Finite.8.8 Decision Problems with Observations.8.9 Construction of Bayes Decision Functions.8.10 The Cost of Observation.8.11 Statistical Decision Problems in Which Both ? and D contains Two Points.8.12 Computation of the Posterior Distribution When the Observations Are Made in More Than One Stage.Exercises.Chapter 9. Conjugate Prior Distributions.9.1 Sufficient Statistics.9.2 Conjugate Families of Distributions.9.3 Construction of the Conjugate Family.9.4 Conjugate Families for Samples from Various Standard Distributions.9.5 Conjugate Families for Samples from a Normal Distribution.9.6 Sampling from a Normal Distribution with Unknown Mean and Unknown Precision.9.7 Sampling from a Uniform Distribution.9.8 A Conjugate Family for Multinomial Observations.9.9 Conjugate Families for Samples from a Multivariate Normal Distribution.9.10 Multivariate Normal Distributions with Unknown Mean Vector and Unknown Precision matrix.9.11 The Marginal Distribution of the Mean Vector.9.12 The Distribution of a Correlation.9.13 Precision Matrices Having an Unknown Factor.Exercises.Chapter 10. Limiting Posterior Distributions.10.1 Improper Prior Distributions.10.2 Improper Prior Distributions for Samples from a Normal Distribution.10.3 Improper Prior Distributions for Samples from a Multivariate Normal Distribution.10.4 Precise Measurement.10.5 Convergence of Posterior Distributions.10.6 Supercontinuity.10.7 Solutions of the Likelihood Equation.10.8 Convergence of Supercontinuous Functions.10.9 Limiting Properties of the Likelihood Function.10.10 Normal Approximation to the Posterior Distribution.10.11 Approximation for Vector Parameters.10.12 Posterior Ratios.Exercises.Chapter 11. Estimation, Testing Hypotheses, and linear Statistical Models.11.1 Estimation.11.2 Quadratic Loss.11.3 Loss Proportional to the Absolute Value of the Error.11.4 Estimation of a Vector.11.5 Problems of Testing Hypotheses.11.6 Testing a Simple Hypothesis About the Mean of a Normal Distribution.11.7 Testing Hypotheses about the Mean of a Normal Distribution.11.8 Deciding Whether a Parameter Is Smaller or larger Than a Specific Value.11.9 Deciding Whether the Mean of a Normal Distribution Is Smaller or larger Than a Specific Value.11.10 Linear Models.11.11 Testing Hypotheses in Linear Models.11.12 Investigating the Hypothesis That Certain Regression Coefficients Vanish.11.13 One-Way Analysis of Variance.Exercises.PART FOUR. SEQUENTIAL DECISIONS.Chapter 12. Sequential Sampling.12.1 Gains from Sequential Sampling.12.2 Sequential Decision Procedures.12.3 The Risk of a Sequential Decision Procedure.12.4 Backward Induction.12.5 Optimal Bounded Sequential Decision procedures.12.6 Illustrative Examples.12.7 Unbounded Sequential Decision Procedures.12.8 Regular Sequential Decision Procedures.12.9 Existence of an Optimal Procedure.12.10 Approximating an Optimal Procedure by Bounded Procedures.12.11 Regions for Continuing or Terminating Sampling.12.12 The Functional Equation.12.13 Approximations and Bounds for the Bayes Risk.12.14 The Sequential Probability-ratio Test.12.15 Characteristics of Sequential Probability-ratio Tests.12.16 Approximating the Expected Number of Observations.Exercises.Chapter 13. Optimal Stopping.13.1 Introduction.13.2 The Statistician's Reward.13.3 Choice of the Utility Function.13.4 Sampling Without Recall.13.5 Further Problems of Sampling with Recall and Sampling without Recall.13.6 Sampling without Recall from a Normal Distribution with Unknown Mean.13.7 Sampling with Recall from a Normal Distribution with Unknown Mean.13.8 Existence of Optimal Stopping Rules.13.9 Existence of Optimal Stopping Rules for Problems of Sampling with Recall and Sampling without Recall.13.10 Martingales.13.11 Stopping Rules for Martingales.13.12 Uniformly Integrable Sequences of Random Variables.13.13 Martingales Formed from Sums and Products of Random Variables.13.14 Regular Supermartingales.13.15 Supermartingales and General Problems of Optimal Stopping.13.16 Markov Processes.13.17 Stationary Stopping Rules for Markov Processes.13.18 Entrance-fee Problems.13.19 The Functional Equation for a Markov Process.Exercises.Chapter 14. Sequential Choice of Experiments.14.1 Introduction.14.2 Markovian Decision Processes with a Finite Number of Stages.14.3 Markovian Decision Processes with an Infinite Number of Stages.14.4 Some Betting Problems.14.5 Two-armed-bandit Problems.14.6 Two-armed-bandit Problems When the Value of One Parameter Is Known.14.7 Two-armed-bandit Problems When the Parameters Are Dependent.14.8 Inventory Problems.14.9 Inventory Problems with an Infinite Number of Stages.14.10 Control Problems.14.11 Optimal Control When the Process Cannot Be Observed without Error.14.12 Multidimensional Control Problems.14.13 Control Problems with Actuation Errors.14.14 Search Problems.14.15 Search Problems with Equal Costs.14.16 Uncertainty Functions and Statistical Decision Problems.14.17 Sufficient Experiments.14.18 Examples of Sufficient Experiments.Exercises.References.Supplementary Bibliography.Name Index.Subject Index.

Optimal Statistical Decisions

Consider a group of individuals who must act together as a team or committee, and suppose that each individual in the group has his own subjective probability distribution for the unknown value of some parameter. A model is presented which describes how the group might reach agreement on a common subjective probability distribution for the parameter by pooling their individual opinions. The process leading to the consensus is explicitly described and the common distribution that is reached is explicitly determined. The model can also be applied to problems of reaching a consensus when the opinion of each member of the group is represented simply as a point estimate of the parameter rather than as a probability distribution.

Reaching a Consensus

1. Introduction to Probability. The History of Probability. Interpretations of Probability. Experiments and Events. Set Theory. The Definition of Probability. Finite Sample Spaces. Counting Methods. Combinatorial Methods. Multinomial Coefficients. The Probability of a Union of Events. Statistical Swindles. Supplementary Exercises. 2. Conditional Probability. The Definition of Conditional Probability. Independent Events. Bayes' Theorem. Markov Chains. The Gambler's Ruin Problem. Supplementary Exercises. 3. Random Variables and Distribution. Random Variables and Discrete Distributions. Continuous Distributions. The Distribution Function. Bivariate Distributions. Marginal Distributions. Conditional Distributions. Multivariate Distributions. Functions of a Random Variable. Functions of Two or More Random Variables. Supplementary Exercises. 4. Expectation. The Expectation of a Random Variable. Properties of Expectations. Variance. Moments. The Mean and The Median. Covariance and Correlation. Conditional Expectation. The Sample Mean. Utility. Supplementary Exercises. 5. Special Distributions. Introduction. The Bernoulli and Binomial Distributions. The Hypergeometric Distribution. The Poisson Distribution. The Negative Binomial Distribution. The Normal Distribution. The Central Limit Theorem. The Correction for Continuity. The Gamma Distribution. The Beta Distribution. The Multinomial Distribution. The Bivariate Normal Distribution. Supplementary Exercises. 6. Estimation. Statistical Inference. Prior and Posterior Distributions. Conjugate Prior Distributions. Bayes Estimators. Maximum Likelihood Estimators. Properties of Maximum Likelihood Estimators. Sufficient Statistics. Jointly Sufficient Statistics. Improving an Estimator. Supplementary Exercises. 7. Sampling Distributions of Estimators. The Sampling Distribution of a Statistic. The Chi-Square Distribution. Joint Distribution of the Sample Mean and Sample Variance. The t Distribution. Confidence Intervals. Bayesian Analysis of Samples from a Normal Distribution. Unbiased Estimators. Fisher Information. Supplementary Exercises. 8. Testing Hypotheses. Problems of Testing Hypotheses. Testing Simple Hypotheses. Uniformly Most Powerful Tests. Two-Sided Alternatives. The t Test. Comparing the Means of Two Normal Distributions. The F Distribution. Bayes Test Procedures. Foundational Issues. Supplementary Exercises. 9. Categorical Data and Nonparametric Methods. Tests of Goodness-of-Fit. Goodness-of-Fit for Composite Hypotheses. Contingency Tables. Tests of Homogeneit. Simpson's Paradox. Kolmogorov-Smirnov Test. Robust Estimation. Sign and Rank Tests. Supplementary Exercises. 10. Linear Statistical Models. The Method of Least Squares. Regression. Statistical Inference in Simple Linear Regression. Bayesian Inference in Simple Linear Regression. The General Linear Model and Multiple Regression. Analysis of Variance. The Two-Way Layout. The Two-Way Layout with Replications. Supplementary Exercises. 11. Simulation. Why is Simulation Useful? Simulating Specific Distributions. Importance Sampling. Markov Chain Monte Carlo. The Bootstrap. Supplementary Exercises.

Probability and statistics

In this paper we present methods for comparing and evaluating forecasters whose predictions are presented as their subjective probability distributions of various random variables that will be observed in the future, e.g. weather forecasters who each day must specify their own probabilities that it will rain in a particular location. We begin by reviewing the concepts of calibration and refinement, and describiing the relationship between this notion of refinement and the notion of sufficiency in the comparison of statistical experiments. We also consider the question of interrelationships among forecasters and discuss methods by which an observer should combine the predictions from two or more different forecasters. Then we turn our attention to the concept of a proper scoring rule for evaluating forecasters, relating it to the concepts of calibration and refinement. Finally, we discuss conditions under which one fore- caster can exploit the predictions of another forecaster to obtain a better score. In this paper we describe some concepts and methods appropriate for evaluating and com- paring forecasters who repeatedly present their predictions of whether or not various events will occur in terms of their subjective probabilities of those events. The ideas we describe here are relevant in almost any situation in which forecasters must repeatedly make such probabilistic predictions, regardless of the particular subject matter or substantive area of the events being forecast. The forecaster might be an economist who at the beginning of each quarterly period makes predictions about unemployment, the rate of inflation, or Gross National Product in that quarter based on the values of various economic indicators; the forecaster might even make predictions using a large-scale econometric model of the United States economy based on hundreds of variables and econometric relations. In a different field, the forecaster might be the weatherman for a television station who at the beginning of each day must announce his probability that it will rain during the day. For ease of exposition, we present our discussions here in the context of such a weather forecaster who day after day must specify his subjective probability x that there will be at least a certain amount of rain at some given location during a specified time interval of the day. We refer to the occurrence of this well-specified event simply as "rain". Thus, at the begin- ning of each day the forecaster must specify his probability of rain and at the end of each day he observes whether or not rain actually occurred. The probability x specified by the forecaster on any particular day is called his prediction for that day. We shall make the realistic, and simultaneously simplifying, assumption that the

The Comparison and Evaluation of Forecasters.

Consider a situation in which it is desired to gain knowledge about the true value of some parameter (or about the true state of the world) by means of experimentation. Let $\Omega$ denote the set of all possible values of the parameter $\theta$, and suppose that the experimenter's knowledge about the true value of $\theta$ can be expressed, at each stage of experimentation, in terms of a probability distribution $\xi$ over $\Omega$. Each distribution $\xi$ indicates a certain amount of uncertainty on the part of the experimenter about the true value of $\theta$, and it is assumed that for each $\xi$ this uncertainty can be characterized by a non-negative number. The information in an experiment is then defined as the expected difference between the uncertainty of the prior distribution over $\Omega$ and the uncertainty of the posterior distribution. In any particular situation, the selection of an appropriate uncertainty function would typically be based on the use to which the experimenter's knowledge about $\theta$ is to be put. If, for example, the actions available to the experimenter and the losses associated with these actions can be specified as in a statistical decision problem, then presumably the uncertainty function would be determined from the loss function. In Section 2 some properties of uncertainty and information functions, and their relation to statistical decision problems and loss functions, are considered. In Section 3 the sequential sampling rule whereby experiments are performed until the uncertainty is reduced to a preassigned level is studied for various uncertainty functions and experiments. This rule has been previously studied by Lindley, [8], [9], in special cases where the uncertainty function is the Shannon entropy function. In Sections 4 and 5 the problem of optimally choosing the experiments to be performed sequentially from a class of available experiments is considered when the goal is either to minimize the expected uncertainty after a fixed number of experiments or to minimize the expected number of experiments needed to reduce the uncertainty to a fixed level. Particular problems of this nature have been treated by Bradt and Karlin [6]. The recent work of Chernoff [7] and Albert [1] on the sequential design of experiments is also of interest in relation to these problems.

/pdf/uncertainty-information-and-sequential-experiments-ciebrbv9an.pdf

Morris H. DeGroot

Papers

Optimal Statistical Decisions

Reaching a Consensus

Probability and statistics

The Comparison and Evaluation of Forecasters.

Uncertainty, Information, and Sequential Experiments