scholarly journals Sequential Monte Carlo Methods to Train Neural Network Models

2000 ◽  
Vol 12 (4) ◽  
pp. 955-993 ◽  
Author(s):  
J. F. G. de Freitas ◽  
M. Niranjan ◽  
A. H. Gee ◽  
A. Doucet
Keyword(s):  
Monte Carlo ◽  
Sequential Monte Carlo ◽  
Probability Distributions ◽  
Network Models ◽  
Computational Time ◽  
Optimization Strategy ◽  
Neural Network Models ◽  
Monte Carlo Techniques ◽  
Monte Carlo Algorithms ◽  
Importance Resampling

We discuss a novel strategy for training neural networks using sequential Monte Carlo algorithms and propose a new hybrid gradient descent/sampling importance resampling algorithm (HySIR). In terms of computational time and accuracy, the hybrid SIR is a clear improvement over conventional sequential Monte Carlo techniques. The new algorithm may be viewed as a global optimization strategy that allows us to learn the probability distributions of the network weights and outputs in a sequential framework. It is well suited to applications involving on-line, nonlinear, and nongaussian signal processing. We show how the new algorithm outperforms extended Kalman filter training on several problems. In particular, we address the problem of pricing option contracts, traded in financial markets. In this context, we are able to estimate the one-step-ahead probability density functions of the options prices.

Bayesian Estimation of DSGE Models

2015 ◽  
Author(s):  
Edward P. Herbst ◽  
Frank Schorfheide
Keyword(s):  
Monte Carlo ◽  
Sequential Monte Carlo ◽  
Likelihood Function ◽  
Academic Research ◽  
Dynamic Stochastic General Equilibrium ◽  
Computational Techniques ◽  
Dsge Models ◽  
Monte Carlo Techniques ◽  
Dynamic Stochastic ◽  
Theoretical Foundations

Dynamic stochastic general equilibrium (DSGE) models have become one of the workhorses of modern macroeconomics and are extensively used for academic research as well as forecasting and policy analysis at central banks. This book introduces readers to state-of-the-art computational techniques used in the Bayesian analysis of DSGE models. The book covers Markov chain Monte Carlo techniques for linearized DSGE models, novel sequential Monte Carlo methods that can be used for parameter inference, and the estimation of nonlinear DSGE models based on particle filter approximations of the likelihood function. The theoretical foundations of the algorithms are discussed in depth, and detailed empirical applications and numerical illustrations are provided. The book also gives invaluable advice on how to tailor these algorithms to specific applications and assess the accuracy and reliability of the computations. The book is essential reading for graduate students, academic researchers, and practitioners at policy institutions.

scholarly journals Stochastic Order and Generalized Weighted Mean Invariance

Entropy ◽  
2021 ◽  
Vol 23 (6) ◽  
pp. 662
Author(s):  
Mateu Sbert ◽  
Jordi Poch ◽  
Shuning Chen ◽  
Víctor Elvira
Keyword(s):  
Monte Carlo ◽  
Probability Distributions ◽  
Stochastic Order ◽  
Arithmetic Mean ◽  
Stochastic Orders ◽  
Arithmetic Means ◽  
Increasing Convex Order ◽  
Monte Carlo Techniques ◽  
Multiple Importance Sampling ◽  
Invariance Properties

In this paper, we present order invariance theoretical results for weighted quasi-arithmetic means of a monotonic series of numbers. The quasi-arithmetic mean, or Kolmogorov–Nagumo mean, generalizes the classical mean and appears in many disciplines, from information theory to physics, from economics to traffic flow. Stochastic orders are defined on weights (or equivalently, discrete probability distributions). They were introduced to study risk in economics and decision theory, and recently have found utility in Monte Carlo techniques and in image processing. We show in this paper that, if two distributions of weights are ordered under first stochastic order, then for any monotonic series of numbers their weighted quasi-arithmetic means share the same order. This means for instance that arithmetic and harmonic mean for two different distributions of weights always have to be aligned if the weights are stochastically ordered, this is, either both means increase or both decrease. We explore the invariance properties when convex (concave) functions define both the quasi-arithmetic mean and the series of numbers, we show its relationship with increasing concave order and increasing convex order, and we observe the important role played by a new defined mirror property of stochastic orders. We also give some applications to entropy and cross-entropy and present an example of multiple importance sampling Monte Carlo technique that illustrates the usefulness and transversality of our approach. Invariance theorems are useful when a system is represented by a set of quasi-arithmetic means and we want to change the distribution of weights so that all means evolve in the same direction.

scholarly journals The time machine: a simulation approach for stochastic trees

2011 ◽  
Vol 467 (2132) ◽  
pp. 2350-2368 ◽  
Author(s):  
Ajay Jasra ◽  
Maria De Iorio ◽  
Marc Chadeau-Hyam
Keyword(s):  
Variance Reduction ◽  
Sequential Monte Carlo ◽  
Likelihood Function ◽  
Computing Time ◽  
Point Of View ◽  
Computational Time ◽  
Recent Common Ancestor ◽  
Theoretical Point ◽  
Monte Carlo Techniques ◽  
Most Recent Common Ancestor

In this paper, we consider a simulation technique for stochastic trees. One of the most important areas in computational genetics is the calculation and subsequent maximization of the likelihood function associated with such models. This typically consists of using importance sampling and sequential Monte Carlo techniques. The approach proceeds by simulating the tree, backward in time from observed data, to a most recent common ancestor. However, in many cases, the computational time and variance of estimators are often too high to make standard approaches useful. In this paper, we propose to stop the simulation, subsequently yielding biased estimates of the likelihood surface. The bias is investigated from a theoretical point of view. Results from simulation studies are also given to investigate the balance between loss of accuracy, saving in computing time and variance reduction.

Computational Models in Developmental Psychology

2013 ◽  
pp. 476-504
Author(s):  
Thomas R. Shultz
Keyword(s):  
Neural Networks ◽  
Computational Modeling ◽  
Computational Models ◽  
Developmental Psychology ◽  
Probability Distributions ◽  
Network Models ◽  
Developmental Theory ◽  
Negative Evidence ◽  
Developmental Robotics ◽  
Neural Network Models

Computational modeling implements developmental theory in a precise manner, allowing generation, explanation, integration, and prediction. Several modeling techniques are applied to development: symbolic rules, neural networks, dynamic systems, Bayesian processing of probability distributions, developmental robotics, and mathematical analysis. The relative strengths and weaknesses of each approach are identified and examples of each technique are described. Ways in which computational modeling contributes to developmental issues are documented. A probabilistic model of the vocabulary spurt shows that various psychological explanations for it are unnecessary. Constructive neural networks clarify the distinction between learning and development and show how it is possible to escape Fodor’s paradox. Connectionist modeling reveals different versions of innateness and how learning and evolution might interact. Agent-based models analyze the basic principles of evolution in a testable, experimental fashion that generates complete evolutionary records. Challenges posed by stimulus poverty and lack of negative examples are explored in neural-network models that learn morphology or syntax probabilistically from indirect negative evidence.

scholarly journals Independent Resampling Sequential Monte Carlo Algorithms

2017 ◽  
Vol 65 (20) ◽  
pp. 5318-5333 ◽  
Author(s):  
Roland Lamberti ◽  
Yohan Petetin ◽  
Francois Desbouvries ◽  
Francois Septier
Keyword(s):  
Monte Carlo ◽  
Sequential Monte Carlo ◽  
Monte Carlo Algorithms

scholarly journals A Comparison of Bootstrap and Monte Carlo Approaches to Testing for Symmetry in the Granger and Lee Error Correction Model

2013 ◽  
Vol 5 (5) ◽  
pp. 240-244
Author(s):  
Henry de-Graft Acquah
Keyword(s):  
Monte Carlo ◽  
Monte Carlo Methods ◽  
Large Error ◽  
Small Error ◽  
Small Samples ◽  
Monte Carlo Techniques ◽  
Large Samples ◽  
Monte Carlo Algorithms ◽  
Power Of The Test ◽  
Simulation Results

In this paper, I investigate the power of the Granger and Lee model of asymmetry via bootstrap and Monte Carlo techniques. The simulation results indicate that sample size, level of asymmetry and the amount of noise in the data generating process are important determinants of the power of the test for asymmetry based on bootstrap and Monte Carlo techniques. Additionally, the simulation results suggest that both bootstrap and Monte Carlo methods are successful in rejecting the false null hypothesis of symmetric adjustment in large samples with small error size and strong levels of asymmetry. In large samples, with small error size and strong levels of asymmetry, the results suggest that asymmetry test based on Monte Carlo methods achieve greater power gains when compared with the test for asymmetry based on bootstrap. However, in small samples, with large error size and subtle levels of asymmetry, the results suggest that asymmetry test based on bootstrap is more powerful than those based on the Monte Carlo methods. I conclude that both bootstrap and Monte Carlo algorithms provide valuable tools for investigating the power of the test of asymmetry.

scholarly journals On the convergence rates of some adaptive Markov chain Monte Carlo algorithms

2015 ◽  
Vol 52 (3) ◽  
pp. 811-825
Author(s):  
Yves Atchadé ◽  
Yizao Wang
Keyword(s):  
Monte Carlo ◽  
Markov Chain ◽  
Markov Chain Monte Carlo ◽  
Convergence Rates ◽  
Mixing Time ◽  
Regularity Conditions ◽  
Monte Carlo Algorithms ◽  
Mcmc Algorithms ◽  
Variation Distance ◽  
Importance Resampling

In this paper we study the mixing time of certain adaptive Markov chain Monte Carlo (MCMC) algorithms. Under some regularity conditions, we show that the convergence rate of importance resampling MCMC algorithms, measured in terms of the total variation distance, is O(n-1). By means of an example, we establish that, in general, this algorithm does not converge at a faster rate. We also study the interacting tempering algorithm, a simplified version of the equi-energy sampler, and establish that its mixing time is of order O(n-1/2).

Application of Two Bayesian Filters to Estimate Unknown Heat Fluxes in a Natural Convection Problem

2012 ◽  
Vol 134 (9) ◽  
Author(s):  
Marcelo J. Colaço ◽  
Helcio R. B. Orlande ◽  
Wellington B. da Silva ◽  
George S. Dulikravich
Keyword(s):  
Natural Convection ◽  
Sequential Monte Carlo ◽  
Nonlinear Models ◽  
Probability Distributions ◽  
Heat Fluxes ◽  
Large Set ◽  
Sampling Importance Resampling ◽  
Filter Methods ◽  
Importance Resampling ◽  
Non Gaussian

Sequential Monte Carlo (SMC) or particle filter methods, which have been originally introduced in the beginning of the 1950s, became very popular in the last few years in the statistical and engineering communities. Such methods have been widely used to deal with sequential Bayesian inference problems in the fields like economics, signal processing, and robotics, among others. SMC methods are an approximation of sequences of probability distributions of interest, using a large set of random samples, named particles. These particles are propagated along time with a simple Sampling Importance distribution. Two advantages of this method are: they do not require the restrictive hypotheses of the Kalman filter, and they can be applied to nonlinear models with non-Gaussian errors. This paper uses two SMC filters, namely the SIR (sampling importance resampling filter) and the ASIR (auxiliary sampling importance resampling filter) to estimate a heat flux on the wall of a square cavity encasing a liquid undergoing natural convection. Measurements, which contain errors, taken at the boundaries of the cavity were used in the estimation process. The mathematical model as well as the initial condition are supposed to have some errors, which were taken into account in the probabilistic evolution model used for the filter. Also, the results using different grid sizes and patterns for the direct and inverse problems were used to avoid the so-called inverse crime. In these results, additional errors were considered due to the different location of the grid points used. The final results were remarkably good when using the ASIR filter.

Sign in / Sign up

Export Citation Format

Share Document