Results 1  10
of
297
Boosting with the L_2Loss: Regression and Classification
, 2001
"... This paper investigates a variant of boosting, L 2 Boost, which is constructed from a functional gradient descent algorithm with the L 2 loss function. Based on an explicit stagewise re tting expression of L 2 Boost, the case of (symmetric) linear weak learners is studied in detail in both regressi ..."
Abstract

Cited by 208 (17 self)
 Add to MetaCart
regression and twoclass classification. In particular, with the boosting iteration m working as the smoothing or regularization parameter, a new exponential biasvariance trade off is found with the variance (complexity) term bounded as m tends to infinity. When the weak learner is a smoothing spline
ETH Zurich
, 2002
"... This paper investigates a computationally simple variant of boosting, L2Boost, which is constructed from a functional gradient descent algorithm with the L2loss function. As other boosting algorithms, L2Boost uses many times in an iterative fashion a prechosen tting method, called the learner. Bas ..."
Abstract
 Add to MetaCart
. Based on the explicit expression of retting of residuals of L2Boost, the case with (symmetric) linear learners is studied in detail in both regression and classication. In particular, with the boosting iteration m working as the smoothing or regularization parameter, a new exponential biasvariance
Biasvariance error bounds for temporal difference updates
 In COLT
, 2000
"... Abstract We give the first rigorous upper bounds on the error of temporal difference ( ) algorithms for policy evaluation as a function of the amount of experience. These upper bounds prove exponentially fast convergence, with both the rate of convergence and the asymptote strongly dependent on the ..."
Abstract

Cited by 16 (1 self)
 Add to MetaCart
on the length of the backups or the parameter . Our bounds give formal verification to the longstanding intuition that methods are subject to a "biasvariance" tradeoff, and they lead to schedules for and that are predicted to be better than any fixed values for these parameters. We give preliminary
Averaged LeastMeanSquares: BiasVariance Tradeoffs and Optimal Sampling Distributions
"... We consider the leastsquares regression problem and provide a detailed asymptotic analysis of the performance of averaged constantstepsize stochastic gradient descent. In the stronglyconvex case, we provide an asymptotic expansion up to explicit exponentially decaying terms. Our analysis lea ..."
Abstract
 Add to MetaCart
sis leads to new insights into stochastic approximation algorithms: (a) it gives a tighter bound on the allowed stepsize; (b) the generalization error may be divided into a variance term which is decaying as O(1/n), independently of the stepsize γ, and a bias term that decays as O(1/γ2n2); (c) when
LeastSquares λ Policy Iteration: BiasVariance Tradeoff in Control Problems
"... In the context of large space MDPs with linear value function approximation, we introduce a new approximate version of λPolicy Iteration (Bertsekas & Ioffe, 1996), a method that generalizes Value Iteration and Policy Iteration with a parameter λ ∈ (0,1). Our approach, called LeastSquares λ Pol ..."
Abstract

Cited by 11 (5 self)
 Add to MetaCart
iteration or to know a model of the MDP. We provide a performance bound that shows the soundness of thealgorithm. Weshowempiricallyonasimple chain problem and on the Tetris game that this λ parameter acts as a biasvariance tradeoff that may improve the convergence and the performance of the policy
Author manuscript, published in "International Conference on Machine Learning (2010)" LeastSquares λ Policy Iteration: BiasVariance Tradeoff in Control Problems
, 2010
"... In the context of large space MDPs with linear value function approximation, we introduce a new approximate version of λPolicy Iteration (Bertsekas & Ioffe, 1996), a method that generalizes Value Iteration and Policy Iteration with a parameter λ ∈ (0,1). Our approach, called LeastSquares λ Pol ..."
Abstract
 Add to MetaCart
iteration or to know a model of the MDP. We provide a performance bound that shows the soundness of thealgorithm. Weshowempiricallyonasimple chain problem and on the Tetris game that this λ parameter acts as a biasvariance tradeoff that may improve the convergence and the performance of the policy
Bayesian Exponentially Tilted Empirical Likeliood
 Biometrika
, 2005
"... Newey and Smith (2001) have recently shown that Empirical Likelihood (EL) exhibits desirable higherorder asymptotic properties, namely, that its O ¡ n −1 ¢ bias is particularly small and that biascorrected EL is higherorder efficient. Although EL possesses these properties when the model is correc ..."
Abstract

Cited by 23 (1 self)
 Add to MetaCart
. Since ET does not share the higherorder asymptotic properties of EL, there is a need for an estimator that combines the qualities of both estimators. This paper introduces a new estimator called Exponentially Tilted Empirical Likelihood (ETEL) that is shown to have the same O ¡ n −1 ¢ bias and the same
Variable Length Markov Chains
 Annals of Statistics
, 1999
"... We study estimation in the class of stationary variable length Markov chains (VLMC) on a finite space. The processes in this class are still Markovian of higher order, but with memory of variable length yielding a much bigger and structurally richer class of models than ordinary higher order Markov ..."
Abstract

Cited by 134 (5 self)
 Add to MetaCart
power by finding a better tradeoff between model bias and variance and allows better structural description which can be of specific interest. The latter is exemplified with some DNA data. A version of the treestructured context algorithm, proposed by Rissanen (1983) in an information theoretical set
Target Independent Variance/Ambiguity for the Linear Opinion Pool and the Connection to the Exponential Family
"... The decomposition of a generalization error into a bias and a variance term is of great importance. So is the decomposition of error functions into an error and an ambiguity term. A fact not generally recognized is that these two decomposition are intimately connected. For both decompositions a cent ..."
Abstract
 Add to MetaCart
central concept is the \average predictor". In this paper we will discuss the connection between the bias/variance decomposition and the error/ambiguity decomposition. We will describe two decompositions of error functions. The rst decomposition gives target independent variance for all error
"BiasVariance" Error Bounds for Temporal Difference Updates
"... We give the first rigorous upper bounds on the error of temporal difference (td) algorithms for policy evaluation as a function of the amount of experience. These upper bounds prove exponentially fast convergence, with both the rate of convergence and the asymptote strongly dependent on the lengt ..."
Abstract
 Add to MetaCart
on the length of the backups k or the parameter . Our bounds give formal verification to the longstanding intuition that td methods are subject to a "biasvariance" tradeoff, and they lead to schedules for k and that are predicted to be better than any fixed values for these parameters. We give
Results 1  10
of
297