Stone, M. (1974). Cross-validation choice and assessment of statistical procedures. Journal Royal of Statistical Society, 36:111--147. Home/Search Document Not in Database Summary Related Articles Check |
This paper is cited in the following contexts:
Towards Robust Model Selection using Estimation and - Approximation Error Bounds (Correct)
....a whole range of numerical procedures for assessing the loss of a hypothesis are based on the so called resampling methods, where the data is used both for training as well as for testing, in order to determine the true behavior. A well known method operating along these lines is cross validation [20]. Having described four major approaches to model selection based on loss estimation, we mention briefly some of the drawbacks of these approaches so as to motivate our proposal below. Consider first the methods based on distribution free upper bounds. These methods are usually very general, ....
M. Stone, "Cross-validation choice and assessment of statistical predictionss", J. Royal. Statzs. Soc, vol. B36, pp. 111-147, 1974.
The Design And Evaluation Of Web Prefetching and Caching Techniques - Davison (2002) (1 citation) (Correct)
....will present and analyze a novel algorithm that satisfies these characteristics and additionally improves on the performance of our previous work [DH97a, DH97b] 3.3.1. 1 Evaluation Criteria In most machine learning experiments that have a single dataset of independent examples, cross validation [Sto74, Sto78, BFOS84] is the standard method of evaluating the performance of an algorithm. When cross validation is inappropriate, partitioning the data into separate training and test sets is common. For sequential datasets, then, the obvious split would have the training set contain the first ....
M. Stone. Cross-validation choices and assessment of statistical predictions. Journal of the Royal Statistical Society, Series B, 36:111--147, 1974.
Prototype Selection and Feature Subset Selection.. - Sierra, Lazkano.. (Correct)
....backpropagation (BP) based Artificial Neural Network[13] and C4.4 classification tree[16] Validation is done by the Leave One Out (LOO) technique, which consists on classifying each case of the database by using the model obtained by the rest of the training cases. In other words, it is the Stone[20] X Fold Cross validation method in which the X equals the number of cases of the training database, 107 in our case. This validation technique is applied usually when the size of the database is minor than 1000. Table 2. Well classified cases obtained using all the cases and all the variables, ....
M. Stone (1974): Cross-validation choice and assessment of statistical procedures. Journal Royal of Statistical Society 36, 111-147.
Estimating the Genetic Architecture of Quantitative Traits - Zeng, Kao, Basten (1999) (Correct)
.... C p method (Mallows, 1973) the Bayes information criterion (BIC) Schwartz, 1978 ; Hannan Quinn, 1979) the final prediction error method (Shibata, 1984) the generalized information criterion (Rao Wu, 1989) and its analogues (Potscherm, 1989) the delete one cross validation (Allen, 1974 ; Stone, 1974), the generalized cross validation (Craven Wahba 1979) the delete d cross validation (Shao, 1993 ; Geisser, 1975 ; Burman, 1989 ; Zhang, 1993) the bootstrap model selection (Shao, 1996) and the minimizing posterior predictive loss (Gelfand Ghosh, 1998) As pointed out by Broman (1997) QTL ....
Stone, M. (1974). Cross-validation choices and assessment of statistical predictions. Journal of the Royal Statistical Society, Series B 36, 111--147.
Constructing New Attributes for Decision Tree Learning - Zheng (1996) (3 citations) (Correct)
....pruning and reduced error pruning use this method. However, the problem is that some available data must be reserved for pruning, so the original tree can only be built using a smaller subset. If the dataset is not large, this can lead to an inferior tree. The v fold cross validation method [Stone, 1974; Breiman et al. 1984] can help to mitigate this problem but with a drawback that v trees, instead of one tree, must be built. Another type of tree evaluation method is to use only the training set that is used to build the raw tree. However, the resubstitution error (on the training set) is not ....
....test sets. ffl In the Monks domains, since the fixed training set and test set are given for each problem by the problem designers, we follow this methodology and run experiments once on the given training set and test set for each domain. ffl In the real world domains, 10 fold cross validations [Stone, 1974; Breiman, Friedman, Olshen, and Stone, 1984; Weiss and Kulikowski, 1990; Zhang, 1992; Bailey and Elkan, 1993] are conducted on the entire datasets. Each experimental result is 11 The time complexities of algorithms in some, not all, domains are reported, as the time issue is not the main concern ....
M. Stone, Cross-validation choice and assessment of statistical predictions. Journal of the Royal Statistical Society, 36, 111-147.
Using Bayesian networks in the construction of a.. - Sierra, Serrano.. (2000) (Correct)
....of cases, we learn the model for each of the level 0 classifiers, and then we learn the structure of the Bayesian network that maximizes the performance as classifier system. 5 Experimental Results In order to give a real perspective of applied methods, we use 10 Fold Cross validation (Stone [40]) in all the experiments. The data has been collected at the ICU service of the Canary Islands University Hospital, and have been used by Serrano et al. 37] in another kind of medical experiments. 5.1 Datafile The datafile used in this experimentation contain data about ICU patients at Canary ....
M. Stone (1974): "Cross-validation choice and assessment of statistical procedures". Journal Royal of Statistical Society 36, 111-147. 21
Bayesian Networks, Rule Induction and Logistic.. - Larrañaga, .. (2000) (Correct)
....Rule Induction and Bayesian Networks in order to compare their performance in a real problem of Oncology. In order to avoid obtaining an estimate too optimistic of the percentage of wellclassi ed for each model, this parameter has been estimated with the 10 fold cross validation method [26]. While both, the Logistic Regression approach as well as the Rule Induction algorithm used CN2 are well known, there are several new approaches that are related with the automatic learning of the Bayesian Networks. In these approaches, the search of the best structures of the Bayesian ....
....95.0 , 80.0 and 69.0 . 6 Experimental Results In order to give a proper estimate of the accuracy of a classi er, several methods train and test, bootstrap, cross validation are the the usually employed. In this case a variant of the last has been used, the the 10 fold cross validation [26]. In this method the cases are randomly divided into 10 mutually exclusive test partitions of approximately equal size. The cases thsat are not found in each test partition that are are used for training and the resulting classi er is tested on the corresponding test partition. The average error ....
Stone, M. (1974) Cross-validation choice and assessment of statistical procedures. Journal of Royal Statistical Society, 36 pp. 111-147. 13
Learning Bayesian Networks by Genetic Algorithms.. - Larranaga.. (1997) (1 citation) (Correct)
....given a file with cases, the a posteriori most probable structure. The second model used is the called Naive Bayes. This model assumes independence among predictor variables. In both models the estimations of the rate of well classified individuals have been obtained using 10 fold cross validation[26]. The propagation of the evidence has been done using the software HUGIN [1] Model I. The a posteriori most probable structure. CH GA. Figures 7, 8 and 9 show the FIGURE 7. The a posteriori most probably structure for the one year case. FIGURE 8. The a posteriori most probably structure for the ....
Stone, M. (1974): "Cross-validation choice and assessment of statistical procedures". Journal Royal of Statistical Society, vol 36, pp. 111-147.
Visualization and Implementation of Feedforward Neural Networks - Klinke, Grassmann (1996) (1 citation) (Correct)
....get E d = E S d X ij w 2 ij with w ij the weights in the network in order to penalize large weights during the training process. Before applying this regularization method the input variables should be normalized. One way to choose the weight decay parameter d is by cross validation (CV) (Stone 1974, Efron Tibshirani 1993) that is, by minimizing an estimate of the generalization ability with respect to the following algorithm. We divide the training set D in a specified number of subsets D j of sizes N j ffl D = D j ; D j = P N j = N , train the network for all but one of these ....
Stone, M. (1974). Cross-validation choice and assessment of statistical predictions, Journal of the Royal Statistical Society pp. 111--147.
Predicting the Survival in Malignant Skin Melanoma Using .. - Sierra, Larrañaga (1998) (5 citations) (Correct)
....and again we use the well classified percentage as goal function. The fourth model is the so called Naive Bayes. This model assumes independence among predictor variables. In both models the estimations of the rate of well classified individuals have been obtained using 10 fold cross validation[19]. The propagation of the evidence has been done using the HUGIN software [1] Model I. The a posteriori most probable structure. CH GA. Figure 6 shows the structure of the Bayesian Network induced by the Genetic Algorithm. It corresponds to the predictions of survival after one year of being ....
M. Stone, Cross-validation choice and assessment of statistical procedures, in Journal Royal of Statistical Society 36 (1974) 111-147
A New Notion of Optimality for Model Selection with.. - Ratsaby, Meir, Maiorov (Correct)
....to circumvent this problem several schemes have been proposed in the literature. These can broadly be described as follows: i) Resampling methods, where the loss is directly estimated through some kind of resampling of the data. Two well known methods belonging to this class are cross validation [19] and bootstrap [4] ii) Distribution free upper bounds, where an upper bound on the loss is determined using some form of the uniform law of large numbers [24] iii) Asymptotic expansions, which usually rely on some form of asymptotic normality of estimators, which then permit the exact ....
M. Stone. (1974). Cross-validation choice and assessment of statistical predictions. J. Royal. Statis. Soc, vol. B36, pp. 111-147.
Towards Robust Model Selection using Estimation and.. - Ratsaby, Meir, Maiorov (1996) (Correct)
....a whole range of numerical procedures for assessing the loss of a hypothesis are based on the so called resampling methods, where the data is used both for training as well as for testing, in order to determine the true behavior. A well known method operating along these lines is cross validation [20]. Having described four major approaches to model selection based on loss estimation, we mention briefly some of the drawbacks of these approaches so as to motivate our proposal below. Consider first the methods based on distribution free upper bounds. These methods are usually very general, ....
M. Stone, "Cross-validation choice and assessment of statistical predictionss", J. Royal. Statis. Soc, vol. B36, pp. 111-147, 1974.
BAYES-NEAREST: a new Hybrid Classifier Combining Bayesian.. - Lazkano, Sierra (2003) (Correct)
No context found.
Stone, M. (1974). Cross-validation choice and assessment of statistical procedures. Journal Royal of Statistical Society, 36:111--147.
Edited Naive Bayes - Martnez-Otzeta Sierra Lazkano (Correct)
No context found.
Stone, M. (1974). Cross-validation choice and assessment of statistical procedures. Journal Royal of Statistical Society, 36:111--147.
K Nearest Neighbor Edition to Guide Classification.. - Martinez-Otzeta.. (Correct)
No context found.
Stone, M. (1974). Cross-validation choice and assessment of statistical procedures. Journal Royal of Statistical Society, 36:111--147.
Prototype Selection and Feature Subset - Selection By Estimation (Correct)
No context found.
M. Stone (1974): Cross-validation choice and assessment of statistical procedures. Journal Royal of Statistical Society 36, 111-147.
K Nearest Neighbor Edition to Guide Classification .. - Martinez-Otzeta.. (Correct)
No context found.
M. Stone. Cross-validation choice and assessment of statistical procedures. Journal Royal of Statistical Society, 36:111--147, 1974.
Door crossing behavior for a mobile robot using Bayesian - Networks Elena Lazkano (Correct)
No context found.
M. Stone. Cross-validation choice and assessment of statistical procedures. Journal Royal of Statistical Society, 36:111-147, 1974.
Iterated Probabilistic Weighted k-Nearest Neighbor Algorithm - Martinez-Otzeta, Sierra (Correct)
No context found.
Stone, M.: "Cross-validation choice and assessment of statistical procedures". Journal Royal of Statistical Society 36 (1974) 111-147
Experimental Evaluation of Tree-Based Algorithms - For Intonational Breaks (2005) (Correct)
No context found.
Stone, M.: Cross-validation choice and assessment of statistical predictions, Journal of the Royal Statistical Society, 36, 111-147, 1974.
Center for Automated Learning and Discovery - Advisor Manuela Veloso (Correct)
No context found.
M. Stone. Cross-validation choice and assessment of statistical procedures. Journal Royal of Statistical Society, B(36):111--147, 1974.
Experimental Evaluation of Tree-Based Algorithms - For Intonational Breaks (Correct)
No context found.
Stone, M.: Cross-validation choice and assessment of statistical predictions, Journal of the Royal Statistical Society, 36, 111-147, 1974.
Iterated Probabilistic Weighted k-Nearest Neighbor Algorithm - Martinez-Otzeta, Sierra (2003) (Correct)
No context found.
Stone, M.: "Cross-validation choice and assessment of statistical procedures". Journal Royal of Statistical Society 36 (1974) 111-147
Prototype Selection and Feature Subset - Selection By Estimation (Correct)
No context found.
M. Stone (1974): Cross-validation choice and assessment of statistical procedures. Journal Royal of Statistical Society 36, 111-147.
Door crossing behavior for a mobile robot using Bayesian.. - Lazkano, Sierra, Zarauz (Correct)
No context found.
M. Stone. Cross-validation choice and assessment of statistical procedures. Journal Royal of Statistical Society, 36:111-147, 1974.
Combining Nearest Neighbor Classifiers Versus Cross-Validation.. - Paik, Yang (2004) (Correct)
No context found.
Stone, M. (1974) Cross-validation choices and assessment of statistical predictions, J. Roy. Statist. Soc Ser B, 36, 111--147.
k. Results indicate that this procedure is very effective .. - Banded Sinusoidal Tasks (1994) (Correct)
No context found.
M. Stone. Cross-validation choice and assessment of statistical predictions. Journal of the Royal Statistical Society, 36:111--147, 1974. 150
Prior Information and Generalized Questions - Lemm (1996) (Correct)
No context found.
Stone, M. (1974) Cross--validation choice and assessment of statistical predictions. Journal of the Royal Statistical Society B 36, 111-147.
Prior Information and Generalized Questions - Lemm (1996) (Correct)
No context found.
Stone, M. (1974) Cross--validation choice and assessment of statistical predictions. Journal of the Royal Statistical Society B 36, 111-147.
Regularisation in the Selection of Radial Basis Function Centres - Orr (1995) (6 citations) (Correct)
No context found.
M. Stone. Cross-validation choice and the assessment of statistical predictions. Journal of the Royal Statistical Society (B), 36:111--147, 1974.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC