Cestnik, B.: Estimating probabilities: A crucial task in machine learning. In Proceedings of the European Conference on Artificial Intelligence (1990) 147-149 Home/Search Document Not in Database Summary Related Articles Check |
This paper is cited in the following contexts:
First 50 documents Next 50
Ai Miei Genitori Ii - Acknowledgements First Would (Correct)
....in which both n (C) and n(C) are 0, the probability is 2 , which reflects the fact that an empty training set can not alter our a priori assumptions that positive and negative examples have the same probability. However, this assumption is rarely true in practice. Therefore the m estimate [Ces90] was introduced that takes into account as well the prior probabilities of the classes: C) m Theta p a ( Phi) n(C) m where the prior probability p a ( Phi) can be estimated by the relative frequency of positive examples in the initial training set n . The value of m expresses our ....
B. Cestnik. Estimating probabilities: A crucial task in machine learning. In Proceedings of the Ninth European Conference on Artificial Intelligence, pages 147--149, London, 1990. Pitman.
Improved Class Probability estimates from Decision Tree Models - Margineantu, Dietterich (2002) (3 citations) (Correct)
....total number of training examples that reach the leaf, N k is the number of examples from class k reaching the leaf, K is the number of classes, and # k is the prior for class y k . In their experiments, they set # k uniformly to 1.0 for all k = 1, K. The e#ect of the Laplace correction [18, 8, 3] is to shrink the probability estimates toward 1 K. For example, a leaf containing two examples (both from class k = 1) will estimate P(1 x) 2 1) 2 2) 0.75 when K = 2 instead of 1.0. Finally, they apply an ensemble method known as Bagging [5] Bagging constructs L decision trees and then ....
B. Cestnik. Estimating probabilities: A crucial task in machine learning. In L. C. Aiello, editor, Proceedings of the Ninthe European Conference on Artificial Intelligence, pages 147--149. Pitman Publishing, 1990.
Hierarchical Hidden Markov Models for Information Extraction - Skounakis, Craven, Ray (2003) (4 citations) (Correct)
....all punctuation. The same preprocessing is done on test sentences, with the exception that words that were not encountered in the training set are mapped to the OUT OF VOCAB token. The vocabulary is the same for all emitting states in the models, and all parameters are smoothed using m estimates [Cestnik, 1990] . We train all models using the discriminative training procedure referred to in Section 3 [Krogh, 1994] To evaluate our models we construct precision recall graphs. Precision is defined as the fraction of correct tuple instances among those instances that are extracted by the model. Recall is ....
B. Cestnik. Estimating probabilities: A crucial task in machine learning. In Proc. of the 9th European Conf. on Artificial Intelligence, pages 147--150, Stockholm, Sweden, 1990. Pitman.
A Comparative Study of Discretization Methods for Naive-Bayes.. - Yang, Webb (Correct)
....number of values. So does the class label. Accordingly p(C = c) and p(X i = x i C = c) can be estimated with reasonable accuracy from the frequency of instances with C = c and the frequency of instances with X i = x i C = c in the training data. In our experiment: The Laplace estimate [6] was used to estimate p(C = c) n c k N nk , where n c is the number of instances satisfying C = c, N is the number of training instances, n is the number of classes and k = 1. The M estimate [6] was used to estimate p(X i = x i C = c) n ci mp n c m , where n ci is the number of ....
....with X i = x i C = c in the training data. In our experiment: The Laplace estimate [6] was used to estimate p(C = c) n c k N nk , where n c is the number of instances satisfying C = c, N is the number of training instances, n is the number of classes and k = 1. The M estimate [6] was used to estimate p(X i = x i C = c) n ci mp n c m , where n ci is the number of instances satisfying X i = x i #C = c, n c is the number of instances satisfying C = c, p is the prior probability p(X i = x i ) estimated by the Laplace estimate) and m = 2. When a continuous numeric ....
Cestnik, B. Estimating probabilities: A crucial task in machine learning. In Proceedings of the European Conference on Artificial Intelligence (1990), pp. 147-- 149.
Analyzing Attribute Dependencies - Jakulin, Bratko (2003) (Correct)
....2 way interaction gains. Consequently, positive interaction gains may in small domains indicate only a coincidental regularity. Taking accidental dependencies into consideration is a well known cause of overfitting, but there are several ways of remedying this probability estimation problem, e.g. [15]. 7 Conclusion In this paper we studied the detection and resolution of dependencies between attributes in machine learning. First we formally defined the degree of interaction between attributes through the deviation of the best possible voting classifier from the true relation between the ....
Cestnik, B.: Estimating probabilities: A crucial task in machine learning. In: Proc. 9th European Conference on Artificial Intelligence. (1990) 147--149
Deriving Biased Classifiers for Better ROC Performance - Blockeel, Struyf (Correct)
....i are dominated by A. For decision trees we propose the following method. Given a tree with positive and negative predictions, order the leaves according to some estimated probability of an unseen example sorted into that leaf being positive. This probability estimate could e.g. be the m estimate [7], a linear interpolation between the ratio of positives in the leaf and some a priori probability p, which is given a weight m : e = n pm n n Gamma m , with n and n Gamma the number of positive negative examples in the leaf. As a value for p it makes sense to use the proportion of ....
B. Cestnik. Estimating probabilities: A crucial task in machine learning. In Proceedings of the 9th European Conference on Artificial Intelligence, pages 147--149, London, 1990. Pitman.
Using Constraints in Relational Subgroup Discovery - Zelezny, Lavrac, Dzeroski (Correct)
....to enable handling example weights, which provide the means to consider di#erent parts of the instance space in each iteration of the weighted covering algorithm. In the WRAcc computation all probabilities are computed by relative frequencies. Alternatively, the Laplace [4] and the m estimate [3, 7] could be used to estimate the probabilities. In order to include example weights into WRAcc, the following notation is used. Let n(B) stand for the number of instances covered by a rule H B, n(H) for the number of examples of class H , and n(H.B) for the number of examples correctly ....
....discovery Maximum feature length was set to 8, yielding 276 initial features. An example feature is f99(A) ext number(A,B) prefix(B, 0,4,0,7] meaning that the caller s number starts with 0407. Another feature is f115(A) call date(A,B) call time(A,C) ext number(A,D) prev attempt(B,C,D,[3,1], today,unavailable) meaning that the caller (of the current call) has today tried to reach line 31, which was unavailable. We then set the minimum feature coverage to 20 instances, thus obtaining 138 distinct features. With these features, we use the RSD rule induction algorithm with altered ....
[Article contains additional citation context not shown here]
B. Cestnik. Estimating probabilities: A crucial task in machine learning. In Proceedings of the 9th European Conference on Artificial Intelligence, pages 147--149. Pitman, 1990.
Institut D'investigacio En Intel.ligincia - Artificial Csic The (2003) (Correct)
....in the dataset fulfilling Ai = Si and C cl is zero. To avoid this problem, a softening consisting in assigning a small proba bility instead of zero to Oi, j can be done. That softening can improve the accuracy of the classifier. A set of ad hoc not well founded softening methods have been tried [2, 6]. In [7] Kontkanen et al. propose an approach for Instance Based Learning (IBL) and apply it to the Naive Bayes classifier. This approach is based on the Bayesian 0i,v, and arrive model averaging principle [4] More concretely, they define qbi,v,j to the conclusion that if we accept a ....
Bojan Cestnik. Estimating probabilities: A crucial task in machine learning. In Proceedings of the 9th European Conference on Artificial Intelligence, pages 147149, 1990.
RSD: Relational subgroup discovery through first-order.. - Lavrac, Zelezny, Flach (2002) (2 citations) (Correct)
....H B is equal to the conditional probability of head H, given that the body B is satisfied: Acc(H B) p(H B) The accuracy measure can be replaced by the weighted relative accuracy, defined in Equation 1. Furthermore, di#erent probability estimates, like the Laplace [4] or the m estimate [3, 7], can be used for estimating the above probability and the probabilities in Equation 1. Additionally, a rule learner can apply a significance test to the induced rule. The rule is considered to be significant, if it locates a regularity unlikely to have occurred by chance. To test significance, ....
B. Cestnik. Estimating probabilities: A crucial task in machine learning. In L. Aiello, editor, Proc. of the 9th European Conference on Artificial Intelligence, 147--149. Pitman, 1990.
Knowledge Discovery in Multi-Label Phenotype Data - Clare, King (2001) (8 citations) (Correct)
....also required only to learn a set of accurate rules, not a complete classification. This unusual feature of the data made it necessary for us to develop a complicated resampling approach to estimating rule accuracy based on the bootstrap. All accuracy measurements were made using the m estimate [5] which is a generalisation of the Laplace estimate, taking into account the a priori probability of the class. The m estimate for rule r (M(r) is: M(r) p m P P N p n m where P = total number of positive examples, N = total number of negative examples. p = number of positive ....
B. Cestnik. Estimating probabilities: A crucial task in machine learning. In Proceedings of the Ninth European Conference on Artificial Intelligence (ECAI90), pages 147--149, 1990.
Gene Selection By Sequential Search Wrapper.. - Inza, Sierra.. (2002) (Correct)
....regarding the test instance is found: its class is retrieved to predict this as the class of the test instance. As the variables of microarray datasets have continuous values, the measure of the distance between two samples is performed using the euclidean metric. The Naive Bayes (NB) rule [7] uses the Bayes theorem to predict the class for each case, assuming that the predictive genes are independent given the category. To classify a new sample characterized by d genes X = X 1 ; X 2 ; X d ) the NB classi er applies the following rule: c N B = arg max c j 2C p(c j ) d Y ....
B. Cestnik, `Estimating probabilities: a crucial task in machine learning', in Proceedings of the European Conference on Arti cial Intelligence, pp. 147-149, (1990).
Sensitivity Analysis of the Result in Binary Decision Trees - Alvarez (2004) (Correct)
No context found.
Cestnik, B.: Estimating probabilities: A crucial task in machine learning. In Proceedings of the European Conference on Artificial Intelligence (1990) 147-149
BaseNP Chunking using ILP - Konstantopoulos (Correct)
No context found.
Cestnik, B.(1990), Estimating probabilities: A crucial task in machine learning, European Conference on Artificial Intelligence, pp. 147--149.
An Analysis of Rule Learning Heuristics - Johannes Urnkranz Austrian (Correct)
No context found.
Cestnik, B. (1990). Estimating probabilities: A crucial task in Machine Learning. Proceedings of the 9th European Conference on Artificial Intelligence (ECAI-90) (pp. 147--150). Stockholm, Sweden: Pitman.
Selection of the Most Useful Subset of Genes for Gene.. - Paul, Iba (2004) (Correct)
No context found.
Cestnik, B., "Estimating probabilities: a crucial task in machine learning", in Proceedings of the European Conference on Artificial Intelligence, 1990, pp. 147--149.
Research on the Improvement of Efficiency of EDAs for Optimization - Paul (2004) (Correct)
No context found.
B. Cestnik, Estimating probabilities: A crucial task in machine learning, Proceedings of the European Conference in Artificial Intelligence, 1990, pp. 147--149.
Linear and Combinatorial Optimizations by Estimation of.. - Paul, Iba (2002) (Correct)
No context found.
Cestnik,B.(1990).Estimating probabilities: A crucial task in machine learning .Proceedings of the European Conference in Artificial Intelligence, 147-149.
Identification of Informative Genes for Molecular Classification .. - Paul, Iba (2004) (Correct)
No context found.
Cestnik, B.: Estimating probabilities: a crucial task in machine learning. Proceedings of the European Conference on Artificial Intelligence, 1990, 147--149.
Experimental Evaluation of a Trainable Scribble - Recognizer For Calligraphic (Correct)
No context found.
Cestnik B. Estimating probabilities: A crucial task in machine learning. Proceedings of the Ninth European Conference on Artificial Intelligence (pp. 147-149). London: Pitman, 1990.
Feature Selection and Classifier Ensembles: A Study on.. - Yu (2003) (Correct)
No context found.
B. Cestnik. Estimating probabilities: a crucial task in machine learning. In Proc. of the European Conference on Artificial Intelligence, pages 147--149, Stockholm, Sweden, 1990. 78
Arbitrating Among Competing Classifiers Using Learned Referees - Ortega, Koppel, Argamon (1998) (7 citations) (Correct)
No context found.
B. Cestnik. Estimating probabilities: A crucial task in machine learning. In Proc. Ninth European Conference on Artificial Intelligence, pages 147--149, London, 1990. Pitman.
Comparative Experiments on Learning Information.. - Bunescu, Ge, Kate, al. (2004) (4 citations) (Correct)
No context found.
B. Cestnik, Estimating probabilities: A crucial task in machine learning, in: Proc. of 9th European Conf. on Arti cial Intelligence, Stockholm, Sweden, 1990, pp. 147-149.
Process-Oriented Estimation of Generalization Error - Domingos (1999) (9 citations) (Correct)
No context found.
B. Cestnik. Estimating probabilities: A crucial task in machine learning. In Proceedings of the Ninth European Conference on Artificial Intelligence, pages 147--149, Stockholm, Sweden, 1990. Pitman.
Using Clustering to Boost Text Classification - Fang Parthasarathy Schwartz (Correct)
No context found.
Cestnik, B. 1990. Estimating probabilities: A crucial task in machine learning. Proceedings of the Ninth European Conference on Artificial Intelligence (pp. 147149) . London: Pitman.
Experimental Evaluation of a Trainable Scribble.. - Csar Pimentel.. (Correct)
No context found.
Cestnik B. Estimating probabilities: A crucial task in machine learning. Proceedings of the Ninth European Conference on Artificial Intelligence (pp. 147-149). London: Pitman, 1990.
First 50 documents Next 50
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