B.V. Dasarathy. Nearest Neighbor (NN) norms: NN pattern classi cation techniques. IEEE Computer Society Press, 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....to one and only one class; see Section3 for details) arg max c j Pr(c j ) Y t2d Pr(tjc j ) N(t;d) 2) 2.3.2. K Nearest Neighbor (kNN) kNN, an instance based classi cation method, has been an e ective approach to a broad range of pattern recognition and text classi cation problems [8, 25, 26, 28]. In contrast to eager learning algorithms (including Naive Bayes) which have an explicit training phase before seeing any test document, kNN uses the training documents local to each test document to make its classi cation decision on that document. Our kNN uses the conventional vector space ....

Belur V. Dasarathy. Nearest Neighbor (NN) Norms: NN Pattern Classi cation Techniques. McGraw-Hill Computer Science Series. IEEE Computer Society Press, Las Alamitos, California, 1991.

....this problem. This paper shows experimental results from applying editing and selection methods individually, in a cascade, or simultaneously. Keywords: Nearest Neighbor Classi er (1 nn) Data editing, Feature selection, Heuristic optimization. 1 Introduction The nearest neighbor classi er (1 nn) [1, 3] is intuitive and accurate. According to 1 nn, an input is assigned to the class of its nearest neighbor from a stored labeled reference set. Figure 1 explains how the label of a new objects x 2 2 is assigned. The k nn classi er takes the k neighbors of x and infers the class label for x by ....

....Instead of Z we use a subset S 1 Z, and instead of X we use S 2 X . Figure 1 shows the reduction of Z, both row wise and column wise. So far, the two tasks: selection of representative data subset and the feature selection (dimensionality reduction) have been approached separately, e.g. [1, 2]. In a previous study we used a genetic algorithm for a simultaneous selection [5] and found it very successful. Here we give experimental results with these methods and two more heuristic optimization methods 1 : a stochastic hill climbing and an incremental hill climbing (a version of tabu ....

B.V. Dasarathy. Nearest Neighbor (NN) Norms: NN Pattern Classication Techniques. IEEE Computer Society Press, Los Alamitos, California, 1990.

....Recently, though, a number of researchers have given attention to the instance selection problem. A wide variety of techniques have been proposed for the instance selection technique, but most of them require independent binary or multi class data to evaluate the utility of each instance [44, 45]. Similarly, techniques such as Support Vector Machines locate a minimal set of instances which de ne an optimal decision hyperplane, but again they require binary class data to nd such a hyperplane [46] Lewis and Catlett [47] have examined a selective sampling method employing classi cation ....

B. V. Dasarathy. Nearest Neighbor (NN) Norms: NN Pattern Classication Techniques. IEEE Computer Society Press, Los Alamitos, CA, 1991.

....prediction accuracy. Over the years a variety of di erent classi cation algorithms have been developed by the machine learning community. Examples of such algorithms are decision tree based [1, 26, 25] rule based [4, 5] probabilistic [19] neural networks [8, 34] genetic [13] instance based [9, 35], and support vector machines [31, 32] Depending on the characteristics of the data sets being classi ed certain algorithms tend to perform better than others. In recent years, algorithms based on the support vector machines and the k nearest neighbors have been shown to produce reasonably good ....

B. Dasarathy. Nearest neighbor (NN) norms: NN pattern classication techniques. IEEE Computer Society Press, 1991.

....to the CHAPTER 3. TEXT CATEGORIZATION ALGORITHMS USED 29 classify(t,k) t:test instance, k:number of neighbors [1] begin [2] for each class c [3] vote[c] 0 [4] for each feature f [5] put k nearest neighbors of test instance t on feature f into Bag [6] Bag=kBag(f; t; k) [7] for each class c [8] vote[c] vote[c] count[c,Bag] 9] prediction= UNDETERMINED class 0 [10] for each class c [11] ifvote[c] vote[prediction]then [12] prediction=c [13] return(prediction) 14] end. Figure 3.1: Classi cation in the FPTC Algorithm instance t on feature f , ....

....complexity of this process is O(log m k ) Since m k, the time complexity of kBag is O(log m) The nal classi cation requires the votes of each of n features. Therefore, the classi cation time complexity of the FPTC algorithm is O(n log m) 11] 3. 2 k NN Algorithm The k NN classi er [7] classi er is the basis of many lazy learning algorithm and it is sure that k NN is purely lazy. Purely lazy learning algorithms generally are characterized by three behaviors: 2] 1. Defer: They store all training data and defer processing until queries are given that require reply. 2. Reply: ....

Dasarathy, B. V., Nearest Neighbor(NN) Norms: NN Pattern Classication Techniques, , IEEE Computer Society Press, Los Alamitos, 1991.

....selection process that determines a feature subset from which an accurate FRBCS can be obtained. To increase the eciency while maintaining the e ectiveness of the wrapper feature selection algorithms, our proposal uses the precision estimation provided by the k nearest neighbour rule (k NN) [18], which is very sensitive to the presence of irrelevant characteristics. Before describing the components of the genetic feature selection process, four considerations on the process must be pointed out: The feature selection process follows this outline because it is necessary an ecient ....

B.V. Dasarathy. Nearest neighbour (nn) norms: Nn pattern classication techniques. IEEE Computer Society Press, 1990.

....determines a feature subset of this cardinality with the best behaviour for the classi cation problem to solve. To increase the eciency maintaining the e ectiveness of the wrapper feature selection algorithms the proposal uses the precision estimation provided by the knearest neighbour rule (k NN) [11], that is very sensitive to the presence of irrelevant characteristics. The k NN rule is not sensible to redundant characteristics. The previous determination of the feature subset size realised by lter algorithms that do not use the k NN rule, helps to the wrapper selection to select only ....

B.V. Dasarathy. Nearest neighbour (nn) norms: Nn pattern classication techniques. IEEE Computer Society Press, 1990.

....followed by feature selection, feature selection followed by editing, and the individual results from feature selection and editing. Keywords: Editing for the nearest neighbor classi er (1 nn) Feature selection, Genetic algorithms (GAs) 1 Introduction The nearest neighbor classi er (1 nn) [2, 3] is intuitive and accurate. According to 1 nn, an input is assigned to the class of its nearest neighbor from a stored labeled reference set. The main problem using 1 nn is the signi cant time and memory resources that are required. With the developments of modern computational technology (faster ....

B.V. Dasarathy. Nearest Neighbor (NN) Norms: NN Pattern Classication Techniques. IEEE Computer Society Press, Los Alamitos, California, 1990.

.... [22] some types of Radial Basis Function (RBF) networks Research supported in parts by the NRC COBASE program and ONR Grant # N 00014 96 1 0642 [5] 19] 23] a class of fuzzy if then systems [15] Learning Vector Quantization (LVQ) classi ers [11] 12] 14] edited nearest neighbor rules [6], fuzzy nearest neighbor rules [1] 13] 26] multiple prototype classi ers [2] and a number of neural network implementations of the nearest neighbor design [7] 17] 26] Each of these has speci c strategies and algorithms for nding the prototypes. The diagram in Figure 1 groups the GNPCs ....

B.V. Dasarathy, Nearest Neighbor (NN) Norms: NN Pattern Classication Techniques, Los Alamitos, California: IEEE Computer Society Press, 1990.

No context found.

Belur V. Dasarathy, editor. Nearest Neighbor (NN) Norms: NN Pattern Classi cation Techniques. IEEE Comp. Society Press, Los Alamitos, 1991.

No context found.

D. V. Dasarachy, Nearest Neighbor (NN) Norms: NN Pattern Classi cation Techniques, IEEE Computer Society Press, Los Alamitos, 1991. 14

No context found.

B.V. Dasarathy (Ed): Nearest neighbour (NN) Norms: NN pattern classi cation techniques. IEEE Computer Society Press, 1991.

....[109] 113] There is a vast literature on the subject of nearest neighbor classi cation which will not be reviewed here. The interested reader is referred to the comprehensive treatment by Devroye, Gy or and Lugosi [37] and the collected papers in the 1991 volume edited by Dasarathy [24]. For more on the information measures closely related to the measure P e [1 NN] the reader is referred to Mathai and Rathie [68] see also Toussaint [110] In the past many pattern recognition practicioners have unfairly criticized the NN rule on the grounds of the mistaken assumptions that (1) ....

Belur V. Dasarathy, editor. Nearest Neighbor (NN) Norms: NN Pattern Classi cation Techniques. IEEE Comp. Society Press, Los Alamitos, 1991.

....40 years, this simple classi#cation rule has been intensivelyused in a broad rang of patternrecogy#x## applications. In contrast to its conceptual simplicity, the rule has agzq behavior when applied to non trivial problems. In fact, the k NN rule is asymptotically optimal in the Bayes sense [1]. In other words, the k NN rule performs as well as anyother possible classi#er, provided there is an arbitrarilylarg number of (representative) # This work has been partiallysupported bySpanish projects TIC98 677 C02 02, 1FD97 279, TIC2000 1703 C03 03 and a gy## from theAgykSz Espanola de ....

B.V. Dasarathy(Ed.), Nearest Neigsty (NN) Norms: NN Pattern Classi#cation Techniques, IEEE Computer Society

No context found.

B.V. Dasarathy Nearest Neighbor (NN) Norms: NN Pattern Classication Techniques. Los Alamitos, California: IEEE Computer Society Press, 1990.

No context found.

B.V. Dasarathy. Nearest Neighbor (NN) norms: NN pattern classi cation techniques. IEEE Computer Society Press, 1991.

No context found.

Belur V. Dasarathy. Nearest Neighbor (NN) Norms: NN Pattern Classi cation Techniques. McGraw-Hill Computer Science Series. IEEE Computer Society Press, Las Alamitos, California, 1991.

No context found.

B. V. Dasarathy, editor. Nearest neighbor (NN) norms : NN pattern classi cation techniques. IEEE Computer Society, 1991.

No context found.

B.V. Dasarathy. Nearest Neighbor (NN) norms: NN pattern classi cation techniques. IEEE Computer Society Press, 1991.

No context found.

B.V. Dasarathy (Ed): Nearest neighbour (NN) Norms: NN pattern classi cation techniques. IEEE Computer Society Press, 1991.

No context found.

Dasarathy, B.V., Nearest Neighbor (NN) Norms: NN Pattern Classication Techniques, IEEE Computer Society Press, New York, 1991.

No context found.

B.V. Dasarathy, Nearest Neighbor (NN) Norms: NN Pattern Classication Techniques, (IEEE Computer Society Press, Los Alamitos, Calofornia, 1990).

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