Odewahn, S. C., Stockwell, E. B., Pennington, R. L., Humphreys, R. M., & Zumach, W. A. (1992). Automated star/galaxy discrimination with neural networks. Astronomical Journal, 103, 318 -- 331.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....and such that m n. Typically, m 10n. We also demonstrate the e#ectiveness of the LPN algorithm by testing it and comparing it with CPLEX on standard classification test problems: four from the University of California Machine Learning Repository [27] and two publicly available datasets [28]. 4.1 Very large synthetic datasets We first test LPN on very large synthetically generated test problems. To display the simplicity of LPN we give two MATLAB m file codes below. The first, lpgen , is a linear programming test problem generator. The second, lpnewt1 is an implementation of ....

....format short e; epsi delta tol i 1 toc1 toc2 norm(x y,inf) norm(x z,inf) 4. 2 Machine learning test problems In this section we test and compare LPN with CPLEX on four classification datasets from the UCI Machine Learning Repository [27] and two publicly available datasets [28]. Again, we use LOCOP2. 12 [5] oom denotes out of memory . LPN Time is total time, i.e. toc2 from Code 4.2 Problem Size Density Time Iter. Accuracy Time Accuracy z## Seconds r## 0.05 1041.8 26 1.1 0.05 787.0 26 8.8 0.1 840.5 14 5.8 10 14 28716.6 7.0 1.0 228.7 15 ....

S. Odewahn, E. Stockwell, R. Pennington, R. Humphreys, and W. Zumach. Automated star/galaxy discrimination with neural networks. Astronomical Journal, 103(1):318--331, 1992.

....The structure of this dataset with very large n and (m n) results from the DNA microarray dataset used. Hence, feature selection is very desirable in such high dimensional problems. Other tests and comparisons were also carried out on six moderately dimensioned, publicly available datasets [21, 22] and are described Section 4.2. 4.1 Multiple Myeloma dataset Multiple Mycloma is cancer of the plasma cell. The plasma cell normally produces antibodies that destroy foreign bodies such as bacteria. As a product of the Myeloma disease the plasma cells get out of control and produce a tumor. These ....

....of other methods that do not perform feature selection. We tested our algorithm on sLx publicly available datasets. Five from the UCI Machine Learning Repository [21] Iono sphere, Cleveland Heart, Pima Indians, BUPA Liver and Housing. The sixth dataset is the Galaxy Dim dataset available at [22]. The dimensionality and size of each dataset is given in Table 2. 4.2.1 Numerical comparisons using a linear classifier In this set of experiments we used a linear classifier to compare our method NLPSVM with LSVM, NSVM and CPLEX SVM on the six datasets mentioned above. Because m n for these ....

S. Odewahn, E. Stockwell, R. Pennington, lq_ Humphreys, and W. Zumach. Automated star/galaxy discrimination with neural networks. Astronomical Journal, 103(1):318331, 1992.

....space features. The three classifiers MSVM (15) SVMII. I1 (8) and FSV [2, Eqn. 8) were tested on seven datasets, the first five of which, WPBC, Ionosphere, Cleveland Heart, Pima Indians, and BUPA Liver are from the Irvine Machine Learning Repository [18] while the Galaxy Dim dataset is from [19], and the Census dataset is a version of the US Census Bureau Adult dataset, which is publicly available from Silicon Graphics website [6] For WPBC(60) 110 breast cancer patients were classified into those who had a recurrence of the disease within 60 months and those who had not. For the ....

S. Odewahn, E. Stockwell, R. Pennington, R. Hummphreys, and W. Zumach. Automated star/galaxy discrimination with neural networks. Astronomical Journal, 103(1):318 331, 1992.

....Cleveland Heart, Pima Indians, BUPA Liver, Mushroom, Tic Tac Toe. The Census dataset is a version of the US Census Bu reau Adult dataset, which is publicly available from the Silicon Graphics website [4] The Galaxy Dim dataset used in galaxy discrimination with neural networks from [30] Two large datasets (2 million points and 10 attributes) created using David Musicant s NDC Data Generator [29] The Spiral dataset proposed by Alexis Wieland of the MITRE Corporation and available from the CMU Ar tificial Intelligence Repository [37] We outline our computational results ....

....In contrast, SVM at [16] failed on this problem [24] 3. Table 3: Comparison of ISVM, SSVM and LSVM and SVM at, using a Linear Classifier In this experiment we compared four methods: PSVM, SSVM, LSVM and SVM at on seven publicly avail able datasets from UCI Machine Learning Repository [28] and [30]. As shown in Table 3, the correctness of the four methods were very similar but the execution time including ten fold cross validation for PSVM was smaller by as much as one order of magnitude or more than the other three methods tested. Since LSVM, SSVM and PSVM are all based on similar formula ....

S. Odewahn, E. Stockwell, R. Pennington, R. Humphreys, and W. Zumach. Automated star/galaxy discrimination with neural networks. Astronomical Journal, 103(1):318 331, 1992.

....dimension 3 once, i.e. using the anisotropic grid c 1;1;2;1;1;1 , we achieve the best result of 73.9 using linear basis functions. We picked this attribute for re nement through cross validation. 3.2. 2 Galaxy Dim The Galaxy Dim data set is a commonly used subset of the data presented in [37]. It consists of 4192 data points with 14 attributes, the two classes have almost the same number of instances. Again no test data is present and therefore we can only report our ten fold cross validation results. Since this data set is in 14 dimensions we use the combination technique of equation ....

S. Odewahn, E. Stockwell, R. Pennington, R. Humphreys, and W. Zumach. Automated star/galaxy discrimination with neural networks. Astronomical Journal, 103(1):318-331, 1992. 22

....set has 4192 examples and the bright data set has 2462 examples. Each example represents a star or a galaxy and is described by 14 numeric attributes. The bright data set is nearly linearly separable. These two data sets are generated from a large set of star and galaxy images collected by Odewahn [22] at the University of Minnesota. Sonar, Mines vs. Rocks The Sonar data set [13] contains sixty real valued attributes between 0.0 and 1.0 used to define 208 mines and rocks. Attributes are obtained by bouncing sonar signals o# a metal cylinder (or rock) at various angles and rises in frequency. ....

S. Odewahn, E. Stockwell, R. Pennington, R Humphreys, and W Zumach. Automated star/galaxy discrimination with neural networks. Astronomical Journal, 103(1):318--331, 1992.

.... by dividing the coordinate axes into six intervals over the range of the data and choosing two centers as the midpoints of the densest and second densest intervals on the axes [27] We utilized three publicly available databases from the UCI ML Repository [119] and one dataset that appears in [126]. The Wisconsin Diagnostic Breast Cancer Dataset (WDBC) consists of 569 instances each having 32 features. Feature 1 is an identification and was discarded. Feature 2 is diagnosis (M = malignant, B = benign) and was used to assign instances to the two datasets A and B. Features 3 32 are the mean, ....

....of 267 instances representing the voting records for Democratic Representatives. Set Bae R 16 consists of 168 instances representing the voting records for Republican Representatives. Each feature has been normalized to have mean = 0, standard deviation = 1. The Star Galaxy Bright dataset [126] consists of the set Aae R 14 having 1505 examples and the set Bae R 14 having 957 examples. Each of the 14 features has been normalized to have mean = 0, standard deviation = 1. Results are summarized in Table 9. We note that for two of the databases the k Median Algorithm 3.1.3 ....

S. Odewahn, E. Stockwell, R. Pennington, R. Hummphreys, and W. Zumach. Automated star/galaxy discrimination with neural networks. Astronomical Journal, 103(1):318--331, 1992.

....constructed using the unlabeled working set. Note that a much larger and clearer separation margin is found. These computational solutions are identical to those presented in [19] We also tested S 3 VM on ten real world data sets (eight from [14] and the bright and dim galaxy sets from [15]) There have been many algorithms applied successfully to these problems without incorporate working set information. Thus it was not clear a priori that S 3 VM would improve generalization on these data sets. For the data sets where no improvement is possible, we would like transduction using ....

S. Odewahn, E. Stockwell, R. Pennington, R Humphreys, and W Zumach. Automated star/galaxy discrimination with neural networks. Astronomical Journal, 103(1):318--331, 1992.

....different real world data sets. In the next section we will consider artificial data, for which the concept definition can be precisely characterized. Description of data sets Star Galaxy Discrimination. Two of our data sets came from a large set of astronomical images collected by Odewahn et al.[367]. In their study, they used these images to 32 Thanks to Kristin Bennett for providing the code, and for helpful discussions. 33 Though LOQO is a commercial product, academic institutions can obtain a free copy for research purposes only. Contact Robert Vanderbi at rvdb jazz.princeton.edu. 61 ....

....of magnitude (brightness) is helpful: human experts classify brighter objects better, so one would expect automated classifiers to do the same. One can also divide objects into brightness ranges, and build separate classifiers for different brightness ranges. The latter approach was taken in [367]. At first, we tried several heuristic, empirically derived rules to filter out the faint objects. The following is an example of the kind of rules we considered: Rule: Retain only the objects that have a peak intensity value 20 in at least one color band. From these, remove objects that have ....

S.C. Odewahn, E.B. Stockwell, R.L. Pennington, R.M. Humphreys, and W.A. Zumach. Automated star-galaxy discrimination with neural networks. Astronomical Journal, 103(1):318--331, 1992.

....In order to study the performance of k DT on larger datasets, we ran several experiments using astronomical image data collected with the University of Minnesota Plate Scanner. This dataset contains several thousand astronomical objects, all of which are classified as either stars or galaxies. Odewahn et al. 1992) used this dataset to train perceptrons and backpropagation networks to differentiate between stars and galaxies. We did not have access to the exact training and test set partitions used by Odewahn et al. so we used a cross validation technique to estimate classification accuracy. The Odewahn et ....

....were generated by averaging 19 ten fold cross validation trials. The astronomy dataset consists of 4164 examples. Each example has fourteen real valued attributes and a label of either star or galaxy. Approximately 35 of the examples are galaxies. Classification results are shown in Table 3. Odewahn et al. 1992) obtained accuracies of 99.7 using backpropagation and 99.4 with a percep13 tron. It appears, however, that their results were generated with a single trial on a single partition into test and training sets. In fact, we obtained a ten fold cross validated accuracy of 99.1 using a perceptron. We ....

Odewahn, S.C., E.B. Stockwell, R.L. Pennington, R.M. Humphreys, and W.A. Zumach, 1992. Automated star-galaxy discrimination with neural networks. Astronomical Journal, 103(1):318--331.

No context found.

Odewahn, S. C., Stockwell, E. B., Pennington, R. L., Humphreys, R. M., & Zumach, W. A. (1992). Automated star/galaxy discrimination with neural networks. Astronomical Journal, 103, 318 -- 331.

No context found.

S. Odewahn, E. Stockwell, R. Pennington, R. Hummphreys, and W. Zumach. Automated star/galaxy discrimination with neural networks. Astronomical Journal, 103(1):318--331, 1992.

No context found.

S. Odewahn, E. Stockwell, R. Pennington, R. Hummphreys, and W. Zumach. Automated star/galaxy discrimination with neural networks. Astronomical Journal, 103(1):318--331, 1992.

No context found.

S. Odewahn et al., "Automated Star/Galaxy Discrimination with Neural Networks," Astronomical J., vol. 103, no. 1, 1992, pp. 318--331.

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