References in periodicals archive ?
The black line is LASSO-LDA, while red and blue ones correspond to LSVM and RSVM, respectively.
Course Classifier Accuracy Precision Recall F-score DSAA LDA 99.6 50.0 88.9 64.0 LR 99.5 65.0 76.5 70.3 LSVM 99.7 64.3 81.8 72.0 Figure 3: Learning attitude analysis.
In particular, the traditional vehicle detection algorithm were the Haar and Adaboost classifier , the HOG and LSVM classifier , and the Haaris and SIFT algorithm .
Method The number of TPR (%) Time (ms) correctly identified vehicles Haar and Adaboost 256 70.9 13.7 HOG and LSVM 309 85.6 12.5 Haaris and SIFT 281 77.8 10.1 Single DPM 293 81.1 9.2 Ours 342 94.7 8.1
LSVM gives 96.71% classification accuracy after dimensional reduction by SAE 8-7.
And LSVM, RSVM, and CART show much more stability in classification accuracy.
Classifier Classification Max Mean Min accuracy (%) KELM ACC 83.23 71.32 68.49 LSVM ACC 81.05 64.21 44.21 MSVM ACC 84.21 61.98 43.16 RSVM ACC 85.26 74.34 69.47 CART ACC 89.47 73.95 58.95 KNN ACC 90.53 82.76 76.84 LDA ACC 87.37 69.61 53.68 NB ACC 75.79 69.74 61.05 TABLE 4: Results of comparative classifiers with SAE on Oxford Dataset.
Once the features were defined, the classification scheme was selected; two binary linear SVM (LSVM) coupled in cascade were implemented: the former to classify absent versus present dyssynchrony patterns and the latter (for cases with dyssynchrony) to distinguish mild or moderate-severe dyssynchrony.
On iLV dyssynchrony, for the first LSVM, we had balanced classes with absent/present dyssynchrony subjects (45%/55%); however for the second SVM a weighted linear SVM (WLSVM) was used due to an unbalance in mild/moderate to severe cases (15%/85%).
Cross-validation was used in order to evaluate the LSVM performances; it consisted of dividing the training data into subgroups and alternating them as training and testing sets to finally compute the average of classification results .
Acronyms browser ?
Full browser ?