A novel neural approach for unsupervised change detection using SOM clustering for pseudo-training set selection followed by CSOM classifier.
Victor-Emil NeagoeAlexandru-Ioan CiureaLorenzo BruzzoneFrancesca BovoloPublished in: IGARSS (2014)
Keyphrases
- training set
- self organizing maps
- k means
- training data
- random selection
- training samples
- neural network
- kohonen self organizing maps
- feature space
- classification algorithm
- unsupervised learning
- cluster analysis
- input space
- kohonen self organizing map
- test set
- clustering algorithm
- supervised learning
- svm classifier
- support vector machine
- decision trees
- class labels
- training examples
- som neural network
- neural gas
- mixed data
- clustering method
- input data
- network architecture
- neural classifier
- data sets
- nearest neighbor
- cross validation
- training dataset
- pattern recognition
- high dimensional
- classification accuracy
- learning algorithm
- selection criterion
- change detection
- k nearest
- weak classifiers
- test data
- positive and negative examples
- decision boundary
- data clustering
- active learning
- training set size
- clustering procedure
- image analysis
- feature vectors
- feature selection