Improving neural classifiers for ATR using a kernel method for generating synthetic training sets.
Roberto Gil-PitaPilar Jarabo AmoresManuel Rosa-ZureraFrancisco López-FerrerasPublished in: NNSP (2002)
Keyphrases
- kernel methods
- training set
- support vector
- svm classifier
- support vector machine
- feature space
- kernel function
- training examples
- training samples
- classification accuracy
- test set
- training data
- neural network
- decision trees
- kernel matrix
- active learning
- data sets
- learning problems
- reproducing kernel hilbert space
- class distribution
- weak classifiers
- feature set
- rademacher complexity
- real world
- learning tasks
- reproducing kernel
- automatic target recognition
- tree kernels
- data dependent
- cross validation
- machine learning
- machine learning algorithms
- learning algorithm
- training set size
- image processing
- feature extraction