Combining information theoretic kernels with generative embeddings for classification.
Manuele BicegoAydin UlasUmberto CastellaniAlessandro PerinaVittorio MurinoAndré F. T. MartinsPedro M. Q. AguiarMário A. T. FigueiredoPublished in: Neurocomputing (2013)
Keyphrases
- information theoretic
- mutual information
- information theory
- theoretic framework
- support vector
- entropy measure
- information bottleneck
- feature space
- classification accuracy
- feature selection
- log likelihood
- information theoretic measures
- jensen shannon divergence
- relative entropy
- feature vectors
- pattern recognition
- unsupervised learning
- text classification
- machine learning
- supervised learning
- decision trees
- multi modality
- vector space
- kernel function
- image classification
- kullback leibler divergence
- minimum description length
- feature maps
- cost sensitive
- dimensionality reduction
- distributional clustering
- jensen shannon