Toward Provably Correct Feature Selection in Arbitrary Domains.
Dimitris MargaritisPublished in: NIPS (2009)
Keyphrases
- provably correct
- feature selection
- situation calculus
- mutual information
- formal methods
- real world
- text classification
- text categorization
- application domains
- multi class
- unsupervised feature selection
- feature set
- feature selection algorithms
- error estimates
- relational structures
- constraint satisfaction
- cross domain
- method for feature selection
- multi task
- information gain
- microarray data
- dimensionality reduction
- feature space
- feature extraction
- knowledge base
- machine learning