A framework for increasing the value of predictive data-driven models by enriching problem domain characterization with novel features.
Sérgio MoroPaulo CortezPaulo RitaPublished in: Neural Comput. Appl. (2017)
Keyphrases
- data driven
- probabilistic model
- discriminative models
- spatial constraints
- main contribution
- model driven
- multiple features
- modeling framework
- predictive power
- domain specific
- prior knowledge
- domain models
- domain independent
- key features
- feature set
- feature extraction
- machine learning approaches
- image segmentation
- image retrieval
- conceptual models
- abstraction levels
- machine learning algorithms
- generative model
- rich set
- mathematical framework
- predictive features
- cross domain
- classification models
- conceptual model
- statistical models
- process model
- computational model
- domain knowledge
- feature vectors
- multiscale
- training data