Is there a technique for classification, where given a feature vector X = (x_1, x_2,..., x_n)

and a Bayesian network for each class, which for each x_1, x_2,...,x_n there may be a corresponding node in the network. So that I can classify whether the feature vector X belongs to the "class" represented by the network?

Is this sort of strategy common? does it work? It's hard to find something like it on the internet. The motivation behind the strategy is that I want to be able to classify feature vectors even when they have missing features, and also some features only make sense with respect to some classes.

some background: This is an image classification project, where we have many feature extractors (each corresponding to an x_i) but they are costly to run. and some feature extractors only make sense to run on an image only to increase our confidence that that image belongs to a class.

I want to run a feature extractor, then check the confidences in each class and then based on those confidences choose another feature extractor and so on, until I am very sure that the image belongs to a particular class.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.