I am working on Bayesian networks and I tried the search algorithms like K2, hillclimber ,TAN, etc., but I don't know which is the best. Any suggestions? How could Bayesian networks work as classifiers?
If your problem is to build a Bayesian network starting from a learning set, probably Greedy Thick Thinning and Bayesian Search are the best algorithm around. You will find both implementations in the Decision Systems Lab's C++ library SMILE: http://genie.sis.pitt.edu/
There are also several evolutionary algorithms applied to the problem (such as K2GA). To have a broader idea, you could try to search for "Bayesan network structure learning" on Google Scholar.
As for using Bayesian networks as classifiers, you might simply have a variable expressing the probability of belonging to a certain class, and infer the conditioned probability with the information in the considered sample; then assign the sample to the class with the highest probability.
As for using Bayesian networks as classifiers, one might simply have a variable expressing the probability of belonging to a certain class, and infer the conditioned probability with the information in the considered sample; then assign the sample to the class with the highest probability.Since Adaptive Decision boundary algorithm is one finding a decision boundary which separates any two classes and runs through entire data set ,it is a training algorithm which finds true decision boundary as far as I know
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Algorithms