Download
BayesiaLab 4.4

Dynamic Presentations
of BayesiaLab

Static Presentations
of BayesiaLab

News

Search

BayesiaLab

	Presentation	Trial Version Dynamic Presentations Static Presentations   Tutorial   Fact Sheet
	Knowledge Modeling
	Bayesian Network Learning
	Evaluation
	Analysis
	Exploitation
	Action Policy Learning
	Application examples
	Tutorial
	News Letters

From a straightforward learning of probability table, up to learning the network structure:

Unsupervised learning help you discover the whole set of probabilistic relationships existing within the data (association discovery)

BayesiaLab offers three conceptually different search techniques:

- SopLEQ : a quick search based on the whole dataset, and Bayesian Network classes of equivalence
- Taboo : an implementation of the Taboo search
- Taboo Order : a learning technique based on the Taboo search in order to optimize node ordering

Unsupervised learning for new concepts discovery (segmentation, clustering)

- Fixed number of classes: you have to provide the number of classes (or segments) you wish to deal with.
- Automatic selection of the number of classes: we provide the number of classes leading to a best partitioning.
- Refined intitial probabilities: sampling to find the number of classes and to find an initial solution leading to quick convergence.

Supervised learning focussing on one variable

We have several learning algorithms:

- Naive Bayes
- Augmented Naive Bayes
- Sons & Spouses
- Markov Blanket
- Augmented Markov Blanket