Learning

Context

• The Learning menu provides access to a wide range of learning algorithms and related functions:

• The following functions make up the core of the learning menu and are central to BayesiaLab’s machine-learning capabilities.

  • Unsupervised Structural Learning to discover high-dimensional probabilistic relationships in data.
  • Supervised Learning to developing predictive models focused on a Target Node.
  • Clustering to identify latent concepts among nodes and within datasets.

• All learning algorithms employ the Minimum Description Length Score in the search for the best network among all candidate networks.

  • In this context, the Minimum Description Length Score takes into account both the fit of the Bayesian network to the dataset and the structural complexity of the network structure.
  • By minimizing the Minimum Description Length Score, the learning algorithm optimizes the trade-off between fit and complexity.
  • In the Console, you can observe the evolution of the MDL Score as a result of using different learning algorithms.
  • Also, for an existing network, you can obtain its Minimum Description Length Score.

Learning Functions in Detail