BayesiaLab 9.0: New Features & Updates (12/2019)
Here is a small selection of new or updated features released in BayesiaLab 9.0:
- The Target and Function node optimization tools are enhanced with new options and outputs.
- The new Most Relevant Explanations function provides precise and concise explanations for your current set of evidence.
- By setting Structural Priors, you incorporate any available, partial prior knowledge about a structure.
- You can improve the quality of machine-learned models with BayesiaLab's new Smoothed Bootstrapping algorithm, Data Perturbation. It perturbs the sample data not only with the weight of each particle but also with the overall Structural Coefficient.
- You can automatically estimate Structural Priors via Resampling/Bagging. This is particularly powerful for small data sets as you no longer have to search for the best Structural Coefficient.
- As a step toward learning causal Bayesian networks, you can induce a Partial Order among your variables via Resampling/Bagging.
- The Markov Blanket Learning Algorithms can now take into account constraints on arc directionality as defined by Temporal Indices or Forbidden Arcs.
- The cross-validation of Variable Clustering now features Purities to estimate to quality of the Factors.
- The Code Export function (optional), can now produce Python code. This code, when embedded into your own program, can compute the posterior probability of a Target node given its Markov Blanket.
- By separate subscription, a new Media tool gives you access to presentation slides and recorded videos of the 3-Day Introductory and Advanced BayesiaLab Courses.
- In the 3D Mapping tool, you can now apply textures to nodes and use auto-rotate for creating visually appealing animations of your Bayesian network models.
BayesiaLab 9.0 presented at the 7th BayesiaLab Conference