Child pages
  • BayesiaLab XBL File

Contents

Context

Tools | WebSimulator Editor

Within BayesiaLab's Validation Mode, you can enter different types of evidence (hard evidence, virtual evidence, soft evidence, mean) via the Monitors on a subset of nodes, and then obtain the posterior probability distributions for all other nodes. 

There are numerous options for sharing such a probabilistic expert system with end users:

  • The BayesiaLab Analyst Edition is a version of BayesiaLab that runs exclusively in Validation Mode. Using the Analyst Edition, your research clients can use the network you developed, but they cannot modify it.
  • There is a Markov Blanket Export Module, which allows you to export your network as code in a number of formats, e.g. R, JavaScript, SAS, PHP, VBA, etc. However, the code exported that way is limited to computing the posterior probability distribution of the Target Node only. 
  • You can use the Bayesia Engine API to programmatically access your network to perform inference. This enables you to build your own simulator.

The BayesiaLab WebSimulator is a new solution which allows you to share interactive models with your audience, without having to install any software on the their computer. You can simply publish your network, in its native XBL file format, to the Bayesia server. This generates a dedicated webpage for your network. This web page will automatically utilize the Bayesia Engine API to perform exact inference in your network.

New Feature: WebSimulator Editor

The BayesiaLab WebSimulator uses observations/evidence entered on Input Nodes to dynamically update the posterior probability distributions of Output Nodes.

In the WebSimulator Editor, you can link different types of Monitors with nodes.

 Monitor in BayesiaLabMonitor in WebSimulator

 

 

 

Unobserved

 

 

 

Hard Evidence

 Monitor in BayesiaLabMonitor in WebSimulator




Unobserved

 

 

 

Soft Evidence

Fixed Probability Distribution

 Monitor in BayesiaLabMonitor in WebSimulator

 

 

 

Unobserved

 

 

Soft Evidence

Fixed Probability Distribution

 Monitor in BayesiaLabMonitor in WebSimulator

 

 

Unobserved

Soft Evidence

Mean with MinXEnt

Fixed Probabilities

 Monitor in BayesiaLabMonitor in WebSimulator

 

 

Unobserved

Soft Evidence

Mean with MinXEnt

Fixed Probabilities

Monitor in BayesiaLabMonitor in WebSimulator
Monitor in BayesiaLabMonitor in WebSimulator
Monitor in BayesiaLabMonitor in WebSimulator

The value used in the gauge is the Expected Value of the node.

Monitor in BayesiaLabMonitor in WebSimulator

The value used in the bar graph is the Expected Value of the node.

Monitor in BayesiaLabMonitor in WebSimulator

The optimal decision (i.e. the one with the best expected utility) is displayed in orange.

Monitor in BayesiaLabMonitor in WebSimulator

The utility bar varies from the min and max utility values (here, from -100 to 50).

The bar is displayed in red, from 0 to the utility value, for negative utilities.

Otherwise, the bar is green, from 0 to the utility value.

  • For an XBL file to be uploaded to the server, it should contain at least one Input Node and one Output Node.
  • Contraint Nodes cannot be set as Input or Output Nodes.
  • Utility Nodes cannot be set as Input Nodes.

The performance of the simulator depends on the complexity of your network. It is best to avoid highly-connected networks, i.e. networks that contain nodes with multiple parents.

You can improve the performance of your to-be-simulated network by deleting all unconnected and excluded nodes.

  • A node only have one Input Monitor.
  • A node can have multiple Output Monitors.
  • A node can have an Input Monitor and Output Monitors simultaneously.

The layout of the monitors is based on the alphabetic order of the node names. If you desire a specific node order, you can add prefixes to the node names to serve as indices.

Example

 

 

 

Link