Graphical Models: Assignment 5
Due February 26, 2019
- Consider the following problem of diagnosing faulty
equipment. A broken belt or a hot room may cause the engine to
overheat. If the engine is overheated, then the temperature
light will probably turn on, unless the temperature sensor is broken. A
hot room can cause the temperature sensor to break. A broken air
conditioner may cause the room to be hot. The belt is more likely to be
broken and the air conditioner is more likely to be broken if the
organization that owns the machine has poor maintenance practices. An
overheated engine can cause defects in the product produced by the
machine.
- Draw a Bayesian network to represent this problem. Use a software package such as Netica.
- Describe your random variables and states. Explain why you chose the structure you did.
- Define local distributions for your random variables. Explain why you chose the distributions you did.
- Enter evidence for two different scenarios. Do the results seem reasonable for the problem? Discuss.
- Install UnBBayes by unzipping the archive at this link. Follow the Vehicle ID Probabilistic Ontology Tutorial to
construct the Vehicle ID Probabilistic Ontology in UnBBayes. Produce
a screenshot of your MFrags by using the See MTheory button (MTheory
icon with a magnifying glass) at the top of the MTheory pane of
UNBBayes. Also produce screenshots of the two SSBNs you generate
as part of the tutorial. You can save all these screenshots using the
"Save graph as image" button at the top left of the image window (hover
over the button to see help information). Comment on your model
and your experience using the UnBBayes software. (More information on
UnBBayes can be found on the UnBBayes Sourceforge site.)