I'm trying to understand bayesan networks also I created a simple bayesan network according to same sample date.
This is the network (created with Hugin Lite)
There is one class (Failure) and two attributes (light and temperature).
Failures = kinds of failure (three possible values: Mechanic , Electric , None) Light = indicates if a light is on or off when there is a failure (which can be Mechanic OR Electric) Temperature = indicates the temperature (three possible values: low,normal,high)
Light and temperature are independent variables
In the beginning there is NO EVIDENCE about variables. These are the probability tables of "A PRIORI probability":
Now an EVIDENCE is provided. So we'll have a "A POSTERIORI Probability"
We know that:
- Light is ON (so in the first table there will be just one row: Light = ON)
- Temperature is LOW (so in the second tablet there will be just one row: Temperature = LOW)
The column of both table remain the same (Mechanic,Electric,None).
My question is: how should I update conditional table?
I'm sure there is some relationship with "Bayes formula" but I'm a bit confused.
This was an homework but I'll have to a similar (not the same) homework next month for an university project. My aim is to understand how to calculate the new probability.
Thank you in advance for any hint.