Continuing from my last blog post, there are two more steps to finish the training of the system based on first example, as follows:
2. Updating examined rules: In the second step we will update the strength of rules based on their accuracy. In this example the strength of the rule #2 will be increased from 25 to 35 and the strength of the rule #5 will be decreased from 25 to 15.
3. Modifying set of rules: In this step we will update rules. In this example we will withdraw the rule #5 and replace it with new random rule, shown blow.
The old rule #5 will be replaced by the new one in the set of rules in intelligent agent. Then, the intelligent agent gets more intelligent. It’s time to apply the second example. Based on this training example:
“The windshield is dry and the distance from the car ahead is fine and the velocity is high and the engine is on and the lighting is sufficient and the temperature high ð only doors are locked (no other action taken)
The example is coded as: 011111: ####1”
- Searching among 11 rules in intelligent agent in order to find rules that their If-part match with the If-part of the example. What rules do you think will be picked?
- Updating examined rules: How would you update the strength of selected rules?
The set of picked and updated rules is shown in the figure below. In the next step (3rd step), weak rules (when a rule’s strength is lower than 25) will be replaced by some random rules.
In real example, the process of training happens hundreds to thousands time. Each time an intelligent agent gets more intelligent by examining random rules. In my next blog post, I will explain more about real systems and how it differs from our simple example. If you have any question for me, don’t hesitate to message here or on my LinkedIn.