15th July 1992
Only becomes more discriminatory with input when it is more meaningful i.e. cues to punishment or reward.
Some rules for learning compiled at AAAI conference and while listening to Case Based Learning by Kristian Hammond:
- Don't make the SAME mistake twice. - make a different mistake and learn.
- Try what has worked before in similar situations.
- Keep trying.
- Don't stop learning when you succeed, keep on getting better.
- The world is a safe, stable and calm place so you can reuse experiences over and over again.
- If it worked in the past use it again, it will most likely work.
- If it works again, don't worry about it, just do it.
- If it does not work - fix it and remember it and don't do it that way again.
- Doing is more important than proving.
- Combine small schemes to do hard things, learn to do easy things first in order to combine them for learning to do harder things.
- Learning considerations: Context, Efficiency, Continuing to learn, keep track of circumstances - they change. When does it work? Where does it work? What are we trying to do?
- Use cause and effect relationships (associations) about what can go wrong.
One possible storage structure worth some exploring is the use of a Markov tree for the storage of experiences and their recall.
A Deictic pointer is a reference to something such as 'it', or 'the event that took place over there'. It is the encapsulation of a complicated something into a unit concept that can be worked with as a unit or exploded into its components if and when necessary. Short-term memory is up to 7 +/- 2 deictic pointers (markers) of the real world.
Let the real world represent itself (model itself) which it does in the form of the experiences recorded in memory.
To pay attention to something - form a mental image of what is wanted and it matches the real world seen, object that matches the mental object (experienced).
Form an image of the goal and satisfaction obtained when solution matches the image of the goal.