2010-07-21

Adaptron Test Run – 1 - Date: 21st July 2010

In this test case #39 which is very similar to test case #7 (modified to test A_Sequence creation and execution) I have changed the setting up of action habits. When the action to be done is the result of a partial match with the trigger I do not reduce the redo interest on the partial match. To be specific the trigger stimulus is A+B = Sequence object #22. There is no interesting action worth repeating for this sequence. However the second part matches with B = object #20. This has a listen action that has an interesting result. So the listen gets tried as an action after the A+B. However I do not reduce the interest in the B-action-listen sequence because it is not being done. Only its action part was being used. Thus the next time the A+B occurs it retries the partial match action and will continue to do so forever until the B-action-listen sequence is learnt.

One possible solution would be to assume we are executing the partial match action sequence. This is actually being done. The A+B is a more specialized situation and we are doing an action based on a more general situation (just B). The A+B is a special case of just B. But should we be updating the more general situation lesson based on this specialized result? I think we can. If later we find that a different specialized situation, say C+B, does not give the same goal result then this specialized lesson gets remembered. Thus we have discrimination.