***Important***: Some errors in the Project 3 recitation slides..

[Subbarao Kambhampati <rao@asu.edu>]

Folks

 There were two errors in the project 3 recitation slides that the TA 
used, which seem to be confusing several people.
[The TA did ask me to look at her slides before she gave the recitation, 
and so the fact that the errors were not
detected is my fault.]

Anyways, here are the two errors:

1. Goal test function

 The slides say that you should check to make sure that goal test function 
should test for _equality_

 The correct definition--as we discussed in the class as well as in the 
text book--is that you want to check that
all the elements of the current regressed state are _in_ (i.e., members of) 
the init state.

As an example, suppose your goal is to get apples and bananas. In the init 
state you have apples, bananas, persimmons and
asian pears. You can see, using your commonsense, that you already achieved 
the goal (there is nothing wrong in having
more fruit than you needed--fruit is good for you ;-). If you check 
equality, you will say you are not done yet..


2. Set-difference heuristic

The set difference heuristic should take the set difference between the 
current regressed state and the init state.

If the size of this set is zero, clearly, you have the plan. Otherwise you 
need to continue.

Continuing the fruit example, suppose, your current (regressed) state is 
"Apples and Mangoes", and the init state is still
"apples, bananas, persimmons and asian pears", then, we see the set 
difference between the former and the latter  is
{Mangoes}--which is the only (sub)goal that is missing from the init 
state--so we can't terminate. The set-diff heuristic
value here is 1 (corresponding to mangoes).

That is all.

Hope you have a fruitful project 3
Rao

ps: in the second example above, if you regress the state "Apples and 
Mangoes" over the action "Buy mangoes" then
presumably you will have the state "Apples" which does hold in the init 
state and so we can terminate.