Sussman Anomaly

Sussman Anomaly

http://en.wikipedia.org/wiki/Sussman_Anomaly

Sussman Anomaly

Sussman Anomaly

Read this book if you like that.
http://aima.cs.berkeley.edu/

but it is still useful for explaining why planning is non-trivial.

...when thought of in the wrong way.

However, noninterleaved planners typically separate the goal (stack A atop B atop C) into subgoals

That's already wrong because these subgoals are not independent! The structure of the problem is not independent. The blocks must grow from the bottom. The first subgoal should be "get C on the bottom". This is not "tower of Hanoi" and not all problems are Hanois.

A good example of why that "break things into little pieces" is NOT always going to solve the problem or solve it efficiently - the whole needs to be considered, not the parts.

Cudder is all talk and no action.

>>4
The point was that the wrong way was the most intuitive one. If someone doesn't stop to think about it carefully, they would break it into parts.

>>4
What, they actually have good systems engineers where you work?

>>6
If you think that wrong way is "the most intuitive" then something is very wrong with your brain... even a 10-year-old would probably figure out that "get C on the bottom" needs to be the first step after being confronted with the restriction that only one block can be moved at a time.

Cudder is all talk and no action.

>>8
Right. This restriction means atop is right-associative, where the ``intuitive'' interpretation as >>6 puts it is non-associative, which is clearly inefficient regardless of the problem domain.
This doesn't mean the whole needs to be considered more than the parts though, since A atop (B atop C) is easily decomposable in the same way any formula is.
In addition, the way ``any 10-year-old'' would figure it out likely doesn't work for a collection of n blocks randomly piled around, nor is it describable as an algorithm (it's just intuition).

In fact, the only reason this 'anomaly' even exists is because it lacks a base case.
Add the goal C atop table and the planner can easily figure out the order of execution from the trivially satisfiable subgoals and the goals which depend on them.
So, with the initial goals

1. A atop B
2. B atop C
3. C atop table

3. is trivialy satisfiable as it is the base case. Once 3. is satisfied, 2. is trivially satisfiable, and so on.
Thus I win.

>>11
7/10, you almost got me to write a serious reply.

>>12
Is that your way of trying to sound clever, or do you just mean you can't figure out what I'm saying? Are you a believer in the Sussman Anomaly? I am, but only if the planner doesn't have enough information.

Plunging into the mysterious anus.

I wonder. Could the anus poster in fact be Anal Turing's disembodied spirit that escaped from the grave and made it on to the Internet?

Touring my anus

>>15
I'll fuck your ass like I fucked the enigma.

>>17
You mean the singer:
http://en.wikipedia.org/wiki/Sandra_%28singer%29

or the guys themselves:
http://en.wikipedia.org/wiki/Enigma_%28musical_project%29
?