Parts:
•(15 Marks)Introduction

This is all setting the CONTEXT of the problem being addressed.  The
problem is predicting how long A* will run, which we have simplified
to predicting how many nodes A* will generate/expand.  However,
modelling A* is difficult because it uses global duplicate checking.
So it makes sense to model IDA* because it doesn't use it.  However,
in many domains there are a lot of duplicates, and they cause a lot of
the space to be repeatedly searched, so some form of duplicate
checking is needed.  The most local form is grandfather checking which
has been modelled.   The intermediate form is loop elimination, which
currently has not been modelled.  We would like to develop a way of
modelling this level of duplicate checking.  To simplify matters
further, to model loop elimination it does not matter whether we are
using heuristics, so we will not use them.  This means we are simply
trying to model loop elimination in depth-first iterative deepening.
If we can then it will immediately transfer to IDA* and hopefully will
give us some insight into modelling global duplicate checking.

Need to describe IDA*, A*, and ID.

•(15 Marks)Problem Description

This describes the problem itself.  Discusses what are loops in ID,
discusses the common approach to predicting tree sizes, e.g., the
formulas, introduction of classifying states into types and building
type graphs and recurrence relations, and the importance of being able
to calculate the branching factor, and ways of solving the recurrence
relation to come up with the branching factor.

•(25 marks)Related Research 

Discuss the different approaches to computing tree size focusing on
current approaches.  Discuss the main ideas and compare the
similarities and differences of the approaches.

•(10 marks)Description of  test runs between ID with 
           and without loop elimination


Description of what problems were being run (all of the length 10
problems, all of the problems, some other set of problems), how they
were being categorized (all the problems, split into even and odds,
?), how the KRE formula predictions were being calculated,
specifically how the branching factor was calculated, e.g., giving the
recurrence relations, solving analytically the recurrence relations,
solving them by numeric approximations, or experimentally.  Whether
the bipartite nature of the graph was being dealt with or not.


•(15 Marks)Analysis of the experiments 

How the comparisons were being made between the predicted tree sizes
and the actual tree sizes.  Describing what the baseline was for the
accuracy of the predictions were for ID without loop elimination and
how the accuracy of the predictions for ID with loop elimination
compared to the baseline.  And some conclusion about whether one was
significantly more accurate than the other.

•(10 marks)Description of your proposed modification 
to the formula and the rationale behind the modification

Describe how you would modify the KRE formula to make it more accurate
for ID with loop elimination and explain why your modification makes
sense.  You need to relate you change in the formula to the change in
the behaviour of ID that loop elimination brings.


•(10 Marks)Conclusions

repeat of the context, the problem, and the proposed mod.

out of 100