Evaluation of project results for ISC612 MISIC Course I semester 2007/2008:

I evaluated each project precisely out of 20 ("Precise" column) and then
penalized lateness according to the amount of hours of late submission
("Lateness" column) to obtain the awarded markings out of 20 
("Awarded" column). These have been mapped to literal categories
(A,B,C,D,F) with D being the lowest pass mark. Overall I was happy with
the effort put in by most teams. 

Project name	   Percentage	Max  Lateness	Mark            
                                        	Precise	Awarded Category
automata	   78%		18   no		14.0	15	B
eternity	   90%		16   15h	14.4	14	B
lcp		   60%		20   10h	12.0	12	C
networkflow	   0%		14   no	 	0.0	 0	F
rc_scheduling	   85%		15   no		12.7	13	C
sepnet		   90%		20   no		18.0	18	A
spares		   85%		20   no		17.0	17	A
uml		   63%		15   70h 	9.0	 8 	D

Prosvirnova	   A
Rivas		   A
Pharwaha	   B
Albarran	   B
Brahim		   D
Bendahou	   D
Foessel		   B
Jouai		   F (undergoes oral examination)
Crosthwaite	   C
Ferreira	   C
Kinengue	   C
Chahid		   C
Guillet		   C
Matos		   A
Lecourtois	   A

Jouai: oral examination on 080303, mark 8, category D

Cellular Automata:
Very good effort on modelling and good effort on presentation. Since the tested
instances are deemed infeasible by CPLEX, there must be some mistake
in the model or data. I checked the model and couldn't find any mistake.
I am going to award 95% for modelling, 80% for presentation and 60% for
the effort on practical results, with an overall 78%.

Eternity:
The author did quite a lot of work by studying and applying the Choco
solver which has not been seen during the course. The main goal of the
project, i.e. designing a model for the puzzle, has been
achieved. Some remarks concerning the work are given below. The
project report is somewhat too ``concentrated''. This complicates its
understanding. This concerns especially the 4th paragraph in the model
description, it is not very well explained. Nevertheless, the overall
quality of the project is good.
- Nothing is done in the model to break the symmetry. Every
  solution has at least seven symmetric solutions (if you rotate or
  overturn it). Taking this into account would make the resolution by
  up to 8 times faster. The importance of symmetry breaking was
  underlined during the Constraint Programming lecture.
- Somewhat related with the first remark. The size of solved
  puzzles is not very big. Other formulations and/or heuristics
  could have been tried.
Taking into account all this and the fact that the project has been
performed singly, the overall project evaluation is 90%.

LCP:
Good presentation effort but unfinished report with personal comments
in it (e.g. "all crap") (70%), good modelling attempt (80%) but confusing
results, although some code has been written. Mention of C++ code but
this has not been submitted (60%). Overall 60%.

Network Flow:
Completely out of scope, the word "flow" is never even mentioned in the 
report: overall 0% (need oral examination).

Resource Constrained Scheduling:
The authors accomplished a fair amount of work to reach the aims of
the project. Some literature on the problem has been studied. Both
parts of the project are fully completed. Some small defects are
present but they are not very important. Their list can be found
below. The report is written in a good style and using a good English.
Flaws:
- The first algorithm to find a longest path uses a topological
  order. The authors did not explain how to find such an order.
- The complexity of the first algorithm to find a longest path  is
  not analized.
- No proof is given to the statement that the given Djikstra-type
  algorithm actually finds a longest path.
Taking all this into account, the overall project evaluation is 85%.

Separation Network:
A very good modelling effort overall, as the problem was really hard. The
fact that you obtained results at the end is also to be commended. The
presentation is somewhat messy but I must concede that the problem difficulty
makes it hard to present it in a simpler way. I award 100% for the modelling
and practical side and 80% for the presentation, for an overall 90%. Notes:
Question (d):
  - c2.2: why b+1? Is the model only valid for a particular node ordering?
    If so, this is an important part of the model and must be made explicit.

Spares Optimization:
The problem is interesting and novel, the reformulation seems to be
OK.  Congratulations on actually obtaining good computational
results. The report itself was written hastily and not very
convincingly. 100% for results and 70% for presentation; this is an
overall 85% score.

UML Class Diagrams: 
In the model, the objective is sum c_ij y_i y_j, not just sum c_ij as
is written in the report. The modelling effort was not spectacular but
mostly OK, although the model itself is not difficult and was even
explained in class (80%). The practical results are missing: the AMPL
script terminates with various errors, although the syntax is mostly
OK (40%). The presentation is decent but not excellent (70%), for an
overall 63%.