UG Exam Performance Feedback Second Year - Semester 2 COMP20012 Algorithms and Data Structures DER GDG No feedback received - please see DER.Comments: COMP20032 Distributed Computing RiS CCK Q1 (a) IDL = Interface Definition Language. Getting the acronym wrong (invarious ways) was too common! b) The first lab exercise involved running a Java servlet using servletrunner. It is clear that some students didn't understand what they were doing, and weren't curious enough to find out. c) The main assumption is that the time delay in sending a message to the server is equal to the time delay in the server sending it back again. If this is true, Cristian's algorithm works perfectly! Other assumptions, such as shorter round- trip times being better, are secondary. Q2) Sshows that a large number of students still did not fully understand the concepts of ACID transactions - most students got question 2.a and 2.c right but 2.d wrong. Also, most students were able to correctly answer questions related to messaging and service-oriented architectures. Q3)To come from RiS. Q4 (a) Suggesting that semaphores be used to achieve distributed mutualexclusion is not sensible. It just leads to the question of how to implement semaphores in a distributed system - and one answer is to use a single coordinating server in the way this question required you to explain. (b) and (c) - bookwork. Comments: COMP20072 Computer Graphics TLJH Section B. All the candidates answered the Colour Model question, and no-one touched the Bezier Curves question. This surprised me, because I didn't perceive the two questions as being particularly different in terms of difficulty. The Colour question was reasonably well answered, with an average of 13.2/20, or 66%. The best mark was 20/20, the worst 5/20. About 1/3 of the candidates used no diagrams at all, which is plain daft, especially as the paper implored them to. Many people simply didn't provide enough detail in their answers, and this was especially apparent in the section on JPEG, where people would list the steps of the algorithm but give no context or explanation at all. And this of course makes you doubt if there is any real understanding, or if it's simply regurgitated memorisation. Overall I got a strong feeling of two "levels" of answers: those who wrote lucidly and explained things well; and those who listed facts in isolation. I don't know whether this corresponds to people who did or didn't attend lectures, but it would not surprise me. Comments: 07 January 2008 Page 1 of 7 COMP20142 Logic in Computer Science AV No feedback received - please see AV.Comments: COMP20212 Digital Design Techniques EWH PWN COMP20212 Feedback Questions 1 & 2 Q1. On the whole this question was answered poorly. The operation of the required controller was tightly bound, yet a number of (incorrect) variants were produced. Two states which caused problems were the 'Idle' state and the 'Close Tray' state, where a number of required conditions were not tested for. It was apparent that there was a lack of understanding on how to design ASM charts from a textual description of the controllers operation. The simplification of K-maps including map-entered variables was clearly difficult for a number of students, and it was clear that some students were unsure how to derive simplified equations from such maps. In addition, everyone who attempted the question failed to recognise that as the design was one-hot, there were a considerable number of don't care states in the k-map making simplification easier. Q2. This question was relatively straightforward, with two possible solutions for part c). However, there were a few problems: The discussion of the design approaches was limited in most cases, and failed to discuss the technological differences between the three design approaches. A number of students failed to discuss the differences between PLA, PAL and ROM devices; this was discussed in detail in the lectures. The advantage of the PAL being the availability of on-board flip-flops making the implementation of sequential systems much easier. The simplest implementation of the code converter of part c) is using the ROM, as the logic table can be implemented directly in memory. However, a number of students failed to recognise this. In addition, a few simply listed the gray code on the decoded outputs of the 4 to 16 line decoder, which is clearly incorrect as the output is BCD! Comments: 07 January 2008 Page 2 of 7 COMP20252 Mobile Systems DAE 36 candidates sat the COMP20252 exam on May 30 2007. The performance on the exam was generally encouraging, and indicated that the candidates had absorbed a considerable proportion of the core course material. Question 1 was a compulsory 'shotgun' question, asking for answers to ten out of a dozen short questions each worth 2 marks and between them covering all aspects of the course. The great majority of candidates performed well on this question, achieving marks in double figures. Question 2 asked for an essay on the material covered in the practical laboratories. In the lab students worked in teams with each individual responsible for a component of the lab work, and here they were asked to write about their component. The essay nature of the question resulted in a relatively narrow spread of marks, with no-one doing really well or really badly. Question 3 was a technical question on error detection and correction. Candidates confident enough to answer this generally did well. Question 4 began with JPEG image compression bookwork but then moved on to mobile system issues, and require candidates to think outside the material presented in the course. Answers were generally encouragingly good. This is the first year COMP20252 has run, and it has deliberately been set up in an experimental way involving teamwork, self-organisation, a technical laboratory, and lecture material presented in a very dense format in half the usual number of lectures. Although there have been problems, and some complaints about work-loads, the exam performance reinforces my view that the course has been effective in that the students have absorbed the material well and have emerged with real practical knowledge of the subject matter. They have worked hard, but their generally good marks reflect their achievement. Comments: 07 January 2008 Page 3 of 7 COMP20312 Fundamentals of Databases SH Overview COMP20312 was taken by 118 undergraduate students mainly from the School of Computer Science but also including BCS, ABIS, with BM, Geology with IT, Physics with IT, and 3rd year Mathematics students. I'm pleased to say that this was one of the best answered COMP20312 exams I have set with the average score being 37.86, making 63.1%, with a standard deviation of 4.4. Students answered 3 out of 4 questions with topic areas spread through all questions avoiding the problem of answering single topic areas and better testing the spread of a students knowledge. I am happy to say that the average standard deviation per student over 3 questions was only 3.0 meaning that students scoring low - scored low over all questions, and students scoring high, scored high over all questions. Some anomalies did exist but these were mostly due to students running out of time and not answering all parts of their last question. This year however there has been one major change regarding the exam format and two changes regarding the unit delivery; both have had an impact on the examination results for this year. Firstly, the examination has now been changed into two parts, one compulsory - comprising 10 multiple choice questions (MCQ) valued at 2 marks each; the whole representing one conventional 20 mark question; and one part comprising 3 longer style (conventional) questions of which two must be answered. Part one aims to test Blooms (Cognitive) Taxonomy on Knowledge while part two aims to test Blooms (Cognitive) Taxonomy on Comprehension, Application, and Synthesis over each question. Secondly, in last years examination report I said: "Q1 part c covering functional dependencies and normalisation proved most problematic for students. Normalisation always proves to be difficult for students and I intend to move it from an example topic to a required lab exercise next year to expose the students to it in a more practical setting...Q2 part c covering SQL and relational algebra proved most problematic for students. Relational algebra was the main problem and always proves to be difficult for students, again, I intend to move it from an example topic to a required lab exercise next year to expose the students to it in a more practical setting." I therefore made substantial changes to the unit so that relational algebra, functional dependencies, and normalisation where included in both the examples clinics (by adding an additional class) and to the laboratory (by removing some the simpler SQL work). The effect of both these changes can now be seen in this summers 2007 examination. Part A Questions 1-10 (2 marks each) Marks: Average 15.36 / Mode 18.00 / Median 16.00 / Average % 76.78 / Standard Deviation 3.32 - Answered by 118 (all) students. This was the highest scoring question mainly due to the fact that each question investigated 'bookwork'. However, the question skewed the results slightly because the conventional 3 questions previously allowed a total of 12 marks to be earned for bookwork overall - this year we allowed 20 but over a much broader spectrum of topics. This skewing will not occur next year as I intend to create another 10 part MCQ testing Blooms (Cognitive) Taxonomy - Application (re-evaluating laboratory and examples clinic work) and move the testing to 2 large MCQ questions testing all topics; and then a choice of 1 (from 2) conventional questions testing Blooms (Cognitive) Taxonomy - Comprehension and Synthesis. Part B Questions 1 (20 marks) Marks: Average 10.80 / Mode 14.00 / Median 11.00 / Average % 54.01 / Standard Deviation 3.48 - Answered by 96 students. This was the third most answered question and tested relational schema's, functional dependencies and normalisation, and database performance. I must say that I think this question was a success and the decision to move functional dependencies and normalisation into a practical setting has been vindicated. Indeed, this part of the question was answered very well by most students. The main problem here was with regard to the 'original thought' (synthesis) question which always proves problematic. Questions 2 (20 marks) Comments: 07 January 2008 Page 4 of 7 Marks: Average 9.44 / Mode 6.00 / Median 10.00 / Average % 47.20 / Standard Deviation 4.85 - Answered by 25 students. The least answered question and for the most part seen as a question of last resort by most students. Those that seemed to attempt this out of choice did well, but those who chose it out of desperation did very badly (born-out by the mode / median disparity). The questions tested consistency and completeness, the SQL language (application in detail), and ER diagramming from an original thought aspect. The consistency and completeness question brokered some differing interpretations which I accepted and marked for their substance, the SQL question was answered well, again the original thought question proved problematic. Questions 3 (20 marks) Marks: Average 12.13 / Mode 13.00 / Median 12.00 / Average % 60.66 / Standard Deviation 2.62 - Answered by 114 students. By far the best answered long style question also chosen by the most students. Very little to say on this question, most students answered it very well indeed. It tested transactions (taught last and so fresh), (E)ER diagramming (application in detail), and relational schema as original thought (synthesis). FIRE ALARM DISRUPTION This years examination was disrupted by a fire alarm. Accurate details have been lodged with the schools Special Circumstances Committee, however, the general details are that the exam started on time at 09:45 on Tuesday 05th June 2007, paused between 10:30 - 11:05 for a fire alarm (false) and then restarted finishing at 12:15. I can understand that some students 'thought' may have been disrupted although having 30 minutes of extra revision time after knowing what the questions are I would expect most students to see this disruption as more of a 'gift'. Looking across student answers and over question sets the standard deviations are pretty constant forming an approximately normally distributed population (following, the classical, central limit theorem). I see no reason to make adjustments to the scoring / grading or re-run the examination. In Summary I was very pleased with the students performance this year (well done) and I am further convinced that practical exposure to subjects, active learning, the full unit text (no ad-hoc lecture notes), the weekly SAQ's, and a change in the exam style (as stated above) is the correct direction for the unit. I will continue to pursue these in the coming years to hopefully increase the students comprehension of, and practical exposure to, databases in support of their future careers. 07 January 2008 Page 5 of 7 COMP20352 Software Engineering 2 SME JTL KKL Report on Student Performance in COMP20352, Summer 2007 Question 1 About half the class answered this question. The average mark was 13.2 out of 20. a)(i) Most candidates correctly identified all 6 partitions, although acouple of people were confused about what a partition actually is. A small number of people explained the principle of equivalence partitioning in general, without making reference to the specific example mentioned in the question. Unfortunately, since the question did not ask for such general comments, no marks could be awarded for answers that did not get into the specifics of the example. (ii) Most candidates managed to give a sensible list of test casesthat covered all the partitions. But very few correctly identified the minimum number of tests needed to covere all the partitions, which was just 3. b)(i) Everyone who attempted this part of the question scored full marks. (ii) Some good CFGs were given but this question was also a source oflost marks for many, due to errors in identifying which lines of the program should appear as nodes in the CFG and in specifying the control flow between them. The flow pattern for the for loop at line 11 tripped up a number of students. Note that after the statement on line 15, the flow of control returns to line 11, and that when the condition at line 11 is false, the flow of control jumps to the statement on line 17. (iii) Answers to this part of the question were marked as though theCFG given in answer to part b (ii) was correct, so that students were not penalised twice for incorrect CFGs. Marks were lost for incorrectly remembering the equation for computing the McCabe Cyclomatic Complexity Metric, and for failing to identify a complete set of independent paths. (iv) Everyone who attempted this part of the question got full marks,except for a handful of candidates who failed to give an actual value for the test case input. Recall that giving a description of the set of suitable values for the test case (e.g. saying "some value for n less than or equal to 1") is not sufficiently detailed for a true test case. True test cases must be specified with actual literal values for inputs and outputs. (Check back to your COMP2034 notes if you are unsure why this point is so crucial.) Comments: 07 January 2008 Page 6 of 7 COMP20442 Artificial Intelligence Programming IEP Generally, performance on the exam was much poorer than in previous years. There is no obvious explanation for this, as the course hardly changed at all. The students complained throughout that the labs were very difficult; however, lab marks seemed to hold up quite well. Q1. This was generally very badly done, and (mercifully) attempted by relatively few students. a Most students got this. I turned a blind eye to minor errors. b This was done well. c A few students got this: many struggled. d A few students got this: many struggled. Q2. This question was attempted by all and exceedingly well done by most. There was some sloppiness in presenting grammar rules, in c, and some students gave the semantics for sentence (i), instead of sentence (ii) as instructed, which simply lost them the marks. Q3. a Pretty well everyone could do this, but a disturbing minority couldn't. Frankly, for this minority, I would say there is no hope. b A simple question, which most students made unnecessarily difficult for themselves. There was no requirement in the question to conform to any linguistic theory. The students had also seen a very similar example in the lectures! c I was pleased that quite a few students got this (or nearly got it). d There were few serious attempts at this. Q4. This was generally badly done. a Easy, and well done. b Only about half the students got this; the remainder have no excuse. c Only the best students attempted the all-important second part of the question: I was looking for `equisatisfiable, not logically equivalent'. d Most could do this. e Most could do this. f I was very gentle with people who had accumulated errors on part b. g A complete wipe-out. Only the top student really got this right. Comments: 07 January 2008 Page 7 of 7