Visa resultat

Stochastic optimization algorithms 2010, FFR105

Teaching

1. How do you rate the lectures?

Not good at all»		0		0%
Not so good»		1		2%
Neither good nor bad»		8		20%
Good»		11		28%
Very good»		19		48%

Genomsnitt: 4.23

- It was much better if you use slides» (Not so good)
- Mycket copy & paste från bok till tavla till anteckningar.» (Neither good nor bad)
- It is easy to get lost in some of the lectures also in some of the lectures, there are unnecessary materials. » (Neither good nor bad)
- The lectures are more or less textbook...no new information....» (Neither good nor bad)
- I think the lecturer mixes theory, background and applications in a very good way.» (Very good)
- One minor thing: please add numbers to your slides.» (Very good)
- Good pedagogic skills by the teacher» (Very good)
- T.ex.: Väldigt bra att gå igenom applikationer osv.» (Very good)

2. To what extent have the lectures been of help for your learning?

Not at all»		1		2%
Some extent»		15		39%
Large extent»		10		26%
Very large extent»		12		31%

Genomsnitt: 2.86

- Even though the lectures I"ve been to were good, I prefer reading on my own.» (Not at all)
- Actually the course book is enough !» (Some extent)
- Mycket som gås igenom finns såklart i boken, och vissa gånger i bättre format, men föreläsningarna ger indikation om vad som är viktigast (vilket är väldigt bra, eftersom det är svårt att lära sig exakt allt i en hel bok). Dessutom gås det igenom saker som inte står i boken osv» (Large extent)

3. Do you think that the lectures complemented the course literature well?

No, not at all»		1		2%
Not so much»		10		27%
Yes»		15		40%
Yes, very much»		11		29%

Genomsnitt: 2.97

- It was basically only a repition of the book!» (No, not at all)
- I princip samma» (Not so much)
- Since the course literature is based on the lectures, the book and the lectures was very much the same. I think the book was good, but I prefer that the course literature isn"t written by the lecturer.» (Not so much)
- The lectures are a bit too similar to the book. It would help if the same information was given from a different angle in the lectures.» (Not so much)
- the lectures=the course literature (still both very good)» (Not so much)
- Basically were the same.» (Not so much)
- The lectures shouldn"t cover the course literature . It was much better to use the lecture time for solving problems and discussing about the usage of algorithms» (Yes)
- If you mean whether you repeat the book without any modification. » (Yes, very much)

4. To what extent has the Matlab introduction been of help for your learning?

Not at all»		5		13%
Some extent»		13		34%
Large extent»		8		21%
Very large extent»		12		31%

Genomsnitt: 2.71

- I know Matlab much earlier or rather I should say the way we required to use Matlab in this course made my Matlab programming skill drop quite a lot. I my opinion, we need to take full advantage of all the possible merits of Matlab and this is the purpose of using Matlab for fast prototyping. » (Not at all)
- I new Matlab before» (Not at all)
- copying the matlab script is not what consider a guided introduction!» (Not at all)
- I have three years experience with MATLAB, so it wasn"t really necessary. The code generated was good to have later though.» (Some extent)
- It was good to get some help in structuring the program.» (Some extent)
- As usual teachers only solve simple problems as a sample for students and leave the difficult ones for students. It was much better if instead of GA some how LGP or few parts of third question of second examle sheet was solved» (Some extent)
- It made it easier to understand the coding standard and rules, as well as a few pointers and tips. I did it at home, though.» (Large extent)
- I"ve worked with matlab before but forgot most of it! It was very good to start of easy.» (Very large extent)
- It set the coding standard for the course very well and it made the first home problem a lot easier because we had a coding example.» (Very large extent)
- Matlab-intron var väldigt bra, för då visste man hur ni vill att man ska programmera i matlab, och dessutom så var plot-algoritmerna skitbra att ha i HP 1 och 2. Dessutom tror jag att vissa små frågor besvarades, som man inte skulle få reda på om man gjorde GA-algoritmen själv genom att bara läsa boken (t.ex.: om jag bara hade läst boken hade jag gjort så här för att forma en ny generation om pop.size varit jämnt: skapa en kopia av den bästa individen, mutera och korsa N-2 individer, använd asexuell reproduktion för 1 individ, forma den nya generationen med de nya individerna. Detta är inte exakt samma som i matlab intron. Det är såklart inte stor skillnad, men matlab-introns version känns lite bättre. Fler exempel finns).» (Very large extent)
- Excellent idea to brush up on matlab skills.» (Very large extent)

5. To what extent has the Introductory programming problem been of help for your learning?

Not at all»		5		12%
Some extent»		14		35%
Large extent»		8		20%
Very large extent»		12		30%

Genomsnitt: 2.69

- Probably more helpful if you are not used to Matlab.» (Not at all)
- Since I knew MATLAB it was useless for me» (Not at all)
- Det var bra att få feedback på hur man kodade innan man gjorde HP, och det är ett stort plus. Men jag tycker att jag hade ganska stor koll på det ändå.» (Some extent)
- It helps with homework 1.» (Large extent)
- Helps to understand how to make the programs for HP1. It would have been problematic otherwise.» (Very large extent)
- It helped in the ways the » (Very large extent)
- very good idea!» (Very large extent)

6. To what extent have the home problems been of help for your learning?

Small extent»		0		0%
Some extent»		1		2%
Large extent»		14		35%
Very large extent»		24		61%

Genomsnitt: 3.58

- If you consider if the student by only solving the classical problems will learn you are totally wrong (you can find a lot of codes about them). My opinion about learning is that you should provide students with code but deliberately put some programming mistakes inside the code where the students instead of writing code have to debug the code and figure out what the correct result should look like that is how that you will confirm that the student understand how the algorithms should provide the results in each step. » (Some extent)
- I think the home problems were very good and interesting to do. They provided much understanding.» (Very large extent)
- HUGE!!! Keeps the knowledge fresh and helps to really remember the things that are thought through implementation. » (Very large extent)
- Though the last dead line were on the same day as in neural networks.» (Very large extent)
- I guess that in your point of view, it is a notable relationship between programming and optimization course, and it is surely true. However, I think that this course does not really demand any examination as far as you can present more home problems. I mean if programming has a crucial part in optimization, having more programming homework would be more appropriate in theory and practice. » (Very large extent)
- It"s much easier to understand algorithms thoroughly after having implemented them.» (Very large extent)
- More time for problem 2 would be good.» (Very large extent)
- HP var episka! Speciellt HP2 var helt fantastisk. HP1 var för mycket av klassisk optimering. Jag accepterar att det kan vara bra med uppgift på klassisk opt också, men två stycken var lite för mkt. Gör så här: ---=== Algoritm 1 ===--- 1. Kläm ihop uppgift 1.1a och 1.1b till en uppgift (=1.1a), och 1.2a och 1.2b till en uppgift (=1.1b). 2. Sätt 1.2 <-- 1.3. 3. Sätt 1.3 <-- 2.1a + 2.1b 4. HP2 <-- HP2 - 2.1a - 2.1b (Kommentar om Algoritm 1: Nr 3 och 4 behöver inte följas exakt, men jag tycker att HP2 innehöll jättemånga olika algoritmer som man ville utforska mera osv., till skillnad från HP1, som bara innehöll en version av GA. Det borde kunna gå att åtgärda).» (Very large extent)
- HP"s were both interesting and educating.» (Very large extent)

7. What is your impression of the course literature?

39 svarande

Very poor»		0		0%
Poor»		0		0%
Average»		6		15%
Good»		18		46%
Very good»		15		38%

Genomsnitt: 4.23

- Covers the content of the course very well.» (Average)
- It should include more theoretical aspects...» (Average)
- Some of the proofs seemed almost overkill to include in the course, but nevertheless very good litterature.» (Good)
- Well written, easy to follow.» (Good)
- What is missed in course literature and lectures is examples. We need a lot of solved simple examples! One properly solved example is better than 100 pages of best explanations.» (Good)
- But one was poor after buying it. The pricing was much too high for a thin book like that.» (Very good)
- Great book! Well chosen contents very relevant. It"s a keeper! =)» (Very good)
- Very well formulated algorithms that are easy to follow» (Very good)
- As we can see in the program"s plan, we have another optimization course which I guess is classical optimization (linear, nonlinear in detail, binary and etc). It would be a bit motivation for students if you ignore some parts of nonlinear optimization in the beginning of the course and concentrate on more stochastic methods or present some more details. If any body wants to know more classical optimization, the so-called course would be applicable. Please pay attention that this was just an imagination, and you can ignore it since I am just a student. Nevertheless, I think that it would work well.» (Very good)
- Great litterature» (Very good)
- Det fanns information om det som kursen innehöll, samt annat som är kul att läsa vid sidan av.» (Very good)
- I like the condensed version and the to the point appendices. Very good indeed.» (Very good)

8. How well did the course administration, web page, communication etc. work?

Very badly»		0		0%
Rather badly»		2		5%
Rather well»		18		46%
Very well»		19		48%

Genomsnitt: 3.43

- Communication between the lecturer and the students were good, but I think it"s very ignorant of the lecturer not to be able to change lecture halls. It was only in the last two reading weeks that the lecture hall was sufficiently large, and that was because people had grown tired of sitting on the floor by then.» (Rather badly)
- As you have mentioned in you web page about the grading system it confuse the students that not mandatory questions have bonus points but they are required points so change that in the way that the student get to know that there is no bonus points at all. » (Rather badly)
- Except the week that both David and Mattias were not here. » (Rather well)
- In retrospect, the info provided in emails could have been posted on a news section on the web page as well.» (Rather well)
- Simple stuff: email/hyperlinks does its job!» (Rather well)
- All relevant info and data collected in one place. Very good with FAQ and grade info etc. Always updated schedule. » (Very well)

Your own efforts

9. How many hours per week did you spend on this course?

38 svarande

At most 15 hours»		4		10%
About 20 hours»		12		31%
About 25 hours»		14		36%
At least 30 hours»		8		21%

Genomsnitt: 2.68

- If you do not count the hours that spent on running the neural network and LGP. » (About 20 hours)
- Väldigt svårt att säga. Man lade ju i en lite högre växel för HP1 och HP2 och tentan, och då kunde det ju gå åt mer än 20h/vecka» (About 20 hours)
- It was 50/50 with Neural Networks probable and I probably averaged 50hr weeks (or a tad more), noting that work was unevenly spread with a larger amount at the end of the quarter.» (About 25 hours)
- Alot! It took over. =/» (At least 30 hours)
- The home problems eat hours.» (At least 30 hours)

10. How large part of the teaching offered did you attend?

39 svarande

Less than 20%»		2		5%
20-40%»		1		2%
41-60%»		5		12%
61-80%»		7		17%
More than 80%»		24		61%

Genomsnitt: 4.28

- I could not find any apartment in Gothenburge. I moved to Stockholm until I would find somewhere.» (Less than 20%)
- I think I skipped out on some lectures in the end, trying to organise my schedule. Actually, I missed the bulk of the stuff about genetic algorithms, which probably wasn"t so good. Maybe split the second assignment in two?» (61-80%)
- All of it! Found it all useful. » (More than 80%)

Study climate

11. How were the opportunities for asking questions and getting help during the course?

Very poor»		0		0%
Rather poor»		0		0%
Rather good»		9		23%
Very good»		24		61%
I did not seek help»		6		15%

Genomsnitt: 3.92

- unfortunately, i could not use it because making relation was very hard for me, so I think it was my problem» (Rather good)
- Got help whenever I needed it. Very pleased!» (Very good)
- Det är bra att man kan fråga på alla möjliga sätt.» (Very good)
- I didn"t ask too much, but there was ample opportunity. Also e-mail worked well for some questions, which is always appreciated.» (Very good)
- Both teacher and TA"s very helpful and kind.» (Very good)

12. How was the course workload?

Too low»		1		2%
Low»		1		2%
Adequate»		15		38%
High»		18		46%
Too high»		4		10%

Genomsnitt: 3.58

- I would have wanted to spend more time on the second home problem, if it hadn"t been for the Neural Networks course where the workload was high.» (Adequate)
- It was high in the sense that it took time to du the home problems. But that was partly because the problems did take rather long time to run.» (High)
- Maybe a bit much work with the HP but I got alot out of it! Would not want it less, considering what I gain.» (High)
- Home problems were really challenging (and quite interesting by the way!), but I"d rather prefer having let"s say five smaller assignments instead of two big ones. But I guess this is mostly due to my bad habit of doing everything in the last moment, which is a wrong strategy when solving such a voluminous problems.» (High)
- I thought this course was very time demanding and I can"t imagine how high the work load was for the guys that took both this course and the neural network course.» (High)
- I mean, I had quite a few late nights in the lab that I really do not see how I would have done without. Alas, this is what one has come to expect from the Eng. Physics programme.» (High)
- HP"s might"ve been a bit too cumbersome.» (High)
- It is time consuming» (Too high)
- Badly managed work load. The introductory didn"t have any work load but the last example sheet was a lot of work load so it was better to distribute the work load of example sheet and give more time for students to solve question.» (Too high)

Course goals, level of difficulty, exam

13. How understandable are the course goals?

39 svarande

I have not read the goals»		6		15%
The goals are difficult to understand»		0		0%
The goals give some guidance, but could be clearer»		5		12%
The goals clearly describe what I am supposed to learn»		28		71%

Genomsnitt: 3.41

- Although I might suspect that the goals are at least to learn different stochastic methods of optimization, understand benefits and drawbacks and of course learn how to implement them. » (I have not read the goals)

14. Is the level of difficulty of the course reasonable, considering your background and the number of credits?

No, the level is too low»		2		5%
Yes, the level is reasonable»		34		89%
No, the level is too high»		2		5%

Genomsnitt: 2

- It is not true to say it was hard or not. it was completely combinational subject, and what we learnt was an ability in computing and special modeling. Although we have to go through the details, however, in practice, being able to use these algorithms seems to be enough. » (?)
- I"ve spent more time coding and debugging Matlab programs and writing reports, rather than learning something really new. The course book is rather simple and can be completely covered in one-two days (except for the exercises), therefore I think that the course itself was not challenging enough. The deadlines were challenging though, but it means that the course was very time-consuming, but not really hard. The same with the exam - I just didn"t have enough time to answer all the questions, though I knew the answers. As the bottom line: I prefer writing less and thinking more instead.» (No, the level is too low)
- was challenging though, with no strong programming background» (Yes, the level is reasonable)
- I"m a bit interested in Matlab programing so that my have been to my benefit. My background is Kandidat in Bioteknik, and currently Master student of Chemistry and Bioscience (MPCBS). » (Yes, the level is reasonable)
- Compared to out courses it is high.» (Yes, the level is reasonable)
- The concepts are easy but the proofs are difficult. I don"t know the use of proofs for students (We as students need to learn some thing that is useful we are not going to defend our papers ,thesis,... to provide proofs)» (No, the level is too high)

15. Did the examination as a whole assess whether you have reached the goals?

39 svarande

No, not at all»		1		2%
Yes, to some extent»		18		46%
Yes, definitely»		20		51%

Genomsnitt: 2.48 (bidrar till totalt genomsnitt/jämförelseindex)

- The mixture is very bad. You should ask students to solve algorithms in MATLAB not in exams . And exams should contain a materials that the students shouldn"t memorize at all. Unfortunetly the exam was designed to only evaluate the students ability of memorizing long formulas and proofs not evaluating how much they understand the concepts. As an example if the students make a little mistake in exam and misplace one of the parameters in the algorithms then he will lose the next part of question which is supposed to solve one problem with what he has memorized and put on paper and this mistake also reflects in the next problem also and he lose points because of little mistake that he made. For this purposes make two types of exams one without book (to evaluate the memorizing capabilities of students ) and next with book to evaluate the problem solving capabilities of students. [As I mentioned the students only have to solve problems without even memorizing one formula]» (No, not at all)
- It was as expected. The questions were as promised about things that we talked about or read about. » (Yes, to some extent)
- As I described, 25p in exam is not really necessary. Homeworks are more important and should be. » (Yes, to some extent)
- I would have needed more time on the exam to answer all the questions. With only 4 hours I only managed to answer about 65% of the questions, which does not reflect my level of knowledge at all. » (Yes, to some extent)
- Some of the proofs for the exam feels like they are just included to have something theoretical in the course. The example calculations on for example GAs with a small population and the descriptional questions are a better test of whether the student has really understood the course content.» (Yes, to some extent)
- The final exam could be removed. It really does not serve much purpose. The problems are fairly simple like doing Lagrangian multipliers and some stuff which involves easy and straightforward calculations. » (Yes, to some extent)
- This exam was in some how better than the earlier exams, but still many hours are spent just on doing silly calculations. Can"t see why that should be of interest.» (Yes, to some extent)
- Home problem 30 Exam 20» (Yes, to some extent)
- But I think the exam was maybe a bit too extensive. There was a little too much to write.» (Yes, definitely)
- The exam covers almost entire course and is well-composed from my point of view.» (Yes, definitely)
Genomsnitt totalt för detta stycke: 2.48

Summarizing questions

16. What is your general impression of the course?

Very poor»		0		0%
Poor»		0		0%
Average»		3		7%
Good»		13		34%
Very Good»		22		57%

Genomsnitt: 4.5

- It could be much better since the topics are interesting and nice but it only depends on your teaching techniques which are still old and classic (not modernized)» (Average)
- The home problems were very challenging for me, coming from a design background, but I probably learned more doing them than I would by reading the course litterature twenty times.» (Good)
- Very interesting topic. Home problems are challenging and fun.» (Good)
- Awsome! Thanks for a very good set of tools and a new way of thinking about problems. » (Very Good)
- One of the most interesting courses I"ve taken.» (Very Good)
- One of my favourite courses at Chalmers» (Very Good)
- Känns som en väldigt vettig kurs, som är användbar nästan överallt, även i fritiden [för att ha kul:)]» (Very Good)
- Very interesting course.» (Very Good)

17. Is there anything you think should be changed until next year?

18. Which areas do you think need more emphasis (or less emphasis) in the course?

- No»
- Less emphasis on classical optimization and more on PSO.»
- The way it"s planned is good.»
- Given the fact that this is an introductory course, I think the current time allocation in each part is very good already.»
- I think the emphasis was good.»
- maybe a bit more on ACO, especially what happens when the last path chosen is one with zero pheromone level.»
- The course is very good as it is, perhaps mention something about hybrid algorithms.»
- Less emphasis: Matlab programming. More emphasis: Classical optimization vs. Stochastic optimization, more advanced math.»
- Why so much focus on the analytical properties of GA? Instead of spending sleepless nights on evaluating infinite sums, I would rather learn more about stochastic algorithms in a practical view, i.e. how to implement constraints etc.»
- I thought it was a good spread»
- I don"t know if all the proofs were really neccessary to understand the algorithms used in the course.»
- The lectures more or less are the same as the content of the book. So, in principle, for this course, the book alone suffices making lectures redundant. Either the prescribed book could be upgraded to an advanced one which has more mathematical rigor and discussion of ideas than just discussing the algorithms or such discussion could be done in class. Basically to make some difference between lectures and the course book :)»
- Helt klart EA, PSO, ACO. Inom dessa algoritmerna är det nog PSO (som skulle vara kul att även använda till andra problem än fuktionsoptimering). Mindre relevans för klassisk optimering.»
- During the second home problem, I realized that PSO is a very effective algorithm. I think that this topic could be expanded giving more examples of practical applications. Maybe, as it was done for ACO or EAs.»
- EAs , PSO and ACO Also Matlab programing»
- Non-Matlab programming? Seriously though, as Mattias pointed out, some more stuff on that super efficient PSO coding?»

19. Additional comments