Aktuella utvärderingar

Visa resultat

Här kan se resultatet från utvärderingen och exportera statistiken till ett annat program. Det går också att göra en enkel filtrering genom att klicka på svarsalternativen och kommentarerna eller en avancerad filtrering genom att använda knappen längst ned.

Artificial Neural Networks, FFR135

Status: Avslutad
Öppen för svar: 2011-10-10 - 2011-12-31
Antal svar: 47
Procent av deltagarna som svarat: ?%
Kontaktperson: Bernhard Mehlig»
Utbildningsprogram som genomför enkäten: Teknisk fysik 300 hp

Your own effort

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

We mean total time, that is, it comprises the time you spent in class and the time you spent on your own work. Try to estimate the average time over the entire study period.

47 svarande

At most 15 hours/week»0 0%
Around 20 hours/week»7 14%
Around 25 hours/week»16 34%
Around 30 hours/week»11 23%
At least 35 hours/week»13 27%

Genomsnitt: 3.63

- Most time spend on debugging algorithms and writing the report» (Around 20 hours/week)
- around 25 at the end» (Around 20 hours/week)
- The home problems take longer than you think.» (Around 25 hours/week)
- Frnlaky I think that"s absolutely good stuff.» (Around 25 hours/week)
- A lot of time is spent in implementing the algorithms, which might not be necesary to grasp the main concepts.» (Around 30 hours/week)
- Varied quite heavily depending on deadlines here and in other courses.» (Around 30 hours/week)
- I think I have spent 30 or 35 hours per week easily, and it should have been even more if I wanted to get everything right in the examples sheets. I just didn’,t have that extra time.» (Around 30 hours/week)
- The study load has been very high with difficult tasks all throughout the course period.» (At least 35 hours/week)
- It took up most of my time.» (At least 35 hours/week)

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

47 svarande

0%»0 0%
25%»4 8%
50%»2 4%
75%»15 31%
100%»26 55%

Genomsnitt: 4.34

- I prefer learn from textbooks rather than lectures, hence the low attendence rate.» (25%)
- I ways away one week, beside that i attended more or less all of the teaching.» (75%)
- most of the lectures but fewer of the excersices» (75%)
- Missed some 2-3 lectures for various reasons, generally not voluntarily.» (75%)
- Had to skip a lecture to do an assignment» (75%)
- Frnlaky I think that"s absolutely good stuff.» (75%)
- One must attend the classes not to miss vital information that is different from that which is specified in the slides given.» (100%)
- Lectures are inspiring, but sometimes more details would be useful when solving the home problems.» (100%)
- missed one lecture» (100%)
- I missed only one lecture because I was working on a homework assignment and had to finish on time.» (100%)
- The teaching complemented the lecture notes well. Both were useful.» (100%)

Goals and goal fulfilment

The course syllabus states the course goals in terms of learning outcomes, i.e., knowledge, skills and attitudes to be acquired by the student during the course.

3. How understandable are the course goals?

47 svarande

I have not seen/read the goals»16 34%
The goals are difficult to understand»1 2%
The goals give some guidance, but could be clearer»15 31%
The goals clearly describe what I am supposed to learn»15 31%

Genomsnitt: 2.61

- Never searched for them» (I have not seen/read the goals)
- no more than "understand what how the excersice sheets are solved and how they"re related to the lectures"» (I have not seen/read the goals)
- Please, put the grade limits on the course home page next time! Students should know from the start what points are required for different grades.» (The goals are difficult to understand)
- Frnlaky I think that"s absolutely good stuff.» (The goals give some guidance, but could be clearer)

4. Are the goals reasonable considering your background and the number of credits?

Answer this this question and the succeeding one, only if you do know the course goals.

34 svarande

No, the goals are set too low»0 0%
Yes, the goals seem reasonable»31 91%
No, the goals are set too high»3 8%

Genomsnitt: 2.08

- I don"t remember the goals, so cannot answer this question.» (?)
- Frnlaky I think that"s absolutely good stuff.» (Yes, the goals seem reasonable)
- Have no idea » (Yes, the goals seem reasonable)
- Had difficulty with the ties to statistical mechanics and physics.» (No, the goals are set too high)

5. Did the examination assess whether you have reached the goals?

37 svarande

No, not at all»0 0%
To some extent»10 27%
Yes, definitely»15 40%
I don"t know/have not been examined yet»12 32%

Genomsnitt: 3.05

- We hadn"t got any comments on one example sheet when handing in the next so we couldn"t use what we have learned. » (To some extent)
- I have the impression that I have learned more than the mark I have got in some examples sheet. But I understand that sometimes it is difficult to evaluate that in a better way.» (To some extent)
- We have not yet compleated the all of the example sheets,but i think they are a very good form of examination.» (Yes, definitely)
- The examination repots were hard work, but truly teached the strenths and weaknesses of neural networks» (Yes, definitely)
- The exercise sheets were mostly good, however the 5th one seems a bit of a letdown. Nothing particularly new and due in the normally sacrosanct exam week (on the same day on which I have an exam in a different course, no less). It"ll require the production of text, charts and figures that it seems, on reading it at least, wont teach me anything new. I"d rather spend that time on much needed exam crunching. Apart from that one caveat the assignments strike me as having been very effective benchmarks and a good way of learning the material.» (Yes, definitely)
- Understanding the theory, implementing in the computer, interpreting results, writing them up... everything helps my understanding.» (Yes, definitely)
- Frnlaky I think that"s absolutely good stuff.» (Yes, definitely)
- All our home problems haven"t been graded yet.» (I don"t know/have not been examined yet)

Teaching and course administration

6. Comparing take-home vs regular exam do you:

47 svarande

Much prefer regular exam»3 6%
Prefer regular exam»3 6%
Have no opinion»8 17%
Prefer take-home exam»25 53%
Much prefer take-home exam»8 17%

Genomsnitt: 3.68

- Frnlaky I think that"s absolutely good stuff.» (Prefer regular exam)
- There is no exam in this course...» (Have no opinion)
- If the form of examination is balanced in terms of difficulty and covered material, either one is acceptable.» (Have no opinion)
- I tend to do better on a regular exam, but I think that I learn much more from a take-home.» (Prefer take-home exam)
- I do prefer the type of exercises which you can take your time with, although in this course the load has been too high.» (Prefer take-home exam)
- I think that the model with the examples sheets is a good way of examination.» (Prefer take-home exam)
- the exercise sheets are good.» (Prefer take-home exam)
- It depends on the course. Why is this not an alternative? I generally prefer regular exams.» (Prefer take-home exam)
- I do prefer this form of examination in principle, however the feedback time should be shortened for this to work truly well.» (Prefer take-home exam)
- in this course it is preferable» (Prefer take-home exam)
- Manny of the tasks where to time consuming.» (Prefer take-home exam)
- The workload is much better distributed over the period» (Much prefer take-home exam)
- One single Examination in this course would not make too much sense» (Much prefer take-home exam)
- This helps one to stay within the shedule as you ar "forced" to learn during the course and not just the las weeks.» (Much prefer take-home exam)
- Individual assignments is a very good examination method. There is no way to escape the hard parts when one has to describe everything. It"s impossible to be vague (in contrast to on a regular exam). Every detail becomes clear when implementing methods in the computer.» (Much prefer take-home exam)
- In a subject like this the whole point is to be able to program and do simulations.» (Much prefer take-home exam)

7. To what extent has the teaching been of help for your learning?

47 svarande

Small extent»4 8%
Some extent»12 25%
Large extent»23 48%
Great extent»8 17%

Genomsnitt: 2.74

- We barely got any help at all in fact, and that was quite a problem.» (Small extent)
- The teacher mainly distracted me, even the sources he introduced were not useful and too old ones specially the main source. The worst part is his lecture note which aweful is not enough for that. Unfortunately, the teacher was expalning the course material for himself and he always was trying to answer his own questions in his mind, not the questions may be student"s questions toward their understanding the whole material. Plus, he wasted most of the time of the lectures with some unuseful proofs which just distracted and make whole idea of the course more vague. I mean the techer tried to focus on just 10-20% trivial part of the course than lots of interesting things in this course. I liked so much this course but by attending in his lectures i started to dislike it because of the too poor presentation of the whole course.» (Small extent)
- Had a difficult time sometimes understanding how the current lecture content tied in with what we were meant to be covering. Also confusing when variable naming (symbols) and indexes changed half way through the lecture.» (Some extent)
- It has been very good to attend the lectures.» (Large extent)
- Lectures have been generally good, however perhaps a mite to concerned with analyzing networks and not sufficiently concerned with efficiently implementing them for my tastes. It should perhaps be noted here that I"m visiting from one of the CS masters. The expected course on the subject is "Machine Learning and Inference" given by the CS department rather than this one, however the timing simply worked far better with ANN.» (Large extent)
- Lectures help a lot with the broader perspective.» (Large extent)
- SOme of the teaching, has been a bit too dificult mathematically for me, though when doing the examples sheets I have been able to understand it while consulting the lecture notes» (Large extent)
- Frnlaky I think that"s absolutely good stuff.» (Large extent)

8. To what extent has the course literature and other material been of help for your learning?

46 svarande

Small extent»9 19%
Some extent»12 26%
Large extent»20 43%
Great extent»5 10%

Genomsnitt: 2.45

- The lecture notes are very good, maybe they could be available online as well? I bought the book and my intention was to read it but it was hard to dive into the book when hopfield networks was something that was introduced in the book when general neural networks was alredy defined» (?)
- Did not buy course literature» (Small extent)
- The lecture notes was great though!» (Small extent)
- I highly recommend and I wanted to steal and burn the main lecture note of the teacher to do this honor to him to re-write and study new books and change his old and less applicable lecture note.» (Small extent)
- the lecturenotes was the only literature i used.» (Small extent)
- The course book (haykin) was a little messy. A better would have been nice.» (Some extent)
- The lecture notes helped a bit.» (Some extent)
- Very good lecture notes. Combined with wikipedia articles it made it easy to learn the parts covered in the week i missed.» (Large extent)
- The slides given are a good source to find different algorithms and equations.» (Large extent)
- I did not buy the book. But the lecture notes handed out were great» (Large extent)
- The book has been a good and concise reference, although not exactly something I"ve been devouring for the fun of it. The lecture notes are good, although I must admit that there are some errors in them that have tripped me up. See: (*) on p. 172.» (Large extent)
- Hertz" book is an excellent read. Too bad it"s not in print anymore. Lecture notes are obviously similar to Hertz, and a good way to remember lectures.» (Large extent)
- Frnlaky I think that"s absolutely good stuff.» (Large extent)

9. How well did the course administration, web page, handouts etc work?

46 svarande

Very badly»2 4%
Rather badly»10 21%
Rather well»25 54%
Very well»9 19%

Genomsnitt: 2.89

- As stated above, we have to recieve comments on our hand-ins to know what to correct for the next hand-in. There are also very few actual questions in the example sheets, it just says, "implement" and similar. What do you want to know from us? If it is something that holds for all of the course, it should be part of the teaching. » (Very badly)
- there is sometimes difficuty accessing the exercise sheets of the webpage. Also: I was unable to find the lecturenotes on the website» (Rather badly)
- Web page was a bit slow on the uptake and scheduling has been generally quite messy. Requiring physical paper hand-ins seem vaguely pointless to me and a waste of both time and perfectly good trees.» (Rather badly)
- The web page is tiny and contains very little information. Check the CSE web pages (TIN092 for example) to see what I mean. The handouts should really be in digital form, rather than printed scanned handwriting.» (Rather badly)
- There were initially some confussion regarding how many examples sheets there were and how much points one needed to pass. I believe that it should, from the beginning of a course, be clear when and what deliverables is needed of a student to pass the course» (Rather badly)
- It felt like the example sheets did not contain the information needed inorder to complete the problems in the way Marina desired. A lot of additional information was given during the thursday example classes. Which plots one should make, which parameters that was good and so forth. It made it hard for people who did not attend these session to complete the tasks.» (Rather badly)
- It would be helpful if there was a schedule showing what chapters and parts of the lecture notes are planned for different weeks. The SOA homepage was excellent in this regard, maybe you could get some inspiration from that.» (Rather badly)
- Corrected reports were returned too late, giving students no possibility to learn from previous mistakes. » (Rather badly)
- I would like it to be easier to get my results on the hand-outs. For example, you could bring the reports to a lesson.» (Rather well)
- VERY useful with the lecture notes, so thats something you should keep handing out :) The examples sheet were corrected very slowly, and most annoying of all: The specification of the tasks are often very vague. It was often hard to know what was to be done, or what type or results to include in the report.» (Rather well)
- One needs feedback from the examples sheets much faster. This in order to improve the writing of them.» (Rather well)
- Web page update was a bit late once or twice. Examples sheet 4 was handed out only a week before deadline... this is a problem since we have assignments in other courses as well.» (Rather well)
- Although the handouts were difficult to understand sometimes. (poorly copied or bits missing)» (Rather well)
- Frnlaky I think that"s absolutely good stuff.» (Rather well)
- I did not like having the deadline on tuesdays, because you can not ask questions during the weekend. It would be better if they were on fridays.» (Rather well)

Study climate

10. How were the opportunities for asking questions and getting help?

47 svarande

Very poor»1 2%
Rather poor»5 10%
Rather good»17 36%
Very good»15 31%
I did not seek help»9 19%

Genomsnitt: 3.55

- Most of the time I asked something about example sheets it was said that "thats part of the question to figure out". So thats basically it. » (Very poor)
- Even though the exercise sessions were good, they were very unclear and contradicting against what was given in the exercise sheets.» (Rather poor)
- Did not get much help from the questions. "Test yourself"» (Rather poor)
- Marina and Bernhard are very helpful.» (Rather good)
- I would have liked feedback on the examples sheets throughtout the course, now we only received feedback on the first examples sheet and it was impossible to know how well one was doing, when writing the other examples sheets» (Rather good)
- I think the opportunities were ok but didn"t feel that approachable (I wasn"t sure if my questions would be "ok" and it felt like others had this problem.). The timing of Marina"s "opening hours" was quite poor as a lot of the time I had most questions the day before or on the same day that the project was due. Also, is one hour a week really sufficient for 60-odd students?» (Rather good)
- could have used more office hours for questioning.» (Rather good)
- Maybe there were too many students for just one course assistant» (Very good)
- Frnlaky I think that"s absolutely good stuff.» (Very good)
- The twoish times I"ve had questions they"ve been promptly answered, but that"s not much basis for an opinion.» (I did not seek help)

11. How well has cooperation between you and your fellow students worked?

47 svarande

Very poorly»0 0%
Rather poorly»7 14%
Rather well»10 21%
Very well»27 57%
I did not seek cooperation»3 6%

Genomsnitt: 3.55

- We worked alone» (Rather poorly)
- Frnlaky I think that"s absolutely good stuff.» (Rather poorly)
- We weren"t allowed to cooperate.» (Rather poorly)
- Did check in what others was doing and drawing ideas from that but only to a limited extent» (Rather well)
- No complaints.» (Very well)
- I think this should be encouraged more (especially at the beginning of the course).» (Very well)
- without them i would not have passed the course» (Very well)

12. How was the course workload?

47 svarande

Too low»0 0%
Low»0 0%
Adequate»9 19%
High»28 59%
Too high»10 21%

Genomsnitt: 4.02

- Medium-high, but generally effective. Again, no complaints.» (Adequate)
- Frnlaky I think that"s absolutely good stuff.» (Adequate)
- Writing the reports was time-consuming.» (High)
- As is often the case with courses that have a lot of delivarbles during the course. But I think it was a good thing. » (High)
- Especially considering when projects from other courses were due. Also, there is very little time for error, if you had a problem then you fell behind and the other projects fell behind... I would have liked the 7pm deadline to not be a strict one (was it strict?) and also to know if /how many points would be deducted if handed in late. A clearer hand in late procedure would have helped as you could have planned things a bit more if you happened to fall behind.» (High)
- The workload is higher then other courses with the same amount of HP. As I understood the course contained five example sheets this year and four last year. This also add an extra 20+ hours to the workload without adding more HP, seems strange to me.» (High)
- In my opinion, four example sheets are enough.» (High)
- During my five years of Chalmers Ive gone through taking all courses seriously, this was yet the highest workload so far. I spent most of the days to the home assignments, days, evenings weekends... I was lucky to have an easy course besides this one because this course alone felt like 2 courses...» (Too high)
- Too long simulations were required. I"d rather see shorter example sheets or shorter simulations. Also, ES 5 was almost identical to ES 3.» (Too high)
- One problem with the high workload was that I felt I wasn"t able to assimilate current lecture material properly, since I was working on problems based on previous chapters. This meant falling behind on course material and having to review considerably in order to complete the next assignment and so on, and the uncertainties accumulated.» (Too high)
- While it was possible to finish the assignments on time we did not have enough time to think about the implications of what we were doing.» (Too high)

13. How was the total workload this study period?

46 svarande

Too low»0 0%
Low»0 0%
Adequate»12 26%
High»26 56%
Too high»8 17%

Genomsnitt: 3.91

- Only because the workload in my other course (not stochastic optimization) was under average.» (Adequate)
- Frnlaky I think that"s absolutely good stuff.» (Adequate)
- Not too high, I like the courses so then it"s no problem to spend time on it.» (High)
- High, but that"s mainly because I have some additional workload as a teaching assistant for 10ish hours a week. I can"t really blame you guys for that.» (High)
- It was a little bit high because of all the simulations one had to do. Simulations are good of course, but for instance, in examples sheet 3 you had to make a LOT of plots, and you did not learn anything at all when you ran the 40th simulation. You got the idea much much earlier, so it was a pain. ES 5 was also like that.» (High)
- Reading stochhastic optimization beside neural network resulted in no free time whatsoever. » (High)
- In addition to another high-load course, this period has had very high load.» (Too high)
- I believe the homework for this course negatively impacted the time left to prepare for other courses, leaving me little time to prepare for the exam or submit full assignments for those.» (Too high)
- But I was not following the regular courses of the study plan.» (Too high)

Summarizing questions

14. What is your general impression of the course?

47 svarande

Poor»2 4%
Fair»3 6%
Adequate»4 8%
Good»30 63%
Excellent»8 17%

Genomsnitt: 3.82

- Just because of the teacher.» (Poor)
- Frnlaky I think that"s absolutely good stuff.» (Fair)
- More topics related to applications should be discussed instead of going so deep at some theoretical issues, because these help not linking them to real problems does not help to clarify them.» (Adequate)
- I think the course was quite good but there was a lot of background information that wasn"t mentioned as background information (and therefore made it a bit tricky to understand how to use it, especially at the beginning when you think everything is important).» (Adequate)
- Artificial neural networks is a very very interesting subject, unfortunately I missed a lot from this course. This is what we learned to program: Hopfield model, Backpropagation, Kohonen algorithm. We did not do anything on recurrent neural networks, support vector machines, reinforcement learning and other algorithms I don"t know about. We only touched a small small fraction of this subject and the Haykin book.» (Adequate)
- In general, the course was very good. The only thing which was bad was the rate at which the exercise sheets were graded, which took WAY too long time.» (Good)
- It was good because I learnt much but too high workload.» (Good)
- Generally very interesting» (Good)
- Very good if, as noted, somewhat analytical. I would personally have appreciated glossing over some of the detailed derivations in favor of discussing strategies for efficient implementation. Some of the lectures that were entirely devoted to examining, say, a single stability criteria did not feel like time optimally spent.» (Good)
- I study the IT-programme and I am maybe more interested in focusing more on how and when one uses what, rather than goind in to mathematical details on the proofs. But it was interesting as well» (Good)
- I enjoyed the topic very much. Example Sheets instead of exam suited me perfect.» (Good)
- The style of the lectures is very appealing and enlightening, if sometimes not very precise or organized. I also enjoyed having to research and figure out things for myself. My only complaints are the too large amount of time I had to allot to this course, including lack of rest time, and the lack of comprehensive reference for the home assignments.» (Good)
- Great, enthusiastic teacher! » (Good)
- Any course becomes interesting with an enthuiastic lecturer and especially this one.» (Excellent)
- Became rather hard towards the end but overall I enjoyed this course.» (Excellent)
- Interesting, fun, a bit challenging. I learned both about neural networks, and more general about simulation/research/presentation.» (Excellent)

15. What should definitely be preserved to next year?

- The inspired lectures and the hand-ins.»
- The course :)»
- the lecture notes!»
- The exam method. The kind of questions on the example sheets.»
- The scope of the home problems.»
- the exercise sheets»
- Hertz course book»
- examination sheets, handout of lecture, motivation of course staff»
- Bernhard.»
- The book, lectures and most of the exercises.»
- I think that the exercise classes should be preserved. »
- The ideas behind the course. In-depth mathematics of how those networks work might be a turn-off at first, but really it sticks once you get it.»
- Examples sheets»
- Change the teacher or beg him (he is the king in this field and no one can change him, unfortunately) to change his way of teaching and focusing more on the applications, real examples than just proofs and using the old fashioned material (1991, the book he is still using). I highly recommend him to see new books of this course and see how the idea is evolving and how he is destroying the main aim of the course. There are lots of online courses which presented by other teachers from other universities, even one of the indian ones was much much much better than him and his way. If there were someone who did not have any idea and information about the Mobile phones and technology and if we wanted Bernhard to explain him the Mobile technology. After his explanation, that poor person would get to this conclusion that, the mobile is a kind of calculator for doing the physical math or a device like GPS than the real mobile itself. I do admit that Bernhard is so knowledgeable person, but he can not could not been able to express and transfer his knowledge to other people. It is much better to use him for P.h.D courses or in research groups than introducing and teaching courses like this.»
- The lectures are excellent. They bring a lot of insight that is not possible to get from the book. Examples sheets is how we learn all the details. Very good, no nonsense.»
- The examination form. Sometimes it felt like most of my time was spent doing the charts, but programming the networks and not just reading about them is very good »
- the hand ote compendium»
- The lecture notes and the lecture books were very helpful. »
- Frnlaky I think that"s absolutely good stuff.»
- The examination form.»
- The teacher. :)»
- course literature»
- It is good that lecture notes are handed out.»
- The examples sheets are really interesting, for the most part (last one, not so much, it"s too similar to the third one).»
- Bernhard Mehlig.»

16. What should definitely be changed to next year?

- Some lessons, above all the concluding ones on the hopfield model, could have been more well prepared, making it easier to follow.»
- Recieving the corrected sheets before the next is due so that one knows what to expect in terms of grading.»
- The examples sheets should be more well definet in that sense that when one has made some working implementation of a problem, one should know what information (plots, data) to include in the results of the report.»
- Much faster correction of the example sheets. It is very bad to not get your points on previous example sheets before the handout of the next.»
- The correcting of home problems could be quicker, could two people do this? Feedback is useful and would be appreciated»
- Help the students and dont say things are meant to be figured out, this is no easy subject. Furthermore, many lectures were way too theoretical and didn´,t relate to the example sheets in a practical way, there were theoretical reasonings that were very advanced but nothing to really apply sometimes.»
- stimulate more group work from the beginning»
- The questions for the examples sheets could be a bit more specific and aimed at understanding rather than performing. Less "Do this:", more "Do this and reflect on..." or "Doing this, why should you...".»
- Organize better to get to the topics about Recurrent networks and time series analysis. »
- Think the example sheets should be somewhat clearer on what you should do, or the excersise lectures could have better explained. In some instances I was lost on what I should do.»
- Need more exercises! Need more opportunities to practice and to go through what we learned in the classes.»
- 1: More course assistants. A quicker feedback cycle and letting me see the grading of one sheet before the next one is due would both let me improve my work and help me learn the material. As is it looks like much of the feedback will arrive a week or two after the end of the course. 2: Do away with the mandatory paper hand-ins. 3: Clarify the requirements on the assignments and make sure the same requirements are stated on the sheet and on the preparatory lecture. Sometimes crucial information is relayed only on the lecture (i.e. "plot results for both activation functions" for sheet 3) and sometimes the lecture and sheet state different things (plots on sheet 3, again). This could harmonize far better.»
- I think that 5 example sheets is too much.»
- Improvements to handouts and the webpage.»
- The lectures needs a better connection to what the students are doing in the examples sheets. Sometimes it feels like the lectures are there to show how much you can do with neural networks and that it is a very complex topic. This is not tested/examined at all, because it is not part of the examples sheets. Also, the examples sheets needs to be corrected faster. Why not having two phd:s correcting them?»
- 1. Teacher 2. Teacher 3. Teacher 4. Lecture Note 5. Lecture Note 6. The way of teaching 7. Course material»
- Get a bigger lecture room from the beginning. Something like FB or HC2 is very good. If it"s possible, please provide reading instructions (page references) on the web page before lectures.»
- Please give some feedback on the examples sheets during the course»
- later hand in dat fore the task»
- I would like to get faster respons on the home problems so that it"s possible to learn from previous mistakes. In this case we just got fast response from the first one and the all other ones were left until the end of he course.»
- Digital hand in only! Also I think a bit more time for the final hand in (Ex.5) would be nice.»
- 3-4 example sheets and an exam seem better evaluation for me. perhaps more exercises that require demonstrations and analytical thinking rather than just programming tasks. »
- Plan a spare lecture or two from the beginning so the theory doesn"t cram up in the end. We addressed Kohonen Network the day before the Example Sheet on these was due which is too close.»
- I might be good idea to actually grade the example sheets. it is study week 7 now and we still havn"t got back example sheet 2 yet»
- Frnlaky I think that"s absolutely good stuff.»
- The lectures on the Kohonen network were only finished one day before the examples sheet on this topic was due in. This could be placed earlier.»
- The example sheets need to be look over.»
- The information needed for completing the examples-sheets is taught a very short time before the deadline (might be hard to change).»
- I believe a lighter workload would be necessary. Perhaps only one (more difficult) problem per home assignment each week. I understand the need to cover as much as possible of the (diverse) topics in the course, but perhaps it would be possible to compress the topics in fewer problems. Furthermore, it would be of great help to be able to have reference (typical solution or expected outcome) for the home problems, even after the deadline, especially since some depend on previous assignments. This dependency is a problem if the necessary problems were inadequately solved or left unclear. The problem I faced was being presented with a certain behaviour in the programs I made and not being able to decide if, indeed, there was a problem to begin with (if what I was seeing was expected), or if the problem was in the coding or the theory. It turned out most of the issues were with my application of theory, but this would be clear frequently only after grading. Up until that point I could only speculate, without obvious reference results to consider.»
- Website, clearer guidance and detailed introduction and programming basics. »
- The examples sheets were not fun, except ES4 and ES2. Come to think of it, the first 2 were both on the Hopfield model, and the last 3 all included back-propagation! Ridiculous. And in ES3 there were too many plots. There should be a lot more variation. In the course stochastic opt algorithms the problems were a lot more interesting. Why not include fun and "real" problems like word recognition or recognising handwriting. The hopfield model is used for pattern recognition, but I never learned how well it would perform to recognise images (I guess it would do very poor), which algorithms are used for image recognition? In short: more fun problems and more variation.»
- Please put grade limits and a detailed course plan on the homepage.»
- An examples sheet must be corrected and returned before the next one. »

17. Additional comments

- I"m a MPALG student and as such my mathematics skills are not even close to being enough for this course.»
- Started my third year on IT and this course required more mathematical knowledge than I have, that you probably get after three years on F. Much in the lectures is expected of you to understand.»
- I"m still not really sure to what extent the course is intended to be an algorithms and machine learning course and to what extent it is to be a modeling of natural systems course, to whichever extent a dichotomy exists. It"s certainly available to CS majors (and IT undergrads, even!) and if that"s to be accommodated perhaps some of the tasks should be specified in more detail. Note that I certainly don"t mean simplified, none of the tasks in hand-ins contain anything particularly troublesome. However, there seems to be an assumption that the person reading them is familiar with, say, magnetic dipoles and the presentation of experimental data. Physicists certainly are, computer scientists perhaps not so much. Alternatively, perhaps you should make sure that the CS department are nice enough to run Machine Learning and Inference in study period 1 and send all the programmers that way.»
- Our teacher in this course is so knowledgeable but he did not have the knowledge of teaching, unfortunately. Meanwhile, somebody must tell him that the main usage of the neural networks is not in physics and he must not present the course only in favor of the physics and let others to use it in other fields.»
- Bernhard is good at drawing pretty pictures on the blackboard.»
- Difficult to see what Marina wrote on the black board sometimes (small writing).»
- Frnlaky I think that"s absolutely good stuff.»
- I had a lot of trouble with debugging the programs because the expected results were not known in advance. The course structure is very different from what i am used to, so it took some time to get used to it.»
- I enjoyed having the possibility to reseach and to figure out things for myself in order to solve the assignments. And I understand that it is realistic to not have answers to fall back to in research or real life situations, but this becomes a source of frustration when the grading for the course depends greatly on what could be idle speculation on my part. While there may be no clear-cut solutions in some situations, attempts at solving could be given higher merit, if, of course, they are presented with justification for the choices made.»

Kursutvärderingssystem från