Menu
▾
▴
memaid-users — General discussion and help for MemAid
You can subscribe to this list here.
| 2003 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
(19) |
Dec
(9) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2004 |
Jan
|
Feb
|
Mar
|
Apr
|
May
(1) |
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
|
From: <ben...@id...> - 2004-05-25 08:31:12
|
Dear Open Source developer I am doing a research project on "Fun and Software Development" in which I kindly invite you to participate. You will find the online survey under http://fasd.ethz.ch/qsf/. The questionnaire consists of 53 questions and you will need about 15 minutes to complete it. With the FASD project (Fun and Software Development) we want to define the motivational significance of fun when software developers decide to engage in Open Source projects. What is special about our research project is that a similar survey is planned with software developers in commercial firms. This procedure allows the immediate comparison between the involved individuals and the conditions of production of these two development models. Thus we hope to obtain substantial new insights to the phenomenon of Open Source Development. With many thanks for your participation, Benno Luthiger PS: The results of the survey will be published under http://www.isu.unizh.ch/fuehrung/blprojects/FASD/. We have set up the mailing list fa...@we... for this study. Please see http://fasd.ethz.ch/qsf/mailinglist_en.html for registration to this mailing list. _______________________________________________________________________ Benno Luthiger Swiss Federal Institute of Technology Zurich 8092 Zurich Mail: benno.luthiger(at)id.ethz.ch _______________________________________________________________________ |
|
From: David C. <ne...@gm...> - 2003-12-03 13:42:29
|
On Wednesday 03 of December 2003 12:50, Axel Kittenberger wrote: > > - cross promotion on project home pages > > Valid! I have just created a cross-link to memaid on the webspace of > pauker. Same here. :) (link to pauker.sf.net created) -- David Calinski http://neodave.civ.pl |
|
From: David C. <ne...@gm...> - 2003-12-03 13:28:01
|
On Wednesday 03 of December 2003 07:21, Ronny Standtke wrote: > There are now two ways of repeating a flashcard in Pauker, by remembering > and by typing. In the first case the user just tells the program if he > could remember the answer right. In the second case the program puts the > card automatically to a higher level if the answer was spelled correctly. > If the spelling was wrong the user can decide if it was a mistake or just a > typo. In MemAid (Fox Gui version) user can also type answer (which is good - helps remembering by giving instant feedback to user, ie. if pressed key is not the right (next on the hidden part of the answer) there is a 'beep' instead of showing next character). But final decision (about grade) belongs always only to user (there is 0-5 scale). And final decision (what will be the next "level" - interval) belongs to neural network which has many data about items (including this grade and data about past repetitions - so e.g. NN can see how difficult this item is treat it accordingly.) So we have different approaches, Pauker can save time for grading, can be better for "faster reviews" - now I better understand Andreas' points about MemAid's limitation here. > And I havent tried memaid. Maybe we should change this as soon as possible. I Agree. :) > Speaking about mutual benefits... I support everything what helps both > projects. These topics come spontaneously to my mind: > > - cross promotion on project home pages > - unified lesson format > - central lesson repository on the web > > Do you have any other suggestions? Nothing so specific comes to my mind at the moment, but as we know about each other - there is no hurry to think about it. Generally: co-operation (e.g. in pursuing file-lesson formats, new features that could be useful in both projects, etc.). Also, Pauker could use MemAid ANN engine. MemAid could use much Pauker's Java code (mainly graphical interface which is much more advanced in Pauker) to be more platform-independent. Just some ideas. -- David Calinski http://neodave.civ.pl |
|
From: Axel K. <ax...@ki...> - 2003-12-03 09:40:50
|
> Speaking about mutual benefits... I support everything what helps both > projects. These topics come spontaneously to my mind: > > - cross promotion on project home pages Valid! I have just created a cross-link to memaid on the webspace of pauker. Greetings, Axel |
|
From: Andreas G. <agr...@ya...> - 2003-12-03 09:34:48
|
On Wed, 2003-12-03 at 07:21, Ronny Standtke wrote: > Am Mittwoch, 3. Dezember 2003 02:44 schrieb David Calinski: > > > Who (user or software?) and how decides when an item go to a higher level? > > There are now two ways of repeating a flashcard in Pauker, by remembering and > by typing. In the first case the user just tells the program if he could > remember the answer right. In the second case the program puts the card > automatically to a higher level if the answer was spelled correctly. If the > spelling was wrong the user can decide if it was a mistake or just a typo. > > > Now, maybe I still miss something? (I haven't tried Pauker, and I might > > have just missed your points.) > > And I havent tried memaid. Maybe we should change this as soon as possible. Ronny, in addition to trying the software, I would humbly suggest that you as well read the documentation on the projects site and stuff at www.supermemo.com to see other concepts of staggered repetition. Regarding Pauker, a good introduction (imho) to the underlying learning concept can be found at http://www.flashcardexchange.com/leitner.php Andreas |
|
From: Andreas G. <agr...@ya...> - 2003-12-03 09:27:17
|
On Wed, 2003-12-03 at 02:44, David Calinski wrote: > On Tuesday 02 of December 2003 22:19, Andreas Grau wrote: > > (...) > In MemAid there is (in theory) countless number of "levels". Level is just > "interval", isn't it? Good point. > In practice, current MemAid is limited to maximum granulation of 1 day. Which is one of the deficits, memaid has in my eyes: When I have more time than repetitions, I can not continue with memaid. To use my time, I have to switch to new input, which is not bad in itself, but I would rather prefer to have the choice. > I see easy, though not ideal, solutions: > 1. instead of making hundred of repetitions of one item in 1 day, you should > focus and make just one repetition, but carefully and try to make sure you > really remember it. (it is always the best policy anyway - don't make > careless, very fast repetitions. *Try* to really remember every item. Sacrife > this time - it really pays off better than many careless reviews.) Again, memaid is trying to force me into a style of doing things, it doesn't give me enough lee-way to learn how I want. I am not saying that you are wrong, but this just isn't my style. I would rather have the choice to either doing 1 carefull repetition or 100 careless ones. Or even both. > There can be better solutions: This is what I am aiming at. Pauker's concept of learning plus memaid's support for recall would be a fine solution as well. Best, Andreas |
|
From: Ronny S. <Ron...@gm...> - 2003-12-03 06:21:48
|
Am Mittwoch, 3. Dezember 2003 02:44 schrieb David Calinski: > Who (user or software?) and how decides when an item go to a higher level? There are now two ways of repeating a flashcard in Pauker, by remembering and by typing. In the first case the user just tells the program if he could remember the answer right. In the second case the program puts the card automatically to a higher level if the answer was spelled correctly. If the spelling was wrong the user can decide if it was a mistake or just a typo. > Now, maybe I still miss something? (I haven't tried Pauker, and I might > have just missed your points.) And I havent tried memaid. Maybe we should change this as soon as possible. > (I don't look for any wars, nor I see it as a touchy topic. I just think > that "borrowing" ideas, solutions, and discussing about all that, can be > beneficial, for both projects, obviously.) Speaking about mutual benefits... I support everything what helps both projects. These topics come spontaneously to my mind: - cross promotion on project home pages - unified lesson format - central lesson repository on the web Do you have any other suggestions? Greetings Ronny Standtke (author of Pauker) |
|
From: David C. <ne...@ci...> - 2003-12-03 01:44:48
|
On Tuesday 02 of December 2003 22:19, Andreas Grau wrote: > (...) > > This is where "Pauker" at sourceforge.net/projects/pauker/ comes in. I > easily grasp the underlying idea and enjoy the system as it stands. It > offers repetions of items by transfering items from one level (or batch) > to the other. Only by getting the same item right several times, the > items gets "promoted" to higher levels/batches. Who (user or software?) and how decides when an item go to a higher level? In MemAid there is (in theory) countless number of "levels". Level is just "interval", isn't it? And it's neural network who decides when an item go to a higher (or lower, e.g. if you was on vacation and forgot item). Neural networks have it's own plasticity - NN can learn how user learn, and modify itself to better schedule repetitions. In practice, current MemAid is limited to maximum granulation of 1 day. I see easy, though not ideal, solutions: 1. instead of making hundred of repetitions of one item in 1 day, you should focus and make just one repetition, but carefully and try to make sure you really remember it. (it is always the best policy anyway - don't make careless, very fast repetitions. *Try* to really remember every item. Sacrife this time - it really pays off better than many careless reviews.) 2. Use "Finall Drill" feature and here you will be presented with items you don't remember until you remember them. There can be better solutions: Peter's 'drill-badly-known' feature or smart-final-drill that will show up in the future. Now, maybe I still miss something? (I haven't tried Pauker, and I might have just missed your points.) If Pauker still have something better then obviously I would like to know about it. :) (I don't look for any wars, nor I see it as a touchy topic. I just think that "borrowing" ideas, solutions, and discussing about all that, can be beneficial, for both projects, obviously.) -- David Calinski http://neodave.civ.pl |
|
From: Peter B. <Pet...@UG...> - 2003-12-02 21:50:13
|
Hi Andreas, As you know, that's precisely what 'drill-badly-known' in pyqt memaid does: it will keep presenting you the items until they get a high enough grade. After that, they will just go in the regular neural network learning process. Cheers, Peter On Tuesday 02 December 2003 22:19, Andreas Grau wrote: > Now, this might be a slippy subject I am touching. Ok, <flame on> for > who dares - I will defend myself :-) > > I recently stumbled upon "MemAid" at sourceforge.net/projects/memaid/ > and must admit that I am fascinated by the underlying concept of an > adopted staggered re-learning taught by a neural-network. > > However, I must admit that I consider memaid a poorer solution when it > comes to the initial learning process. Memaid, as I understand it, > relies on finding the best possible *INDIVIDUAL* repetition schedule for > not forgetting items. While in the process of initial learning, however, > memaid fails because of current inabilities to differentiate between > "learning" and "recalling". Personally, I feel there is a difference. > > This is where "Pauker" at sourceforge.net/projects/pauker/ comes in. I > easily grasp the underlying idea and enjoy the system as it stands. It > offers repetions of items by transfering items from one level (or batch) > to the other. Only by getting the same item right several times, the > items gets "promoted" to higher levels/batches. > > Now, my wildest dreams would come true if there were a bridge between > the two: Using Pauker 'till, say, Batch 3, when items have survived my > short- to mid-term memory deficiencies :-) These items I would like to > hand over to MemAid to keep them stable in my memory for eternity (or > until my harddisk dies - whichever comes first :-) > > I'd like to get a discussion going on the differences between learning > and recalling, and where the projects have their individual strengthes > and weaknesses. I am aware it's a touchy subject, as project initiators > and their followers usually are quite sentitive towards their respective > "babies" - but let's keep it rollin' anyway. Remember: it's "Free, as in > freedom" > > Best, Andreas > > > > ------------------------------------------------------- > This SF.net email is sponsored by OSDN's Audience Survey. > Help shape OSDN's sites and tell us what you think. Take this > five minute survey and you could win a $250 Gift Certificate. > http://www.wrgsurveys.com/2003/osdntech03.php?site=8 > _______________________________________________ > Memaid-users mailing list > Mem...@li... > https://lists.sourceforge.net/lists/listinfo/memaid-users -- ------------------------------------------------ Peter Bienstman Ghent University, Dep. of Information Technology Sint-Pietersnieuwstraat 41, B-9000 Gent, Belgium tel: +32 9 264 34 45, fax: +32 9 264 35 93 WWW: http://photonics.intec.UGent.be email: Pet...@UG... ------------------------------------------------ |
|
From: Andreas G. <agr...@ya...> - 2003-12-02 21:19:00
|
Now, this might be a slippy subject I am touching. Ok, <flame on> for who dares - I will defend myself :-) I recently stumbled upon "MemAid" at sourceforge.net/projects/memaid/ and must admit that I am fascinated by the underlying concept of an adopted staggered re-learning taught by a neural-network. However, I must admit that I consider memaid a poorer solution when it comes to the initial learning process. Memaid, as I understand it, relies on finding the best possible *INDIVIDUAL* repetition schedule for not forgetting items. While in the process of initial learning, however, memaid fails because of current inabilities to differentiate between "learning" and "recalling". Personally, I feel there is a difference. This is where "Pauker" at sourceforge.net/projects/pauker/ comes in. I easily grasp the underlying idea and enjoy the system as it stands. It offers repetions of items by transfering items from one level (or batch) to the other. Only by getting the same item right several times, the items gets "promoted" to higher levels/batches. Now, my wildest dreams would come true if there were a bridge between the two: Using Pauker 'till, say, Batch 3, when items have survived my short- to mid-term memory deficiencies :-) These items I would like to hand over to MemAid to keep them stable in my memory for eternity (or until my harddisk dies - whichever comes first :-) I'd like to get a discussion going on the differences between learning and recalling, and where the projects have their individual strengthes and weaknesses. I am aware it's a touchy subject, as project initiators and their followers usually are quite sentitive towards their respective "babies" - but let's keep it rollin' anyway. Remember: it's "Free, as in freedom" Best, Andreas |
|
From: David C. <ne...@ci...> - 2003-11-30 21:33:45
|
On Sunday 30 of November 2003 21:40, Peter Bienstman wrote: > On Sunday 30 November 2003 20:06, David Calinski wrote: > > But I would like to focus on commercial version for Windows > > Oops, as you know I was planning a free Windows version of PyQt MemAid. > I've already done some experiments on how hard it would be to backport it > to Qt 2.x which is needed for the free edition of Qt under Windows. Turns > out it's not that hard, but I don't want to do these changes everytime I > make a new release, so I was planning to wait until I've implemented all > the features I want in the main code, which won't happen very soon. > > If you want, we can discuss this further privately. Feel free to release anything you want, on platform you want. I don't mind it, just the opposite. Especially if it will be free on GPL license. I will try to never block development of Open Sourced version. And I will be happy if there will be free MemAid version for Windows. (I expect that my commercial version, if I will make it, will be better, have support, great documentation (maybe a book about learning), etc. so some people would still choose to pay money for it. But if not, and your open sourced version will be better, then it will be only my *personal* problem (I would have to work harder to create something better, so it would be good for end-users after all - and this is the most important).) -- David Calinski http://neodave.civ.pl |
|
From: Peter B. <Pet...@UG...> - 2003-11-30 20:40:23
|
On Sunday 30 November 2003 20:06, David Calinski wrote: > But I would like to focus on commercial version for Windows Oops, as you know I was planning a free Windows version of PyQt MemAid. I've already done some experiments on how hard it would be to backport it to Qt 2.x which is needed for the free edition of Qt under Windows. Turns out it's not that hard, but I don't want to do these changes everytime I make a new release, so I was planning to wait until I've implemented all the features I want in the main code, which won't happen very soon. If you want, we can discuss this further privately. Cheers, Peter ------------------------------------------------ Peter Bienstman Ghent University, Dep. of Information Technology Sint-Pietersnieuwstraat 41, B-9000 Gent, Belgium tel: +32 9 264 34 45, fax: +32 9 264 35 93 WWW: http://photonics.intec.UGent.be email: Pet...@UG... ------------------------------------------------ |
|
From: Andreas G. <agr...@ya...> - 2003-11-30 20:07:51
|
Forgot to say, that the referenced article and the quote below is written by the guru himself. <Snip> Strictly speaking, ``selling'' means trading goods for money. Selling a copy of a free program is legitimate, and we encourage it. </Snip> On Sun, 2003-11-30 at 20:58, Andreas Grau wrote: > On Sun, 2003-11-30 at 20:54, Andreas Grau wrote: > > > Problems is... well, for example in free software world some people consider > > > writting commercial software a crime. :) > > > Sounds funny, but it's true. Richard Stallman is for many people a guru, and > > > it's his opinion. > > http://www.gnu.org/philosophy/selling.html > > <Snip> > > The word ``free'' has two legitimate general meanings; it can refer > either to freedom or to price. When we speak of ``free software'', we're > talking about freedom, not price. (Think of ``free speech'', not ``free > beer''.) Specifically, it means that a user is free to run the program, > change the program, and redistribute the program with or without > changes. > > Free programs are sometimes distributed gratis, and sometimes for a > substantial price. Often the same program is available in both ways from > different places. The program is free regardless of the price, because > users have freedom in using it. > > </Snip> > > Go ahead David, and make tons of money ! |
|
From: Andreas G. <agr...@ya...> - 2003-11-30 20:01:45
|
On Sun, 2003-11-30 at 20:06, David Calinski wrote: > > Problems is... well, for example in free software world some people consider > writting commercial software a crime. :) > Sounds funny, but it's true. Richard Stallman is for many people a guru, and > it's his opinion. http://www.gnu.org/philosophy/selling.html <Snip> The word ``free'' has two legitimate general meanings; it can refer either to freedom or to price. When we speak of ``free software'', we're talking about freedom, not price. (Think of ``free speech'', not ``free beer''.) Specifically, it means that a user is free to run the program, change the program, and redistribute the program with or without changes. Free programs are sometimes distributed gratis, and sometimes for a substantial price. Often the same program is available in both ways from different places. The program is free regardless of the price, because users have freedom in using it. </Snip> Go ahead, David, and make tons of money ! |
|
From: David C. <ne...@ci...> - 2003-11-30 19:06:31
|
On Sunday 30 of November 2003 19:13, Andreas Grau wrote: > > I am not sure if I get you here, but I definitely don't agree with last > > sentence. You shouldn't have to "enforce" anything. > > It's MemAid role to guess how difficult every item is and schedule > > accordingly! > > I agree that we disagree. Our main difference is that you are talking > about difficulty, and I am *also* talking about importance. I agree about importance. But importance is highly related to retention rate. (retention rate - percent of things you remember). In fact, for important items you care about high retention rate. As I said, I would like to see "aimed retention rate" option for every category (when categories will be implemented.) Higher the 'aimed retention rate', more frequent MemAid will schedule repetitions. (so you could set higher retention rate for something that is important for you. Retention and importance is almost like synonyms here.) I do want to put all work, as much as possible, to ANN. And I don't want to think "oh, I need to do some more repetitions from this category now". I prefer to think that ANN worries about such things. :) If I really need to do some more repetitions from particular category - it's ANN-MemAid who should tell me about that - simple schedule these repetitions. > During a (hopefully) normal and structured learning process, memaid > helps a lot. No dissent here. > > But what, if I absolutely and unconditionally need some knowledge > tomorrow? I understand. But MemAid simply won't help much here. Optimum spacing of repetitions shows its strength over longer periods. If one needs to learn (cram) for an exam tomorrow, then the only thing in MemAid that would be helpful is "active recalls". The main advantage (ANN and optimum repetitions) is simply lost. In other words: although I admit sometimes there is a need to cram something, it will never be main MemAid aim to be such cram tool. But categories and "Forced repetitions mode" can help in such cramming. > Ok, tomorrow might be a bit too exagerated. Let's say in a > month's time I have an exam on some stuff. How to make sure, exams' > contents will be covered by memaid. Again, the problems lies partially > in difficulty. but mainly in importance. I would happily forget some > other items (scheduled for repetition) if I would recall the items at > hand for the exam. Yeah. Here categories could be helpful. > > Especially new code I am working on - among other things it will have a > > smart "Finall Drill" feature that will work automagically, and there will > > be no normal "Finall Drill" - everything should be automated and it is > > MemAid 'who' should worry about enforcing and proper scheduling > > repetitions, not human. > > Agree here. But still then: Man overrides Machine - Assimov's 2nd law of > robots? - not sure, will have to learn again. Sigh. ;-) Hehe. :) I used a wrong word. Obviously MemAid will never "enforce" anything. :) Hmm. "suggest", or just "schedule", sounds much better. :) > I am not sure I understand this correctly. Is "higher retention rate" > synonymus to "prefered item to learn, overriding other items in > priority, normally scheduled for today" ? More or less. Well, in SuperMemo software (something like MemAid, but doesn't use ANN) in options you can find "retention rate". By default is set to 90%, so SuperMemo algorithm tries to scheduled items in a way that 10% of items on next repetition will be forgotten. This is, more or less, optimal. (100% would be achieved if one would learn endlessly with SuperMemo, which is obviously crazy and impossible. MemAid/SuperMemo aim is to have the most optimal knowledge_acquisition/time_spend_for_learning ratio.) To put it more simply: MemAid will schedule more frequent repetitions of items that have higher "aimed retention rate". Which means that these items (with higher "retention rate") are more important for user and user is willing to pay higher price(more time spend on them) for keeping them in memory (as a result more of them user will remember) > Sorry to hear about your financial distress. Are you saying that you are > about to create a new "commercial" branch of memaid to generate you some > income? This is not certain, but... yes, I have such plans... > If so, where's the problem? Problems is... well, for example in free software world some people consider writting commercial software a crime. :) Sounds funny, but it's true. Richard Stallman is for many people a guru, and it's his opinion. The other problem is how I feel about it. I have always aimed that MemAid code would be totaly free and developed by many people, so many people as possible would benefit from it... (but now I see that lack of marketing and money doesn't help in promoting a product among people, and being "open" for a project doesn't give any warranty about development by others people.) And finally I don't know how some people will react here. But I would like to focus on commercial version for Windows, and I would like to still support and develop free/Open version. But it's still very vague - I don't know how everything will be. > I would definitely keep my fingers > crossed for you. Speaking for myself, I feel you have already given > much, so maybe it's time to think of yourself. Best of luck! Thank you..! -- David Calinski http://neodave.civ.pl |
|
From: Andreas G. <agr...@ya...> - 2003-11-30 18:13:07
|
> > <snip. In my restaurant example, it doesn't help much if I don't feed > > this vocabulary into a separate database using low grades to enforce > > frequent and focused repetition (aka drilling exercise) > > I am not sure if I get you here, but I definitely don't agree with last > sentence. You shouldn't have to "enforce" anything. > It's MemAid role to guess how difficult every item is and schedule > accordingly! I agree that we disagree. Our main difference is that you are talking about difficulty, and I am *also* talking about importance. During a (hopefully) normal and structured learning process, memaid helps a lot. No dissent here. But what, if I absolutely and unconditionally need some knowledge tomorrow? Ok, tomorrow might be a bit too exagerated. Let's say in a month's time I have an exam on some stuff. How to make sure, exams' contents will be covered by memaid. Again, the problems lies partially in difficulty. but mainly in importance. I would happily forget some other items (scheduled for repetition) if I would recall the items at hand for the exam. > Especially new code I am working on - among other things it will have a smart > "Finall Drill" feature that will work automagically, and there will be no > normal "Finall Drill" - everything should be automated and it is MemAid 'who' > should worry about enforcing and proper scheduling repetitions, not human. Agree here. But still then: Man overrides Machine - Assimov's 2nd law of robots? - not sure, will have to learn again. Sigh. ;-) > Now, I understand one may be interested in having a higher retention rate for > particular items (e.g. specific category) - and only "retention rate" should > be in MemAid's options to set (not difficulty of specific items - MemAid > should handle well elements with very various difficulty). I am not sure I understand this correctly. Is "higher retention rate" synonymus to "prefered item to learn, overriding other items in priority, normally scheduled for today" ? Sorry to hear about your financial distress. Are you saying that you are about to create a new "commercial" branch of memaid to generate you some income? If so, where's the problem? I would definitely keep my fingers crossed for you. Speaking for myself, I feel you have already given much, so maybe it's time to think of yourself. Best of luck! Andreas |
|
From: David C. <ne...@ci...> - 2003-11-30 17:49:21
|
On Sunday 30 of November 2003 18:08, Peter Bienstman wrote: > On Sunday 30 November 2003 17:26, David Calinski wrote: > > Especially new code I am working on - among other things it will have a > > smart "Finall Drill" feature that will work automagically, and there will > > be no normal "Finall Drill" - everything should be automated and it is > > MemAid 'who' should worry about enforcing and proper scheduling > > repetitions, not human. > > Sounds interesting! Will it have support for multiple revision of the same > item during the same session? That's one of the main reasons I programmed > 'drill-badly known': when I need to learn a new set of vocabulary, I want > to go through each item several times. Yes. It will try to keep executing these repetitions (of "forgotten" items) until ANN will decide "next interval is not today", which will mean, more or less: until one will put a decent grade (but here won't be a "hard" threshold - ANN will decide what is the threshold for every item, and will correct itself as long as will have more feedback data). And repetitions will be handled in a smart way: there will be priorities for every item. Generally: closer the item is to the begin of linked-list, higher the priority (higher chances that it will appear as next item to repeat in a session) - as it was. It has a strong rationale behind it, as element in linked-list are sorted by... But there will be many tweaks. For example: an item one forgot during a session, can be scheduled on the same session (aka "auto-final drill") by ANN, but it will have a special priority. Priority of such "forgotten" item will depend on, among other things: when it was last executed (but on the same session, on the same day; in other words: how many other items was executed since last review of this item on a session). MemAid will take care that such forgotten items will be repeated in a session until one will remember them, but will also take care that no item will be repeated with a interspace less than few other items, as much as this is possible. (this is not possible when, e.g. there is only 1 or 2 items left in a session.) Sorry if my explanation is too vague. Anyway: items one forgot, instead of going to a "Final drill" queue, will be smartly interchanged with 'regural' items scheduled for repetition. And priority system will take care that there will be high chances that elements one forgot will be executed until one will remember them, but at the same time with a nice inter-items-interval which will help to remember them. P.S. I have also bad a news. Please don't be mad at me... I am in a ... well... not good financial situation and I am not sure if this, and many other things I am working on in a splitted since some time MemAid code/version, I will immediately commit to the free Open Sourced version... sigh. -- David Calinski http://neodave.civ.pl |
|
From: Peter B. <Pet...@UG...> - 2003-11-30 17:08:42
|
On Sunday 30 November 2003 17:26, David Calinski wrote: > Especially new code I am working on - among other things it will have a > smart "Finall Drill" feature that will work automagically, and there will > be no normal "Finall Drill" - everything should be automated and it is > MemAid 'who' should worry about enforcing and proper scheduling > repetitions, not human. Sounds interesting! Will it have support for multiple revision of the same item during the same session? That's one of the main reasons I programmed 'drill-badly known': when I need to learn a new set of vocabulary, I want to go through each item several times. Peter -- ------------------------------------------------ Peter Bienstman Ghent University, Dep. of Information Technology Sint-Pietersnieuwstraat 41, B-9000 Gent, Belgium tel: +32 9 264 34 45, fax: +32 9 264 35 93 WWW: http://photonics.intec.UGent.be email: Pet...@UG... ------------------------------------------------ |
|
From: David C. <ne...@ci...> - 2003-11-30 16:52:14
|
On Sunday 30 of November 2003 17:26, David Calinski wrote: > The main problem I see - here in KMemAid - that if most sessions are very > long (e.g. 200 items, or a person often doesn't turn of KmemAid at all), > then these 10 seconds... could be... well, a bit short. Ah. BTW. It's worth noticing here that ANN "learning rate" changes dynamically according to actual ANN "Error" variable, so actually 10 secs of training ANN on the end of session could be as productive as e.g. 30 secs of training ANN sprinkled all over session.. -- David Calinski http://neodave.civ.pl |
|
From: David C. <ne...@ci...> - 2003-11-30 16:38:53
|
On Sunday 30 of November 2003 10:23, Andreas Grau wrote: > I changed inasmuch as I begin to thing off memaid as a recalling-aid, > not so much a learning tool. memaid keeps knowledge active, there often > needs to be a learning process before - where memaid may or may not help > well. In my restaurant example, it doesn't help much if I don't feed > this vocabulary into a separate database using low grades to enforce > frequent and focused repetition (aka drilling exercise) I am not sure if I get you here, but I definitely don't agree with last sentence. You shouldn't have to "enforce" anything. It's MemAid role to guess how difficult every item is and schedule accordingly! Especially new code I am working on - among other things it will have a smart "Finall Drill" feature that will work automagically, and there will be no normal "Finall Drill" - everything should be automated and it is MemAid 'who' should worry about enforcing and proper scheduling repetitions, not human. (Normal "Finall Drill" sux, mainly because you alter, in a way, learning process manually. Also because all "badly known" items are not intermixed with other items, which ... is bad, trust me. New automatical "Final Drill" will have support in ANN itself, and it's ANN-MemAid who will decide which items go to the "Final Drill". These elements, from "auto-final drill", will be intermixed, in a smart way, with current scheduled repetitions. It will have many benefits...) I consider putting elements into different databases (or "categories"), and keeping separate ANN for each own a bad idea. Now, I understand one may be interested in having a higher retention rate for particular items (e.g. specific category) - and only "retention rate" should be in MemAid's options to set (not difficulty of specific items - MemAid should handle well elements with very various difficulty). Currently there is no "retention rate" option thus. MemAid just sticks to the default - most optimal, for every item, but this (aimed "retention rate") is one of the features I want to see implemented. > In the fox code, you use user input time to train the net for 2 seconds, It's actually only 1 second, IIRC. (Was changed from 2 secs since time ago.) > so it get's trained quite often and for a rather big total of seconds. > For kmemaid, this get's done only once on exit (or when saving), but for > a maximum of 10 secs. So a first sight it seems that kmemaid doesn't > spend as much time training the net. Would this be a favourable thing to > change within kmemaid? How important is training time? I would say it's very important. But the more important is to turn on training ANN at all - from time to time. Without training ANN doesn't learn about an user's memory and user's personal accustoms (e.g. one may use grades in a bit different way). And I say that neither 1 second per item, nor 10 second per session, is generally suffice to learn ANN about just added data *immediately*, even with most fastest CPUs currently available. But it's suffice to have peace of mind that ANN will "get your memory model", after all. The main problem I see - here in KMemAid - that if most sessions are very long (e.g. 200 items, or a person often doesn't turn of KmemAid at all), then these 10 seconds... could be... well, a bit short. I would personally use (like in first MemAid versions) background training ANN in a thread, but I resigned from it in Fox version, as Fox Toolkit doesn't provide any portable threads. -- David Calinski http://neodave.civ.pl |
|
From: Peter B. <Pet...@UG...> - 2003-11-30 11:09:02
|
On Sunday 30 November 2003 10:23, Andreas Grau wrote: > Thanks, Peter and David, for your responses. You both helped me well > improving my usage and view of memaid. > > I changed inasmuch as I begin to thing off memaid as a recalling-aid, > not so much a learning tool. memaid keeps knowledge active, there often > needs to be a learning process before - where memaid may or may not help > well. In my restaurant example, it doesn't help much if I don't feed > this vocabulary into a separate database using low grades to enforce > frequent and focused repetition (aka drilling exercise) True, that's why I implemented the 'drill badly-known' feature in PyQt memaid. When I need to drill e.g. the vocabulary of this week's lesson, I input it with a low grade, and then 'drill badly-known' will keep on presenting me the item until the grade is high enough, while the 'optimally scheduled' learning mode will make sure I don't forget the older items. Hope this helps, Peter ------------------------------------------------ Peter Bienstman Ghent University, Dep. of Information Technology Sint-Pietersnieuwstraat 41, B-9000 Gent, Belgium tel: +32 9 264 34 45, fax: +32 9 264 35 93 WWW: http://photonics.intec.UGent.be email: Pet...@UG... ------------------------------------------------ |
|
From: Andreas G. <agr...@ya...> - 2003-11-30 09:46:34
|
Thanks, Peter and David, for your responses. You both helped me well improving my usage and view of memaid. I changed inasmuch as I begin to thing off memaid as a recalling-aid, not so much a learning tool. memaid keeps knowledge active, there often needs to be a learning process before - where memaid may or may not help well. In my restaurant example, it doesn't help much if I don't feed this vocabulary into a separate database using low grades to enforce frequent and focused repetition (aka drilling exercise) > More or less. So more "NN Training Time" - longer you will wait (e.g. after > pressing "quit"), but *MemAid will be better prepared to schedule repetitions > for you. In the fox code, you use user input time to train the net for 2 seconds, so it get's trained quite often and for a rather big total of seconds. For kmemaid, this get's done only once on exit (or when saving), but for a maximum of 10 secs. So a first sight it seems that kmemaid doesn't spend as much time training the net. Would this be a favourable thing to change within kmemaid? How important is training time? > > Talking about categories, I, for one, would love to see them in memaid. > > Peter's PyQT_MemAid has them, but plain MemAid will probably also have them, > sooner or later - if somebody will have enough lack of categories and will > implement them. :) "enough lack of categories" - huh ? > I understand your need. I could argue a bit about economy of over-working on > (some) items to have higher retention rate - I personally prefer to stick on > normal schedules and have nice feeling that I learn the most-optimally: good > retention with relatively small amount of time spend of learning. I have read your articles and support your points. Once, I heard a saying the "to a child with a hammer, everything looks like a nail". I guess that's what happened here to me :-) Memaid is/might be optimal for consequent, disciplined learning, but fails for ad-hoc, time-pressed item-acquisition. I didn't realize in the beginning - I had a new hammer :-) Keep up the good work, and many thanks for your support, Andreas |
|
From: David C. <ne...@ci...> - 2003-11-29 23:10:22
|
On Saturday 29 of November 2003 00:02, Andreas Grau wrote: > I am using kmemaid for learning french vocs. As I don't understand what > the "NN Training Time" in Settings represents, I started browsing for an > answer. Unsuccessful, but that's a different story. I don't know either, but I would expect, as name suggests, it has something to do with time spend on training ANN. E.g. AFAIK KMemAid uses this time after pressing "exit" to train ANN. On repetitions *MemAid gives some feedback data to ANN (so ANN will better "know" how your memory works, and will be able to better schedule repetitions for you), and training ANN means "learning ANN itself about your memory". More or less. So more "NN Training Time" - longer you will wait (e.g. after pressing "quit"), but *MemAid will be better prepared to schedule repetitions for you. > What I have found instead is David's explanation on the ANN internals. > There is written, that Grade 4 is regarded as the optimum, while 5 > implies the user knows the subject too well. And grade 0 is used for > items never reviewed. > > Up to today, I sticked to the following scheme: > > Grade 0 : Never heard that word before. > Grade 1 : Would have understood it, even without knowing it > Grade 2 : Oh sure! I know it, but just couldn't remember > Grade 3 : After some thinking, I found it > Grade 4 : Immediate recall, but with errors (spelling, sex) > Grade 5 : Immediate and correct recall > > I guess with the ANN internal grade interpretation, the above list could > be improved. But how? Any ideas? Well, I consider the most important this explanation: - If you pick grade 4 - it means that last interval was OK, so by pressing 4 you (also) give ANN feedback: "last interval was OK!" and ANN will "remember" that last time given interval was OK. - If you pick grade less than 4 (you have problems with recollecting answer) - you tell ANN "last interval was too long! please give me shorted next time", so next time ANN would give a slightly shorted interval (given the same data like on last repetition, of course) - Grade 5 tells ANN that last interval was too short (you remember it very well, thus interval could be longer.), so MemAid will "learn" it and next time give slightly shorted interval, if given the same data like on last repetition... Well, I hope it's not too complicated. Now, I think making a scale should be easier: You should use grade "4" when you remember something well, without any problems. Grade 5 when you remember it *so* well, that you think interval (distance from last repetition) should be actually longer. And grades less then 4 when there was some problems with recollecting. > Talking about categories, I, for one, would love to see them in memaid. Peter's PyQT_MemAid has them, but plain MemAid will probably also have them, sooner or later - if somebody will have enough lack of categories and will implement them. :) I am, personally, not much interested in implementing categories (I simply don't need them), but I would be glad if somebody would send a patch with a good solution (implemented categories). ;) > Here's my thinking - and it's all about vocabulary, your mileage may > vary. > > While learning French, I need to improve the overall language > capability, so importing and learning any word-pairs will do just fine > and this is where I am currently. > > Next, I realize, my restaurant vocabulary is just too bad. So I would > like to have these words in my french learning environment, but wish to > work on these words specifically. Or, once learned, just wish to > maintain the recall-level. > > (...) > > A dedicated Restaurant database would do the job to enforce learning > this specific language part - but then how to integrate it later in the > general french database to keep the recall on the language level? > > Any ideas? Comments? Am I too complicated? I understand your need. I could argue a bit about economy of over-working on (some) items to have higher retention rate - I personally prefer to stick on normal schedules and have nice feeling that I learn the most-optimally: good retention with relatively small amount of time spend of learning. But that's a bit different story - I am aware that sometimes it's nice to have separate category and learn some more from it. I just cannot tell you when categories will be implemented in MemAid or KMemAid... sorry. -- David Calinski http://neodave.civ.pl |
|
From: Peter B. <Pet...@UG...> - 2003-11-29 07:02:25
|
On Saturday 29 November 2003 00:02, Andreas Grau wrote: > I am using kmemaid for learning french vocs. As I don't understand what > the "NN Training Time" in Settings represents, I started browsing for an > answer. Unsuccessful, but that's a different story. > > What I have found instead is David's explanation on the ANN internals. > There is written, that Grade 4 is regarded as the optimum, while 5 > implies the user knows the subject too well. And grade 0 is used for > items never reviewed. > > Up to today, I sticked to the following scheme: > > Grade 0 : Never heard that word before. > Grade 1 : Would have understood it, even without knowing it > Grade 2 : Oh sure! I know it, but just couldn't remember > Grade 3 : After some thinking, I found it > Grade 4 : Immediate recall, but with errors (spelling, sex) > Grade 5 : Immediate and correct recall > > I guess with the ANN internal grade interpretation, the above list could > be improved. But how? Any ideas? You should use grade 4 as the optimum (correct recall) and grade 5 as 'it was too easy, don't bother whith scheduling this item for a long time' > Talking about categories, I, for one, would love to see them in memaid. pyqt_memaid has categories ;-) Cheers, Peter -- ------------------------------------------------ Peter Bienstman Ghent University, Dep. of Information Technology Sint-Pietersnieuwstraat 41, B-9000 Gent, Belgium tel: +32 9 264 34 45, fax: +32 9 264 35 93 WWW: http://photonics.intec.UGent.be email: Pet...@UG... ------------------------------------------------ |
|
From: Andreas G. <and...@cl...> - 2003-11-28 23:01:58
|
I am using kmemaid for learning french vocs. As I don't understand what the "NN Training Time" in Settings represents, I started browsing for an answer. Unsuccessful, but that's a different story. What I have found instead is David's explanation on the ANN internals. There is written, that Grade 4 is regarded as the optimum, while 5 implies the user knows the subject too well. And grade 0 is used for items never reviewed. Up to today, I sticked to the following scheme: Grade 0 : Never heard that word before. Grade 1 : Would have understood it, even without knowing it Grade 2 : Oh sure! I know it, but just couldn't remember Grade 3 : After some thinking, I found it Grade 4 : Immediate recall, but with errors (spelling, sex) Grade 5 : Immediate and correct recall I guess with the ANN internal grade interpretation, the above list could be improved. But how? Any ideas? Talking about categories, I, for one, would love to see them in memaid. Here's my thinking - and it's all about vocabulary, your mileage may vary. While learning French, I need to improve the overall language capability, so importing and learning any word-pairs will do just fine and this is where I am currently. Next, I realize, my restaurant vocabulary is just too bad. So I would like to have these words in my french learning environment, but wish to work on these words specifically. Or, once learned, just wish to maintain the recall-level. >From this example, I would even propose yet another structure: I have category "Languages" with sub-category "French". Within "French", I have contexts like "Restaurant", "Internet", "Aviation", "Dating" ... A dedicated Restaurant database would do the job to enforce learning this specific language part - but then how to integrate it later in the general french database to keep the recall on the language level? Any ideas? Comments? Am I too complicated? Best, Andreas -- Andreas Grau 24, rue Lakanal -- F-31000 Toulouse -- France Tel. +33-561-23.96.63 |
