Computer Science

Valentin Tihomirov

07:57

@vzn That is crap. Tell the teacher about "deductive (direct) vs. inductive (indirect) learning". Tell them the purely inductive style is waste of time. That is why we have education and learning from the past instead of science (creating alphabet and everything else yourself from the scratch). You should give mostly theory (deductive knowledge) with some embedding of practice where students are asked to predict your next step.

Valentin Tihomirov

08:14

Also, learn about "learning styles". There are people who are comfortable with direct (deductive, theoretical) learning, even future scientists. Moreover, most people are not going to be scientists. Engeneers and 99% of the rest population just learn to apply the science, using the "maps" that are already built by the scientists. That is why should not waste time to prepare "standard, creative scientists" only with your inductive style.

Raphael

12:14

@ValentinTihomirov I see a false dichotomy here. In my experiments, some situations call for one and some for the other approach. Deductive approaches often work better for transporting information (e.g. a foundation of definitions and such) but learning happens inductively through working with the material.

@Gilles Sorry, I misremembered.

Valentin Tihomirov

12:57

@Raphael I responded to somebody claiming that we must forget the deductive learning as inefficient. We should stop using maps and start building them ourselves in the schools because that constituates the true learning. That is what false is. We use deductive education just because it is more efficient. The science is building the maps (knowledge) indeed but it does not mean that we should discard the knowledge and start over because this is "true science".

Raphael

13:18

"We use deductive education just because it is more efficient." -- citation needed. Afaik, learning science has established that deductive "teaching" alone does not lead to much learning. All I'm saying is, be careful with your criticism. While inductive learning is definitely more resource hungry, it's probably more effective (if done right). Good teaching is probably a lot about striking the right balance here.

One concrete example: most mathematics students I know here (Kaiserslautern, Germany) learn nothing in lectures. Ask them after a session what was on today, they can't tell you. They only copy down what's on the blackboard, maybe take some notes. They learn a) by doing revision and exercises later, and b) cramming before the exam.

They definitely get through lots of material in little time, but what do they have to show for it? Fairly little.

A professor I had the pleasure of being taught by did the exact opposite: his courses had very little content, in the ToC-length metric. But what he had in there, he taught, and we learned. In my estimation, we had learned more after such a course than after "classic mathematics" course, and with less time investment on our part.

(Side note: classic, non-interactive lectures have been shown to not promote learning at all.)

Valentin Tihomirov

14:01

By effective I mean the amount of knowledge per unit of time. I heard it orally on the didactics lecture. I am not going to challenge this reasonable fact. It is obvious to everybody that science builds the compact models (the "maps") of the world. It is much more efficient to use a map rather than to invent your own. It is obvious to every normal person. Even children can easily understand this and I am not going to prove it with authoritive references. We are not at Wikipedia.

Raphael

@ValentinTihomirov Well, when we talk about teaching we should talk about learning. Effective communication of data is not the same as learning. I don't know in what kind of didactics lecture you where, and when, but if all they did was focus on cramming data into students, they were missing a big part of what is important.

I agree that we have to establish a common language first. But we can certainly learn it along the way. There is middle ground between classic lecture (pure data transmission) and "everbody learns on their own inductively", for instance guided inductive learning where the teachers observes and corrects if necessary.

(" It is obvious to every normal person. Even children can easily understand this and I am not going to prove it with authoritive references. We are not at Wikipedia." -- careful with the attitude.)

(Note that this is the chat of a computer science site; if anything, the demands regarding supporting your claims with proof and/or reference are higher than on Wikipedia.)

Valentin Tihomirov

14:20

I will keep telling what I find right over the stupid rules. I am telling the noise. I am telling the reasonable things that make sense and easy to check. You may ask but I am not forced to prove anything. Science is not trust in authority, after all. It is common sense that sane people can see.

Finally, there cannot be more strict policy wrt to referencing. I had experience in WP, trying to tie together several articles. My interwiki references are always rolled back because of "lack of references".

Raphael

@ValentinTihomirov You are, of course, entitled to your own opinion. Your claims that they are unchallenged truth are, equally of course, vacuous. (A person that claims to be a scientists but ignores scientific proof/data is what we call a "crank", by the way. Just so you know.)

Valentin Tihomirov

You basically say that I am illiteral kid because not aware of such simple stuff. You also cast a shadow of crank upon me, telling others that I am a crank. Of course this is what you should do after mine statement Science is not trust in authority, after all. It is common sense that sane people can see. It obviously says that I do not accept the scientific proof/data.

Telling what is the science, that science is common sense rather than preaching after authorigy means that you do not accept accept the scientific proof/data and you are a crank.

user340082710

16:09

What are some good practices to receive more views/attention towards a question you've posted. I feel like there are some questions that receive 100+ views in the matter of a few hours, but mine received 7 views in 12. Is my question perhaps better suited for another stackexchange site perhaps (was thinking maybe math)?

EvilJS

17:20

@user340082710 it is about "bipartite graph"?

@ValentinTihomirov to this and other messages: it is true, that different people have different learning styles, but top-down vs bottom-up approach is not that different among people. Read/write/execute/hear/touch/pictures and convergent vs divergent styles are really making difference. The school experience pushes everyone into convergent (destroying creativity).

@Raphael within one hour after lecture / exam, the memory recovery is unlikely, merging new knowledge with existing one is possible only after about 24h, and the most time 2-3 repeats are needed. The problem is the most visible with eye-witness...

user340082710

@EvilJS Well, the hypercube graph is indeed bipartite, but I am asking about the n-cube graph in particular.

EvilJS

Experience is a weight, but without it we start reinventing wheel, not that I have anything against, but this is strictly inefficient. And about proofs - let me discard some knowledge and try to find even prime number... Bigger than 2.

@user340082710 There are bounties or math stacks which might give bigger attention. I have seen the question, I do not know the answer, but standard way (afaik) is to give it a few days and then if nothing happened migrate.

user340082710

Do you think that the question lacks information or detail? Or perhaps a better title might attract more viewers?

vzn

17:43

@user340082710 its kind of abstract. can you give any more context/ bkg? it could be studied empirically without too much trouble. is an n-cube basically also denoted by the binary strings {0,1}^n? maybe reformulate it in terms of binary strings? as for hits on your question, it is not unusually low if you look at other questions.

EvilJS

@user340082710 Better title is always better. Any improvement, examples and so on helps.

vzn

the question seems to relate to hamming distance...

it looks likely to have transition-point-like behavior...

user340082710

I suspect there is some restrictions of the hamming distance of the pairs (s_i, t_i), but haven't found such a result/reference yet.

What do you mean by transition-point-like behavior?

vzn

my (blanket) suggestion for anyone who wants better attn on the site is to fill out more of their profile.

are you a student? researcher? what bkg/ context does the problem come from? etc

cant the paths be denoted by binary strings with a single bit "flipped" from each string to the next?

ie hamming distance change/ delta of 1.

user340082710

Researcher - I suspect (hope) that a solution to this problem will give me some insight into the generation of certain Gray code - and yes, labeling the vertices by their corresponding binary strings gives the interpretation you mention.

vzn

17:51

transition pt behavior: existence of solution more/ less likely depending on size of sets wrt size of cube etc

@user340082710 undergrad? grad?

user340082710

grad

vzn

masters student? phd? etc?

user340082710

starting my phd in may

vzn

why not formulate it in terms of gray codes?

user340082710

because i suspect that there is a more general result that is applicable

vzn

17:53

there are (somewhat) similar problems in "set systems" and "hypergraphs" etc...

user340082710

I am pretty familiar with the gray code literature, but not nearly as much with the literature regarding 'graph decompositions' (though i am not sure that's the correct term)

vzn

when you say vertex disjoint paths you mean no paths have any common vertices?

user340082710

yes

vzn

there are some studies of message routing on an n-cube, possibly somewhat related

user340082710

i have added that clarification to my question - also, is graph decomposition the name for the problem i am trying to describe?

vzn

17:57

do you have masters degree? there are many equivalent ways to formulate certain problems and sometimes there are different connections in literature (not a single focus) etc.

user340082710

i am finishing it this term

vzn

think there is something call "path coverings"...

Path cover

Given a directed graph G = (V, E), a path cover is a set of directed paths such that every vertex v ∈ V belongs to at least one path. Note that a path cover may include paths of length 0 (a single vertex). A path cover may also refer to a vertex-disjoint path cover, i.e., a set of paths such that every vertex v ∈ V belongs to exactly one path. == Properties == A theorem by Gallai and Milgram shows that the number of paths in a smallest path cover cannot be larger than the number of vertices in the largest independent set. In particular, for any graph G, there is a path cover P and an independent...

user340082710

looks like a good starting point

vzn

:)

user340082710

in my formulation, however, one specifies the starting and ending points of each path, as well as the number of such paths

so i need to see if that formulation has been studied already

vzn

18:04

yes there would be different ways to formulate this. one could either look at all the paths on the cube and say youre limiting them, or you could say that youre starting from 2 sets of start/end points and looking at possible paths.

user340082710

this appears quite similar: sciencedirect.com/science/article/pii/S0304397514001078

EvilJS

Another mechanism is when somebody goes to the exam, after exam if we ask him "please tell us the questions" or "please explain X", even though on the exam he did it, there is just blank, he cannot remember anything - upkeep of short term memory is lost, long term memory is not there yet, because it was kept until this very moment... This does not apply up to such extent to well learned and super calm person.

vzn

@user340082710 wow! nice find. nearly an answer maybe. plz post it if you think so

the "fault tolerance" angle shows up there. maybe that is a basic context the problem arises.

so anyway you see maybe your problem can be described as a message routing problem somehow (etc) to leverage that particular literature.

maybe describe your masters prj sometime if you have time, sounds interesting, have looked into codes some over the yrs. (which reminds me) ps plz +1 this

user340082710

18:29

Done

I am still looking at the reference, and seeing if there are other results that might be closer to what i've described

@vzn my thesis is based on the results of this paper: sciencedirect.com/science/article/pii/S1570866713001020 if you're interested

vzn

ok thx

what country are you in?

user340082710

Canada

vzn

did you just upvote that ad? thx man =D

user340082710

:D

It's a nice pun

vzn

coding theory seems to show up in deep complexity theory somehow, the links still not fully explored. once (few yrs ago) ran into a guy that worked on a P!=NP proof somewhat based on coding theory.

user340082710

18:36

interesting

EvilJS

@vzn Have you seen Raphaels response to change in the text, and +6 comment that asks about relation to computer chat? Maybe it is high-time to at least change the text, or mirror the image, so it would at least look like conversation?

user340082710

i am looking at the problem from the perspective "how many of the bits do i need to read to perform increment operations"

vzn

@EvilJS feel free to go right ahead... admittedly R. suggesting something acceptable was something of a breakthru (thx for your effort) but am not interested in attempting to plz ppl who clearly are not pleased by much of anything... & others seem to have a similar sentiment... hamster on a treadmill anyone? :(

@user340082710 yeah it sounds kind of abstract but presumably it has some application in hardware & also an apparently elegant theoretical problem...

EvilJS

I like the picture, I like the idea, but it honestly looks like he reads paper out loud, so rather monologue not chat... I know the effort counts, and it is easier to critique than draw a new one, but responses are at least constructive and with perfectly valid point. So something pleasing majority and attracting people should at least win the local race.

vzn

@EvilJS so fix it according to demands... and see what happens

user340082710

18:43

well gray codes originated from the conversion of analog and digital signals

vzn

@user340082710 yeah this is all reminding me of some question from a ways back but cant put my finger on it right now... iirc it had to do with optimal hardware at a bit level etc :|

EvilJS

@vzn well, even without artistic skills, I will take the challange.

vzn

@EvilJS will likely upvote whatever you post (thx for your own support). think you may be mistaken if you think this is just about artistic skills or acceptable words/ messages etc

EvilJS

@vzn and this is the part, where this becomes paegant contest. I know that people who somehow likes me or I have not managed to offend are likely to upvote, but in this case, I really would like to help and make it work because it is good...

vzn

@EvilJS go for it EJS/ good luck you havent offended anyone :)

EvilJS

18:53

@vzn totally not true ;) I counted at least five, which I did not intended... Anyway I started yesterday, I put Delanauy triangulation on top of two people playing chess, making chessboard invisible and wrote text about swinging by. Disaster...

Just the sketch - something like this as the outline (not the given one, just general scene).

Fun, witty, Computer Science related, chat like scene, visually appealing, memorable. Taking three at once is doable, all of them, gets harder.

user340082710

19:22

@vzn Think I found what I'm looking for: sciencedirect.com/science/article/pii/S0020019008002238

:)

EvilJS

Anyway now I am struggling with AVL bulk inserts and it seems that when data is scattered it is not beneficial :-/ I mean, when data in the batch is not localized the boost is marginal.

vzn

20:08

@user340082710 amazing! is that exactly the same as your problem? some very impressive googlefu!

user340082710

@vzn I believe it is exactly my problem

vzn

20:27

@EvilJS maybe ask R instead of me. looks fine. defn liked your idea of tieing up/ hooking up the Turing test somehow.

Raphael

22:21

@user340082710 For views, good title (piques interest) and proper tagging (so the question shows up in the correct lists) are most important. Upvotes and answers are another issue.

@EvilJS I seem to remember that delaying rebalancing operations are beneficial in concurrent settings, but I'd have to dig for the reference.

Transcript for