[17:43] hey [17:47] random linux question ... I need to copy a file (or directory) from location a to location b however i need to be user A at a and B at B ... any ideas? [17:51] --> philip_ has joined this channel (i=8064db45@gateway/web/freenode/session). [17:52] Hi all! [17:52] pgornicz, maybe something like scp file username2@localhost:dir [17:52] hi [17:52] hmm [17:53] I don't think Yulia will be joining us, judging by her email, so once Chani and Alex arrive, we can probably get going. [17:54] oh hi [17:54] email? [17:54] oh, that [17:55] The email she sent in the early hours of this morning said that she wouldn't be free until 7pm. [17:55] that clever scp trick is round about but it works :D thx [17:55] ok.. .so... [17:55] clever trick? it's a fairly normal thing to do :) [17:56] then again, I use scp every day, maybe I'm not normal ;) [17:57] philip_: one nitpick on the marking scheme: maybe instewad of going from "occasionally wins" straight to "consistently wins" it should be "usually wins"? [17:57] ya, but i'd never have thought about doing it to the same computer [17:58] pgornicz: ohhh. same computer. yes, that is a neat trick then :) [17:58] Chani: i was thinking i might have to use pipes ... but that would have been a pain for a directory ... anyways, way off topic [17:59] Chani: I was trying to think of a way to do that. My concern is that it's going to be hard to tell if it's actually about the same and just within normal variance or not, without running a lot of games. [17:59] lol [18:00] philip_: that was in response to this question i posted before you showed up "random linux question ... I need to copy a file (or directory) from location a to location b however i need to be user A at a and B at B ... any ideas?" [18:00] pgornicz: I probably would've used chown and logged in as the other user to move it [18:01] But if you want to change it to do something like "wins x out of n games" as the categories, and then hope that it doesn't just get an unlucky run of games, you could try that. [18:01] philip_: well, would winning 4 of 5 games count as "consistently" or "occasionally"? [18:01] hmm. perhaps [18:01] Patrik: I'd guessed it was something like that. [18:03] Chani: I think if you win a sufficiently large percentage of games, we'll view that as consistent. Chances are this won't be an issue anyway, as it'll probably either win all by a small margin or win all by a large margin. :-) [18:04] ok :) [18:04] philip_, for the individual grades, for this: "2 other items combined to 30% (either 15 each or 10/20 split)" do we have to specify what the items are now, or can we do it towards the end of the term? [18:04] Perhaps we could use the margin of victory as a way of deciding marks, too. [18:05] Ioana: You need to specify them now. [18:05] using margins of victory feels a tad to random for me ... [18:06] <-- philip_ has left this server (zelazny.freenode.net irc.freenode.net). [18:06] <-- alexnico has left this server (zelazny.freenode.net irc.freenode.net). [18:06] --> philip_ has joined this channel (i=8064db45@gateway/web/freenode/x-vhrhbqfmprxdzfsm). [18:06] --> alexnico has joined this channel (n=a@207.81.123.181). [18:06] <-- philip_ has left this server (zelazny.freenode.net irc.freenode.net). [18:06] <-- alexnico has left this server (zelazny.freenode.net irc.freenode.net). [18:06] what just happened? [18:06] netsplit [18:07] 15:05 < philip_> Ioana: You need to specify them now. [18:07] 15:05 -!- Netsplit kubrick.freenode.net <-> irc.freenode.net quits: alexnico, philip_ [18:07] --> philip_ has joined this channel (i=8064db45@gateway/web/freenode/x-vhrhbqfmprxdzfsm). [18:07] --> alexnico has joined this channel (n=a@207.81.123.181). [18:07] 15:06 < pgornicz> using margins of victory feels a tad to random for me ... [18:07] aand here they are again [18:07] let's give the network a minute to settle [18:08] Alex: Participation - good point. [18:08] To some extent this is covered by peer reviews. But doing minutes/blog entries could be treated separately. [18:10] yeah, i propose 10% for participation, stealing from 'performance' [18:10] For margins, I'm thinking of "close" versus "huge". The boundary would have to be set, of course. [18:10] philip_: alexnico: you probably missed this line: 15:06 < pgornicz> using margins of victory feels a tad to random for me ... [18:10] In the summer, the best team was often winning by 10 goals, whereas the 2nd best would win by a couple. [18:11] huh [18:11] Of course, "participation" is a bit of subjective thing, and I don't know if Greg will go for it or not. [18:11] best team was dyanamite code though ... lol [18:12] Patrik: I think we can deduct about 20% from your grade for that bad joke! :-) [18:12] not unless you make 20% peer review ... lol [18:13] philip_ : But Greg also mentions that there are writing requirements (blogs, comments, etc.) as part of this course [18:13] interesting point, how do those come in to affecting the mark? [18:14] hmm, yeah [18:15] Well, as I said, it's up to you to specify this. If you just say "10% participation", then it's a bit vague, and I/Steve are probably going to be left with justifying why you all get 10%. :-) [18:15] I think you'll have to say "5% for minutes blog entries, 5% for meetings" or something. [18:16] Although it's probably not going to change much, at least there's a justification for the mark. [18:16] philip_ : that sounds reasonable [18:17] OK, so that's one issue resolved? :-) [18:17] but, where we steal 10% from? [18:17] hmm. participation: coming to meetings, not being silent at the meetings, doing minutes once in a while, talking to people instead of just dumping code (that one's pretty subjective, and we seem to have all done fine at it)... [18:17] The other thing that I see needing some work is the definition of individual tasks and deliverables. [18:17] Alex: I thought that you were going to pinch it from performance? [18:18] yeah. [18:19] philip_ : if everyone agrees, then I am okay with it [18:19] sounds far to me [18:19] So, at present we have 10% participation, 10% peer reviews, 10% performance (roughly sketched out at present), 10% final report, 10% design docs/commented code. [18:19] it's ok with me [18:19] That's 50%. [18:21] Then there's 10% demo, 10% for the code sprint work items and 30% for the remaining work items. [18:21] now we need some way to evaluating whether the tasks we set out to do have been completed [18:22] You all need to (a) decide what the remaining 2 items are for each of you, (b) decide how the 30% gets split between them (could be different for each person), and (c) give some description of the deliverables, including test plans, etc. [18:24] are we going through with neural nets? [18:24] I don't know - Ioana, do you think this is still feasible? [18:25] maybe we could combine it with the passing improvements? [18:26] i can't see that making it easier to complete ... [18:27] if we do use them, we need a specific task, like alex and yulia are using DT for the goalie improvements, and we need to improve passes anyway [18:28] My feeling is that it will be difficult to deliver something useful in 5 weeks. But you're welcome to try! [18:30] though on the other hand i've only really helped with the human control stuff and now the advanced actions (which will be passing and intercepting) though i guess actions could help with the sight changes ... [18:31] how is the sight stuff coming along ioana, any thoughts? [18:31] At the moment people are working on head-turning and DTs. We need 2 or 3 more goals spread across the 5 of you. [18:32] OK, so we've already made a start on passing and intercepting, so it's 1 or 2 more things to put in. [18:32] I would call for shooting/passing improvements [18:33] ooopss, it's being done [18:33] ya, that multi-tick actions are being created to try making passing/dribbling/intercepting better [18:34] yeah. not making much headway though [18:34] So, Chani and Ioana have worked on head-turning, Patrik and Chani on "advanced actions", Yulia and Alex on DTs. Did I miss anything or anybody? [18:34] i'm working on adapting the dainamite algorithm, but they seem to be using a lot of hardcoded params that I guess they got by just trying lots of combinations and seeing what works best ... i feel weird just copying them, but i'm not sure what else to do [18:35] I fixed the spinning-head thing, but again, the new code for findingthe ball appears slightly *worse* than the original code [18:35] Chani seems to have three tasks done or on the go, so she's sorted. [18:35] philip_: no, I have not worked on head turning. I looked up the vision quality optimization and we decided it wasn't worth it [18:35] hmm, i plan to spend the rest of the day riping dvds and working on robocup [18:35] Ioana is at about 2.5, then? :-) [18:36] although I suspect that the out-of-date vision info might be part of the reason they have so much trouble keeping their eye on the ball [18:36] Patrik is working on his second, as are Alex and Yulia, so they definitely need a third task. [18:36] i doupt it out of date by one-ish ticks would do that much [18:37] i only worked on the neck turning thing and the debug gui [18:37] other than multi-tick actions, i think there are other ways to improve shooting/passing/interception, especially shooting [18:37] Ioana: I was taking into account what you said about working on adapting the dainamite algo. [18:38] pgornicz: from what can tell it's always out of date by 1-2 ticks. information that comes in that tick won't be used until the next tick [18:38] philip_, that's for the neck turning [18:38] Ioana: If you wanted to extend that into a full task, you could. And perhaps someone else could work with you. [18:38] Ioana: Ah. [18:40] So, I think that another person could easily work on the advanced actions/passing/etc improvements, as it seems a pretty wide, catch-all sort of task. [18:40] so, I've been doing the advanced action stuff, and picking at the debug ui when I need better output to figure out why the actions don't work [18:41] we've got passign, dribbling, and intercepting in there, and finding the ball when it's lost... [18:41] Ioana could do a neural net as a task if she wanted to, though this might be a bit heavy for one person. [18:41] and I've not been having much luck at either dribbling or finding hte ball [18:41] if she wants to do one I can help [18:42] ya it is rather catch-all ... anything done in role.py with respect to picking when to start preforming an action to how the action gets caried out in action.py kinda fall under it [18:42] Chani: But you should only help if it doesn't detract from the other tasks that you're working on. [18:42] hmm. [18:42] Chani: Or you can help, but then run the risk of not delivering on your own tasks. [18:43] the way I see it, we have two sides to this: the basic actions in actions.py (pass, dribble, intercept) and the logic for choosing which action to perform (role.py) [18:43] Yes, the basic actions probably need some work, and the DT work can pick actions. [18:44] i can work on improving basic actions [18:44] i can work on improving passing, and if we decide using a neural nets would be useful do that, otherwise continue with what chani and patrick have been doing [18:44] I thought that those actions should be simple enough to hardcode in and then have AI pick which ones to do, but it doesn't seem to be working out [18:44] And, of course, a neural network could be good for improving the actions themselves. [18:44] ya it doesn't seem to be as simple as might be expected [18:45] so... errr... we're all working on improving actions? o.0 [18:45] that seems like a LOT of things just got tossed in the air to be done ... [18:46] Could we split some of the basic actions up, so that, say, Patrik and Chani do one set, and Ioana and Alex do another, for example? [18:47] we could. [18:47] possible [18:48] And leave some for Yulia. :-) [18:48] --> yulia has joined this channel (n=yulia@CPE001ff3c58763-CM001ade84f614.cpe.net.cable.rogers.com). [18:48] Speaking of the Devil. :-) [18:48] 1) interception. 2) passing. 3) dribbling. 4) not fucking losing sight of the ball constantly. 5) choosing which action to do [18:48] did i miss something? [18:48] 5) could be shooting [18:48] Hi Yulia! [18:49] hi Philip [18:49] hi everyone [18:49] hi Yulia [18:49] alexnico: right now shooting is just "kick at the net", but yes, we could create a shooting action [18:49] (4) and (5) are already items being worked on. I suggest you split (1), (2) and (3) up between you. [18:49] although iirc, the big problem with shooting is the client nevre asks "is there someone blockinghte net?" [18:50] philip_: who's working on 5? [18:50] can someone send me the log? [18:50] please? [18:50] (5) is basically DTs, I'd say. [18:50] Extended beyond the goalie. [18:50] speaking of DTs [18:50] are we going to extend DTs beyond the goalie, then? [18:50] philip_ : I am not sure if we can expand DT beyong goalie [18:50] i'm going to take a look into (4) today and see if i can find the problem, them possibly move on if things work out [18:50] Alex, we can [18:50] let's make shooting #6 [18:51] when we get proper controls [18:51] which i think i have [18:51] although shooting and passing have a lot in common [18:51] true [18:51] I've worked on 3 and 4 so far but not gotten anywhere. [18:51] we need some kind of priority queue of actions for each player [18:51] Chani: Yes, make shooting a separate goal - no pun intended. :-) [18:52] ioana, alexnico, does one (or both) of you want to take #1? [18:52] yulia: oh? [18:52] did anyone hear me at all? can you copy the log please so that i know what you were talking about? [18:52] Passing is probably a harder version of shooting, as the target moves. [18:52] i am interested with #2 actually [18:52] * Chani was hoping someone else would send the log :) [18:52] i can take 1 [18:52] i'll send it [18:54] <-- alexnico has left this server (Excess Flood). [18:54] lol [18:54] --> alexnico has joined this channel (n=a@207.81.123.181). [18:54] lol [18:54] <-- alexnico has left this server (Excess Flood). [18:54] >.< [18:54] yulia: bah, I'll email you the log [18:54] --> alexnico has joined this channel (n=a@207.81.123.181). [18:54] Just sent the log. [18:54] alexnico: stop [18:54] --> DrEnigma has joined this channel (n=DrEnigma@CPE00032f39283b-CM001cea375aa6.cpe.net.cable.rogers.com). [18:54] ok [18:54] i got it [18:55] ok then [18:55] thank you Philip [18:55] * yulia reading [18:55] did i get disconnected for writing so many lines? [18:56] don't know [18:56] i got 6 lines total [18:56] from you [18:56] http://pastebin.com is great for that. [18:56] yes, (Excess Flood) [18:56] ooohhhh... no wonder [18:56] who's DrEnigma? [18:57] Just a passerby. [18:57] interested in robocup? [18:57] i thought we got a new teammember [18:57] or Steve in disguise [18:58] Do you guys have a website/development log up anywhere? [18:59] Not to speak of. [18:59] Philip, you said you had to leave by 7 or 7.30 [18:59] have we agreed on something? [18:59] I can stay a little longer. I'm in EJ Pratt and have to be in a building five minutes away by 7:30. [18:59] alright [18:59] well, i got some progress in making the human player controllable [18:59] So, I think we have almost agreed on something. :-) [19:00] * DrEnigma vacates. [19:00] <-- DrEnigma has left this server ("Leaving"). [19:00] with the mouse [19:00] And then he was gone. [19:00] well we decided on everything but how to split up the 30% individual mark, and we're discussing what to do with action tasks [19:00] Yulia: Sounds promising. [19:00] hmm. [19:01] it can run after the mouse [19:01] or rather where the mouse points [19:01] for that i had to do an action queue [19:01] and i think that's what we also might want to consider for a regular player [19:01] to be able to do composite actions [19:01] well, I'm guessing 3 and 4 won't be enough to give me 30%... really I expect them to either be working soon or never [19:02] like dash_to or chase_the_ball [19:02] Chani, i have the same problem [19:02] yulia: actions can create other actios and wait for those sub-actions to complete [19:02] that's what i do too [19:02] basically my actions insert themselves or others to the queue [19:02] or clear it [19:03] also I think Philip mentioned neural nets somewhere earlier [19:03] technically they are very similar to DTs [19:04] Neural nets came up again for Ioana. [19:04] does the queue ever have more then 2 things in it ... i can't really think of when an action cannot decide to preform the next action, instead of terminating and having whatever is next on the queue happen [19:04] the data parser that i have can very easily be modified to create input files for python neural nets [19:04] Patrik, I have cases when I have 3 or 4 items [19:05] for instance i'm dribbling the ball to a point and while i'm dribbling i'm setting a target for kicking [19:05] I would see neural nets being more useful, potentially, within actions, rather than in picking them. [19:05] Philip, they can really be used for both [19:05] I had an hour long discussion with one of the profs from machine learning about it [19:05] just use softmax on the outputs [19:06] anyway, i realized that i need another person next to me to help me doing the experiments [19:06] so if Ioana is interested, we could cooperate [19:06] yulia, sure [19:06] I can't really do this stuff remotely [19:06] sorry Alex [19:06] huh [19:06] Yulia: that's okay [19:06] Yulia: I'm sure they can, but given that you're doing DTs for picking actions, I don't think we want to find another way of doing that. [19:07] the split between actions isn't as clean as I thought [19:07] Chani, it depends on how you look at it [19:07] we can split actions into basic ones and composite ones [19:07] then it becomes a little clearer [19:07] Yulia/Ioana: So, do you want to try and do a DT and a neural net for your two tasks? [19:07] how would you define 'basic'? [19:08] Chani, kick in the direction, turn body, turn neck, look in the direction, save, etc [19:08] composite actions are ones that are .. well, composed of basic ones [19:08] like dash_to involves turning first and then dashing_to [19:08] dashing in the direction [19:09] or chase the ball = find ball, turn, run in that direction [19:09] so on [19:09] that's what we are doing with multi-tick actions ... just not using a queue, and trying to only expose the "composite" actions to role.py [19:10] i think it's more intuitive with queue. but general idea is the same [19:10] except in queue you can also make priorities for certain actions [19:10] and change them as conditions change [19:11] OK, let's not get sidetracked too much by having technical discussions here. We need to agree the plan to be submitted to Greg tomorrow. [19:11] yulia: ok... right now most of those basic actions are actually just functions in actions.py used by Actions. I've been tempted to make them into Actions of their own [19:12] hmm. [19:13] Repeating my question: Yulia and Ioana, are you going to do both a DT and a neural net? If so, can the other action-related improvements be split across the rest of the team? [19:14] sure [19:14] i'll do the minutes and create the sample marking scheme for todays meeting [19:14] philip_, sounds good [19:14] Patrik, thanks! [19:14] OK, suddenly we seem to have reached a consensus! :-) [19:14] * Chani should do the minutes for next meeting [19:14] I'm still not clear on what I'm doing. [19:14] Philip, we are not as hopeless as we thought we were [19:15] Patrik: That would be great. If you can get the sample scheme out tonight, that would be excellent! [19:15] Yeah me neither. Does it mean I stop working on DT? [19:15] Re: next meeting. Sunday at 5pm? Remember that the clocks change this weekend. [19:16] philip_: will do [19:16] Alex: Yes, I think so. [19:17] Alex: And don't worry, I still know that you were working on it! [19:17] philip_ : OK [19:17] To a large extent, I think that the marking scheme is somewhat an administrative exercise, and there will be some latitude in the final grading. [19:18] So, I think we're done until such time as Patrik produces the first draft of the scheme. Then we'll have to turn it around asap, in order to get it to Greg by end of day tomorrow. [19:19] Unless there is anything else that desperately needs discussing, I declare the meeting over! :-) [19:19] hehe, now run off you! [19:19] Bye all! [19:19] <-- philip_ has left this server ("Page closed"). [19:20] I have to run too [19:20] bye all [19:20] bye