Disqus - Latest Comments for gperciva

Re: How to recruit open-source contributors

gperciva — Thu, 12 Jul 2012 16:04:17 -0000

I'm very interested in this topic, but could you increase the font size of the graph? Or maybe make that picture a link to a higher-resolution "zoomed in" image? My old eyes can read the percentages, but the meanings of the colors printed on the right-hand side of the graph is illegible to me.

Re: SFU alumni lies (new contact info)

gperciva — Wed, 02 Nov 2011 04:21:52 -0000

Sorry, I'm not familiar with the alumni stuff; my only connection
was via the email account. Try contacting the alumni association directly?

I know that UVic alumni have a "UVic online community" social
media platform they started pushing a few years ago, precisely to address this kind of thing. I wouldn't be surprised if SFU alumni have the same kind of thing now.

Re: Ubuntu ssh agent admitted failure

gperciva — Thu, 13 Oct 2011 22:11:17 -0000

Yikes! I'm sorry to hear about your hard drive crash. :( I hope
you discover a USB key or tarball or anything that has some old
data on it so that you can recover some of your scores.

Re: Becoming social

gperciva — Mon, 29 Aug 2011 08:55:43 -0000

You have a point; it's been about a week since I logged into g+. Granted, that's partly a matter of switching computers and laptops and stuff, but it certainly hasn't grabbed me yet. I'm seeing the same kind of malaise amongst programmers who were really excited about g+ when it first came out (news.ycombinator.com).

That said, as long as google isn't noticeably worse than fbook in terms of UI, and has a better privacy / data export option, I'm not too fussed. (of course, I wouldn't be too fussed either way)

I don't really know what I was expecting. It hasn't lived up to those expectations, but I can't point at anything and say "aha, there's where they failed".

Re: Artifastring 1.4

gperciva — Fri, 08 Apr 2011 22:36:53 -0000

Actually, I had contact with Joshua Koo a few weeks ago. He just
graduated, and one of his final projects was a collaborative html5
music editor; the idea was to make it easy for multiple people to
work on a composition at the same time. Could be useful in school
music classes, or something like that.

Yes, of course Artifastring be used in a game. The game could either
focus on low-level control, or higher-level musicality (if used with Vivi).

But you know what they say: ideas are useless. Code speaks.

about "Free"
...

Yes, Valentin. I consider myself to be "too good" to use the word
"Free". I am devoting my life to sneakily promoting the word
"open", because that's something you do with doors and windows,
and you know about my portal fetish. Oh noes, you have uncovered
my evil secret plan.

Seriously. It's garbage like this that make me consider giving up
linux and switching to BSD. The word "Free" is massively
ambiguous and has little meaning outside of a tiny circle. If
he'd used the word "libre" or "liberty", I might be more
sympathetic. Or even "freedom software".

Actions speak louder than words. I'm making the source code
available to people, with very few restrictions (i.e. those in the
GPLv3+), at no charge. If my use of the word "open" instead of
"free" to describe this process is offensive, then you're welcome
to release your own code and make use of whatever adjective you
like.

Re: Artifastring 1.0

gperciva — Sun, 21 Nov 2010 18:32:33 -0000

That python script is not available yet. Artifastring is a
program for simulating a physical violin; vivi is a program for
simulating a human violinist, and is nowhere near ready for
general use yet.

I'm releasing the violin simulation right now because it can be
used for a whole bunch of other stuff. Anybody working on new musical interfaces (such as
phone-controlled music, or the 3rd year engineering team project
at my university to create a virtual hurdy gurdy) can benefit from
a physical violin simulation. I wish that somebody else -- an
actual expert in this field -- had released their code; I've spent
something like 300-400 hours working on this. Sure, I've learned
a lot about physical modeling and signal processing in C (as
opposed to using marsyas or
matlab) along the way... but none of that is related to my phd
research, and in the UK you only get 4 years to finish. This was
a costly distraction. :(

But the only other publicly-available bowed violin synthesis (that
I could find, at least) was the venerable stk, which
uses the 1986 Smith and 1983 McIntyre, Schumacher, Woodhouse
algorithm. Although the modal synthesis used in Artifastring is
relatively inaccurate, it still sounds better than the mid-80s
stuff. :)

Re: Artifastring 1.0

gperciva — Sat, 20 Nov 2010 22:39:23 -0000

Oh, you were thinking of bowed col legno. I was thinking of the
hitting type. For an initial approximation of bowed col legno, it
would probably be sufificient to bump the friction coefficients
way down.

With respect to bowing, .action files specify the bow-bridge
distance, force, and velocity. That's enough to reproduce a huge
range of violin sounds. There's no "predefined
dynamics/velocities/whatever". There's just the position (as a
ratio of string length), force (in Newtons), and velocity (in
meters per second).

About the downbow vs. upbow, you missed the "infinitely skilled
violinist" part. Somebody with perfect control over the bow could
exert whatever force they wanted on the string, regardless of
where in the bow they were.

Now, you could certainly object that since there's no such thing
as an infinitely skilled musician -- but I already wrote that "in
the hands of a real-life violinist, the contact point's distance
from the frog changes the amount of force the bow exerts on the
string."

As for your comments about the bowmaker, you completely
misunderstand this work. To quote the main page, "this model is
old... does not represent the most accurate simulation known to
researchers... this model was not chosen for accuracy; rather for
being simple yet still 'good enough'... don't criticize it for
being 'not sufficiently accurate' for whatever you want to do".

I am not an acoustics researcher. I have read a few papers on the
subject, and I find it interesting, but it is not my field. I am
also interested in algorithmic graph theory, computer security,
and social philosophy -- but none of those are my field either.
I do not have scientific data to support any claims that this
model is accurate; in fact, I am explicitly saying that it is
*not* accurate. Virtually every single aspect of this model is
inaccurate -- for any part of this simulation, I could probably
find at least 3 papers which provided mathematics which produced
more realistic sounds, with solid empirical evidence.

If somebody offered my a GPL or BSD-licensed program which was
more accurate than this, I might switch. However, in the absense
of such a program, I have written (and will use) this one. To be
nice to other people, I am offering this code, licensed under the
GPL v3. But as the license states, this is offered "WITHOUT ANY
WARRANTY". If you don't like it, then don't use it!

Re: Artifastring 1.0

gperciva — Sat, 20 Nov 2010 21:27:15 -0000

The code currently doesn't allow you to alter the force of the
finger -- at the moment, it's modelled as an infinitely stiff
spring. If you changed that (line 206 in violin_string.cpp), then
you might get harmonics. I'm not certain if the finger width is
important, though. In this model, the finger (and bow) are
infinitely small.

I also want to emphasize that this particular model is not the
most accurate one known to mankind -- I'm sure that physicists and
engineers who specialize in bowed strings have physical
simulations that produce harmonics.

col legno is produced by hitting the string with the stick of the
bow. As such, it's similar to a hammer. If this was a digital
waveguide model, it would be trivial to simulate that action.
With modal synthesis, though, you'd need to decompose that force
into the forces upon each mode. You'd also need to deal with
cross-coupling -- instead of having two discrete forces acting on
the string, you'd have one force (the finger) and a continuous
function-force (the shape of the bow). The math gets way
trickier... in fact, this might be the reason that many (most?)
physicists don't use modal synthesis.

.action files are generated with lilypond and python. Write music
as normal. Run lilypond vivi.ly myfile.ly. Run python vivi.py
myfile.txt. Run action2wav myfile.actions. Yes, I *am* just that
awesome.

The idea of playing violin on an iphone is hardly new; see Ge
Wang's Ocarina and related line of instruments. The idea of
playing virtual instruments on ipod touches in an elementary
school setting is newer, and was what I did in Singapore. I
haven't updated my Research page with the latest publications from
that, but I'll get around to it eventually.

The difference between upbow and downbow (in the real world)
depends on the player's ability to control the bow. In the hands
of an infinitely skilled violinist, there would be no difference
whatsoever. In the hands of a real-life violinist, the contact
point's distance from the frog changes the amount of force the bow
exerts on the string. So within the .actions file, the difference
is in the force column. The precise calculation of those forces
is left as an exercise for the reader.

Re: Firelily

gperciva — Sat, 20 Nov 2010 20:56:10 -0000

As I wrote, "this is old code... it does not represent my current
ability level". This may surprise you, but my music department
did not offer any courses on comparitive programming languages. I
was aware that there was an "easy" language called "perl" -- quite
possibly just from having heard that slashdot was coded in it --
so I went with that.

Re: Tempo experiment over

gperciva — Thu, 07 Oct 2010 21:06:19 -0000

Sorry to disappoint, but I have no way of knowing if you had the
best scores or not. The experiment really was anonymous; I have
no idea which set of taps were done by you, or at what time(s) the
games were played, or even which countries the participants came
from.

I'll post an approximate breakdown of scores, so if you remembered
the types of percentages you got, you could see how many other
game-plays received similar scores... but that's as detailed as
you can get with respect to competition.

I personally wouldn't use those scores for anything significant,
though -- I've already said that the scores are known to be
inaccurate in certain circumstances. Now that I have the data I
need to finish the tempo detection algorithm, improving the
grading algorithm is the next step. We're still a few months away
from having a reasonable computer-graded rhythm score. :)

Re: Time spent on LilyPond

gperciva — Mon, 13 Sep 2010 13:53:19 -0000

./tt.py ls

will list the time spent on each project; that's all I use. I
reset the database every week, after recording any amount of time
over 10 hours. After the reset, I apply that extra time to the
next week's total, and then start counting again.

I suppose it might be interesting to keep a full record, and look
at "hours per day over the past 6 months" or something like
that... but I'm not quite _that_ seriously OCD about my time. :)

Cheers,
- Graham

Re: Tempo Game

gperciva — Sun, 08 Aug 2010 00:03:51 -0000

1. Most people don't want to jump to another level. Sure,
trained pianists can just blast through all these exercises, but
most amateur musicians find even the first couple of levels
challenging. That's why the "looks ok, take me to the next
exercise" option goes to another exercise, rather than changing
the level.

2. The visual metronome wasn't reliable for you!? Wow, that's
surprising. That metronome is just an animated gif. I figured
that everything would be able to render an animated gif correctly.
:(

There's absolutely no way to synchronize an animated gif and audio
in flash. I could try making a "flash movie", but that would
almost certainly use more CPU power than playing an mp3 or
displaying an animated gif, so I don't dare do that. If I knew
that only people running firefox 3.5 or higher on linux-x86 or
linux-64, then I could try things like that... but since I'm going
for wide usage, I can't make those kinds of assumptions.

3. Yeah, you have fun with browser-specific javascript features
and bugs. Yes, I considered using html5. Yes, I considered using
javascript. But almost all desktop users have a recent version of
flash installed so that they can view youtube videos, whereas I
have no idea what level of javascript is available in ie7 or
whatnot. When dealing with flash, there's only 1 (or 2 or 3)
non-free technolgoies to deal with: adobe flash, plus testing it
under windows or osx. If it runs under flash 10 on firefox on
windows, then I don't need to bother testing on ie8 on windows XP,
on ie8 on windows 7, on ie7 on XP, on opera on XP, on safari on
osx...

I'd be delighted if everybody supported html5 and a well-defined
level of javascript, and if they supported it *well*. But I'm
certain that they don't, and a few questions at my web-developer
friends suggest that I'm not being overly pessimistic in my low
estimation of cross-browser javascript support. Flash 10 really
seems like the most reliable solution. :(

Re: Bordeaux city

gperciva — Thu, 15 Jul 2010 11:20:54 -0000

1. Meh, ok, corrected. But a) in mutt I just type "Valentin" to
get your email, and b) do you expect me to think in +30 C ?!

2. foreign tourists? ok, maybe, since we were in the touristy
part of town, but they were speaking French. I suppose they could
have been Quebecois girls, which are also quite yummy. Or maybe
from Belgium, Switzerland, or...?

3. well, the pizzeria we went to called them "exotic". Also, note
that pineapple alone is not a Hawaiian pizza; you also need the ham.

4. wow.

5. I just ran 11.6 km in 53 minutes, which is close enough to 10
km/h. (ok, it's actually 13 km/h). The maximum speed of the tram
was about twice what I could run.

6. I have *never* claimed to be an expert on food. :)

7. Two words: "situational awareness".

8. That's certainly the cariacture, but it was interesting to see
it happen in real life.

9. Ok, "gallettes" it is. If only because then I don't have to
figure out how to type the accent. :)

10. Dude, the cafes we looked at were 10-20 euro. Is 6 less than
10? Yes. Now, I'll admit that 6 is not less than 4... but that
was just part of the situation. We should have spent an hour
walking *away* from the tourist center and towards the student
area (Victoire), instead of wandering around in a circle
in tourist-land.

11. Canadians never get over themselves. We just do it quietly.

Re: Academic culture shock

gperciva — Mon, 01 Mar 2010 07:31:54 -0000

Umm, because this is a course offered by the department of
electrical engineering, to electrical engineering students? And
some students in other branches of engineering, although I'm not
too clear on which ones.

It's definitely a course about computer *programming*, not
computer *science*. If it were supposed to be CS, then I'd
definitely be pushing for a switch to python. No, the main idea
is to prepare them for writing code for microprocessors to control
flashing lights and whatnot.

Re: Academic culture shock

gperciva — Mon, 01 Mar 2010 04:13:24 -0000

Yeah, but the mandate is "to cover the C language". You can't do
(serious) stuff in C if you haven't covered memory management.
The exam also covers reading+writing files.

OTOH, the kids aren't keeping up with the labs as it is. The way
things are going, half of them will have only covered 25% of the
course before the final exam.

The department needs to seriously take a look at their 1st and 2nd
year courses and figure out how everything is supposed to fit
together. It's currently a complete mess -- that's not just my
opinion; the hard fact of a 40% pass rate should be a huge warning
sign. Sadly, the department doesn't seem to care at all about
teaching, so I can't see this kind of serious review happening. I
feel sorry for these kids, but I've already overstepped my
authority (and vastly overworked myself... I think I'm
currently sitting at 60 hours of unpaid work, and I anticipate
another 10 before the end of term).

Re: Academic culture shock

gperciva — Sat, 13 Feb 2010 10:26:09 -0000

Oh, I totally agree that C is a *terrible* language to teach as their first language. A child's first language should either be python or lisp.

However, these _are_ people training to be engineers, and they need to create tiny executables so that their programs fit onto microcontrollers. No language with garbage collection is going to qualify. Also, I've heard that they use C in 2nd year to make lights flash on logic boards and whatnot. So "being stupidly close to the hardware" is actually somewhat of a plus in this case. :|

I've toyed with the idea of doing python for 8 weeks to introduce the major programming concepts (other than memory... although we could fake that by allocating space inside an array, maybe?), then spending the last 2 weeks learning C syntax and how it differs from python. That would be a much more invasive change, though, and I definitely don't have the clout to do that -- even if I were convinced it would be better, which I'm not.
(if I get seriously into the teaching side of things here, I might look into what they need to use programming for in the department. If they only need for() loops and binary operations in C in 2nd year, then it probably _would_ be worth doing the python / C switch. If they need much more, then it's probably best to stay in C.)

Re: SFU alumni lies (new contact info)

gperciva — Fri, 06 Nov 2009 19:14:42 -0000

Wow, I'm amazed that works. It's probably not reliable, though --
if I were google, I'd consider that a bug and change it in the
next version.

Still, as long as it works, it works!

Re: We have heat again!

gperciva — Tue, 03 Nov 2009 19:14:12 -0000

Well, $500 doesn't include accommodation. (which, I admit, is a
somewhat odd definition of "living expenses")

I never claimed that it was beyond my means... just that, to a
Canadian graduate student, $500/month on food and drink sounds
like a lot.

Re: They're so cute!

gperciva — Sat, 17 Oct 2009 08:11:29 -0000

Oops, fixed. I've been working in texinfo so much these days, I
forgot that HTML was totally whitespace-insensitive... texinfo is
insensitive to spaces, tabs, and one newline, but if you have two
newlines, it interprets it as a paragraph separation.

Re: Looking back at 100 blog posts

gperciva — Thu, 15 Oct 2009 19:11:38 -0000

I like this idea! I'm not so interested in tarsnap workflow (since it's still in the "public beta" phase, new website, etc). But I'm quite interested in the FreeBSD security workflow.

How much time do you spend on public/private email, how much time do you spend managing the security team (if at all -- and by "managing", I include both mentoring new members (if you do this at all) and assigning/discussing tasks)... and how much time do you spend actually looking at code?

I find that in LilyPond, actually working on the docs or build system is almost a relief; about 70% of my time is spent on emails and management. I'm wondering if that's something specific to documentation / website / releases (since the first two are much more "front-line" issues that are highly visible to users), or a general curse facing high-ranked members of open-source projects.

Re: Looking back at 100 blog posts

gperciva — Thu, 15 Oct 2009 12:57:00 -0000

Cats. You should write about cats. All blogs are better with cat stories. Plus photos.

Re: I fail at shopping

gperciva — Sun, 27 Sep 2009 16:59:25 -0000

I didn't think to type "mall" into google maps. I've been trying
all sorts of other things, though. And how much walking do you
think I do? The closest of those places looks about 45 minutes
away!

I suppose I could *gasp* take a bus or subway or something.

One of my flatmates suggested that I could probably find slippers
in some 1-pound stores; apparently there's a couple in the city
center. So I guess I'll head there next weekend. Or maybe on a
weekday, just to make sure that places will be open. I've got my
building keycard, so I can go to university on Sat.

- Graham

Re: Glasgow, United Kingdom

gperciva — Sat, 18 Jul 2009 23:26:52 -0000

Hey, I'm just being a good Scotman. Scotland is a *distinct
society* and constitutes a nation within the UK!

Re: Recovered from trip

gperciva — Thu, 18 Jun 2009 14:54:24 -0000

Heh, it goes to show how much of a Vancouverite I am: when I heard
of that study, I just nodded and said "of course". :) Such
studies come out fairly often, and Vancouver is always in the top
5. I didn't see any mention of Glasgow, though.

Glad to hear that you two made it back safely!

Re: Recovered from trip

gperciva — Wed, 10 Jun 2009 15:19:13 -0000

As opposed to assuming that I'm developing dementia? I could well
imagine that the book stated that going barefoot while outside was
a *symptom* of dementia, rather than reducing it... :)