Yeah, lets hear more about project Gallium. I'll do some alpha/beta testing for sure.
Does it have configs? Can you reuse any from G-Force?
We must know more!
Idea to take GForce to an entirely new level...
Moderators: BTT, andy55, b.dwall, juxtiphi
Not at all -- your effort and time thus far is very refreshing and interesting. I just wanted to address mobile1's implied accusation that we at SoundSpectrum are making a poor decision by not falling head over heels for genetic algorithms.chkman wrote:Andy
So what are you saying? Don't bother?
andy
Thank you, Andy. I've enjoyed my G-Force efforts so far.
I don't see how the GA is going to improve G-Force unless we are producing new configs. But it will be interesting to do the GA and to see what people like. I'm anxious to see what configs are new in 3.1
Now what were you saying about Gallium?
I don't see how the GA is going to improve G-Force unless we are producing new configs. But it will be interesting to do the GA and to see what people like. I'm anxious to see what configs are new in 3.1
Now what were you saying about Gallium?
Andy
Thanks for your comment - my comments earlier regarding priorities were not meant to be offensive - just my opinion. I think having worked with all kinds of optimization algorithms for years I am a bit biased thinking that this would be a major improvement.
I remember you opinion on the GA - or different teastes of people. Like I said back then I don't agree with you that everyone has a different taste.
There is a reason why in most pizza restaurants in this world, 80% of their menu looks the same - there are several types of pizza (peperoni, mushroom, margarita and a couple others) that people like and mostly buy. It is not that everyone prefers a custom pizza - most people fall into probably max 10 groups. And there are probably a couple people that like it custom that don't fall in any group but they are a minority (my guess).
Or from a math/optimization point of view, we have local optimums which MOST people prefer - GAs are excellent specifically in that regard, and my guess is that it's the same with visualizations.
This would mean that evolution would work for everone except the custom - minority.
You can see the same principles in art (why do people like certain artists style better), music (why do 80% or more people listen to mainstream music - diferent categories) - it all has to do with taste groups.
I guess the only way to see who is right is to do a simple proof of concept with this test server. I hope Chkman is still on to give this a try. If he does and we have some ratings some simple data mining of the rating data will display these "taste-groups" if I am right.
If there are no groups to be found (statistically) then you Andy would be right.
If you are right Andy I'll send you a nice bottle of wine - If I am right, I'll send a bottle to Chkman.
-------------
chkman - When I mention a config - I mean a combination of a color map, wave, flow field, (and particle). A GA would constantly produce the best combinations of those.
Evolving new flow fields, waves, color maps would be possible as well (would be more genetic programming) but more complicated and I don't think would have as much as an impact as finding good combinations of flows, waves, color maps etc...
Thanks for your comment - my comments earlier regarding priorities were not meant to be offensive - just my opinion. I think having worked with all kinds of optimization algorithms for years I am a bit biased thinking that this would be a major improvement.
I remember you opinion on the GA - or different teastes of people. Like I said back then I don't agree with you that everyone has a different taste.
There is a reason why in most pizza restaurants in this world, 80% of their menu looks the same - there are several types of pizza (peperoni, mushroom, margarita and a couple others) that people like and mostly buy. It is not that everyone prefers a custom pizza - most people fall into probably max 10 groups. And there are probably a couple people that like it custom that don't fall in any group but they are a minority (my guess).
Or from a math/optimization point of view, we have local optimums which MOST people prefer - GAs are excellent specifically in that regard, and my guess is that it's the same with visualizations.
This would mean that evolution would work for everone except the custom - minority.
You can see the same principles in art (why do people like certain artists style better), music (why do 80% or more people listen to mainstream music - diferent categories) - it all has to do with taste groups.
I guess the only way to see who is right is to do a simple proof of concept with this test server. I hope Chkman is still on to give this a try. If he does and we have some ratings some simple data mining of the rating data will display these "taste-groups" if I am right.
If there are no groups to be found (statistically) then you Andy would be right.
If you are right Andy I'll send you a nice bottle of wine - If I am right, I'll send a bottle to Chkman.
-------------
chkman - When I mention a config - I mean a combination of a color map, wave, flow field, (and particle). A GA would constantly produce the best combinations of those.
Evolving new flow fields, waves, color maps would be possible as well (would be more genetic programming) but more complicated and I don't think would have as much as an impact as finding good combinations of flows, waves, color maps etc...
Yeah, I said I'd do it. My friend Joe is supposed to provide the server and so far I've been unable to FTP to it. I am considering paying for a "web hosting" site but will avoid paying money if I can. I will get a domain name for it.I hope Chkman is still on to give this a try.
So I can't get to the server then I realized "dah" I have IIS on my machine and can start coding and testing on it. I need to learn more about ASP or decide if I want to use Perl.
Next question is "are there more configs supplied with Platinum than Gold?".
Andy, good luck on the winning the wine!
Using Platinum / Gold or free version. I guess this would depend on how many have Platinum or Gold. Obviously the more people we have to participate in this the better. So if there are 2000 Platinum user, 5000 Gold and 1 Mio who use the free version - I would suggest to do it using the free version. We can probably only get a percentage to give it a try anyway. If we have more people, we could run larger populations which would allow us to cover more local optimums and cover a wider variety of tastes. Also with having more people test this, they each would have to rate only a couple of visualizations every hour - instead of rating more - and it would still work (we simply hand out each config to multiple people).
Then again I don't know how much better Platinum looks. I'll buy it the next couple of days to find out - $30 is not a lot.
Even if the GA does not work (which would be very surprising to me), I think having a database that shows what people like - will be very interesting by itself. The results would help authors to learn what type of flows, waves, color maps people like. Another result would be that we could actually post a ranking of color maps, waves, flow fields... would be interesting what this would look like - and could help authors what people like. Or as mentioned before, running some data mining on the data to get some rules (what combinations look good and which don't - if you log the bpm (or other parameters) we could even break it down into different styles/speeds of music. Possibilities are endless. Or if someone rates many configs - we could actually derive a custom profile of what visualizations that one particular person likes....
Anyway the sky is the limit if we give this a try.
Then again I don't know how much better Platinum looks. I'll buy it the next couple of days to find out - $30 is not a lot.
Even if the GA does not work (which would be very surprising to me), I think having a database that shows what people like - will be very interesting by itself. The results would help authors to learn what type of flows, waves, color maps people like. Another result would be that we could actually post a ranking of color maps, waves, flow fields... would be interesting what this would look like - and could help authors what people like. Or as mentioned before, running some data mining on the data to get some rules (what combinations look good and which don't - if you log the bpm (or other parameters) we could even break it down into different styles/speeds of music. Possibilities are endless. Or if someone rates many configs - we could actually derive a custom profile of what visualizations that one particular person likes....
Anyway the sky is the limit if we give this a try.
Gallium and Nitrogen are indeed two separate projects. You're correct in saying that Gallium is a ways down the road still.JayPro wrote:#1. I was saying something about Gallium, namely that it's quite a ways down the road yet...unless the new developer he has on board now can pull a rabbit out of his hat.
#2. IIRC It's called Project Nitrogen now (read the latest News headlines).
Andy
-
JayPro
- Posts: 738
- Joined: Sat May 01, 2004 10:51 pm
- Location: Huntington Station, Long Island, New York
'Kay...now I'm kenfewwwwwwwwwwzed.
I thought the next great project was only one thing. Is it that you've got two things going and may the better proposal win?
You can e-mail me or whtever if you want to explain things further...coz the fact that ther're now two projects is news to me.
I thought the next great project was only one thing. Is it that you've got two things going and may the better proposal win?
You can e-mail me or whtever if you want to explain things further...coz the fact that ther're now two projects is news to me.
"God is syntax."
Wow the Platinum version is a major step forward compared to the free version - Thanks Andy, I had no idea!!!! - also the User Interface is very nice. Some of the stuff GForce comes up with is simply MINDBLOWING.
Chkman if you get this GA going - in the worst case we should have a large number of ratings - from which I can create simple rules (of what looks good and/or bad) - which Soundspectrum then can decide to use or not use (if there are rules).
I am wondering once the server is setup, and everything of the GA tested and running, whether Soundspectrum would send out an email to all platinum members, inviting them to participate in this test. Andy?
This would provide enough users to test the GA while at the same time providing data to Soundspectrum of what looks good and what doesnt. This could be valuable information to decide in which direction to develop (even if the GA does not work)
If the GA works, People have no idea what they are in for - this would be the beginning of a new generation of music visualization! We'll see....
Chkman if you get this GA going - in the worst case we should have a large number of ratings - from which I can create simple rules (of what looks good and/or bad) - which Soundspectrum then can decide to use or not use (if there are rules).
I am wondering once the server is setup, and everything of the GA tested and running, whether Soundspectrum would send out an email to all platinum members, inviting them to participate in this test. Andy?
This would provide enough users to test the GA while at the same time providing data to Soundspectrum of what looks good and what doesnt. This could be valuable information to decide in which direction to develop (even if the GA does not work)
If the GA works, People have no idea what they are in for - this would be the beginning of a new generation of music visualization! We'll see....
Designed the layout of the web site early this morning. Got sidetracked for the rest of the day. Just sitting down at the computer now and yawning. I'm going to try coding some of the guts of it, maybe the selection process. Have to setup the database 1st.
I know that your anxious, please be patient. I should be able to spend a good deal of tomorrow on it. Steeler game and finish up installing a car stereo is all I have planned.
I know that your anxious, please be patient. I should be able to spend a good deal of tomorrow on it. Steeler game and finish up installing a car stereo is all I have planned.
I would say 1 particle and 1 background image, 1 wave, 1 flow and 1 color per each entity. However for particle and image there should also be the option to have NO background image or particle - and it would probably make sense that the probability (when starting out) that there is no particle or background image is higher (maybe 80%)... because otherwise if we create random population for the first generation all of them have particles and background image which might be a bit too much... I would do that only 20% have - from there the GA will take over and decide...