We Can Only Ever Know Epsilon
Had a good discussion last night. First off, I brought up the idea of holophonic sound. Click!
Make sure you wear headphones, preferably the studio monitor kind. It's really kinda creepy, and interesting how they do it. Basically they record it all with a dummy head, with replicated ears. So we made two interesting conclusions from it. One, it's obvious that the ear itself changes how we hear things. It's been tried to use a priori methods to replicate actually being in a place, by using mathematical/wave physics to change the volume and delay of the sound, but they are definitely not the optimum solution. As you can probably hear. But also, what we found interesting, was that in theory you should be able to create a signal processing filter that would change normally recorded sound/music into something that is quasi-3D. Cool.
Information theory was next on the plate. Yes, a digital signal does degrade people. There is this current belief that digital cables are great, because you receive exactly what was sent. It is binary, either the signal gets there or it doesn't. It's nice to think theoretically with computers, but the problem with that is that things take place in reality. So we have to deal with that. When a signal is being sent, there is inherent noise. It isn't like analogue noise, but bits will be changed purely because of entropy. This is why we have compression, redundancy and error checking. So, with the best codes, they are supposed to take into account for all of this. It's a probabilistic affair, you know, quantum mechanics; therefore there is always the exception cases. Areas in cryptography do deal with this problem, that being ensuring integrity. But this happens a lot less than you think, and all we can do is minimize it. Basically, I'm trying to justify pricey stereo equipment.
Yann also brought up a cool way of thinking about the least-squares method of estimation. If you have a bunch of linear equations, and it isn't deterministic (i.e. the vectors intersect), you want to find out how close do they get. This is where it gets cool. You can break it down geometrically and look at orthanormal basis's and think of it in terms of creating orthogonal vectors, which is the shortest distance between to other vectors. For me it was a revelation, because I've always seen the least-squares solution as an ugly affair, something strictly for *those* applied math types. But it does have a nice basis (no pun intended) in linear algebra/geometry, so I can respect it now.
Onto something less technical. I went out on the Champs Elysee the other night. Only two real points of interest. Well, I guess girls always love how it looks, but I'm concerned about more practical things. The first being the price of drinks, and food anywhere near the main road. 175 Euros for a dinner!?! 13 Euros for a drink! My god, and people line up to pay it. The second interesting note was how we got into this bar, "Doobies." There was no door handle, just a buzzer and a slit with a slidable metal cover. So, ring the buzzer...and wait. Then I hear a faint series of knocks on the other side of the door, so what the hell? Match it. Then another series, match that again, and the door opens. We're in, and I wonder at my luck for getting into bars in Europe.
I also had my first dream in French the other night. Very cool, but I'm disappointed it took so long. That being said, the ol' French has gotten a lot better in the past couple of weeks. I've managed to have almost all French conversations at lunch and with some coworkers lately. I feel relatively accomplished.
"Garfield becomes an oddly surrealist comic when you remove the dialogue"
In any case, I'll see all you Calgary folk soon enough. I'm glad to be getting back in cowtown and rippin it up. Party on Wayne...
Make sure you wear headphones, preferably the studio monitor kind. It's really kinda creepy, and interesting how they do it. Basically they record it all with a dummy head, with replicated ears. So we made two interesting conclusions from it. One, it's obvious that the ear itself changes how we hear things. It's been tried to use a priori methods to replicate actually being in a place, by using mathematical/wave physics to change the volume and delay of the sound, but they are definitely not the optimum solution. As you can probably hear. But also, what we found interesting, was that in theory you should be able to create a signal processing filter that would change normally recorded sound/music into something that is quasi-3D. Cool.
Information theory was next on the plate. Yes, a digital signal does degrade people. There is this current belief that digital cables are great, because you receive exactly what was sent. It is binary, either the signal gets there or it doesn't. It's nice to think theoretically with computers, but the problem with that is that things take place in reality. So we have to deal with that. When a signal is being sent, there is inherent noise. It isn't like analogue noise, but bits will be changed purely because of entropy. This is why we have compression, redundancy and error checking. So, with the best codes, they are supposed to take into account for all of this. It's a probabilistic affair, you know, quantum mechanics; therefore there is always the exception cases. Areas in cryptography do deal with this problem, that being ensuring integrity. But this happens a lot less than you think, and all we can do is minimize it. Basically, I'm trying to justify pricey stereo equipment.
Yann also brought up a cool way of thinking about the least-squares method of estimation. If you have a bunch of linear equations, and it isn't deterministic (i.e. the vectors intersect), you want to find out how close do they get. This is where it gets cool. You can break it down geometrically and look at orthanormal basis's and think of it in terms of creating orthogonal vectors, which is the shortest distance between to other vectors. For me it was a revelation, because I've always seen the least-squares solution as an ugly affair, something strictly for *those* applied math types. But it does have a nice basis (no pun intended) in linear algebra/geometry, so I can respect it now.
Onto something less technical. I went out on the Champs Elysee the other night. Only two real points of interest. Well, I guess girls always love how it looks, but I'm concerned about more practical things. The first being the price of drinks, and food anywhere near the main road. 175 Euros for a dinner!?! 13 Euros for a drink! My god, and people line up to pay it. The second interesting note was how we got into this bar, "Doobies." There was no door handle, just a buzzer and a slit with a slidable metal cover. So, ring the buzzer...and wait. Then I hear a faint series of knocks on the other side of the door, so what the hell? Match it. Then another series, match that again, and the door opens. We're in, and I wonder at my luck for getting into bars in Europe.
I also had my first dream in French the other night. Very cool, but I'm disappointed it took so long. That being said, the ol' French has gotten a lot better in the past couple of weeks. I've managed to have almost all French conversations at lunch and with some coworkers lately. I feel relatively accomplished.
"Garfield becomes an oddly surrealist comic when you remove the dialogue"
In any case, I'll see all you Calgary folk soon enough. I'm glad to be getting back in cowtown and rippin it up. Party on Wayne...
posted by Poshy at 9:22 AM
0 Comments:
Post a Comment
<< Home