So it is essentially like a Netflix type system, but with books?
I'm curious what length of time they need you to hold the book in order for them to remain profitable with shipping and whatnot. Netflix has stated what that time period is and the person needs to keep movies longer than 7 days or something like that to remain profitable on that user.
Launch.com currently allows you to vote on a scale how much you like/dislike a song - then based on that it recommends other songs to you. But it doesn't analyze anything in the acutal music.
For that, I would recommend FFT and backprop Neural Nets being added to the existing ranking methods that they have - but in the end, your own brain is likely better at it.
What's sad is it is all too true. Instead of innovating, a lot of OSS projects that are supposed to be like MS apps usually just mimick, rather than truly innovate.
Perhaps because most of the time "truly innovating" is a waste of time.
People sure do love to hate MS because they are huge and because of that push people around.
But when it comes to UI design - both they and Apple have the money to do a lot of research into actual usability.
Something that becomes obvious when you use "truly innovative" software - when someone tries to do something new just because it seems right to them.
More often then not, it looks cool, but proves useless for day to day use. (a few of MS's "features" are much like this - fading menus and such, but some people love them. Apple too has much fluff, or dare I say cruft? but for the most part, they have a very strongly researched base of design methods, hence why they are mimicked)
Obviously there are exceptions - but for the most part, MS is oft imitated because they have already invested literally millions of dollars and tons of time in research into making products that people can sit down and use.
(I'm sure someone will chime in and say that vi is far more usable for themselves and that an luser that can't see that is an idiot. But the obvious point should be that when designing for the massees - there are certain techniques that will be seen over and over - because they work.
All that could be summed up in "why reinvent the wheel?"
mod this WAY up - I have no points right now, but this is an excellent question!
it is frustrating to always see mention of how horrible this or that on the net is, but it would be excellent to see some real numbers to gnaw on and fondle.
they have great designers - can't fault them there.
but I feel that they must have drunk engineers - esp. in the power supply department.
they had to recall tons of laptop power supplies because they would get too hot and cause fires. they would exchange them for free with (still warm) improved ones.
then they put out their current laptops that get hot enough to burn - but that isn't all power supply and is largely also in part of the processor and graphics card being warm little things - and then the aluminum covers conducting heat well... as aluminum does:)
and then this - and I would imagine that noise is a HUGE problem with many Mac types since they use it for video and sound editing.
I personally just go for cheap and fast systems that I can cluster - and I spend a little more when I build them to get the quiet stuff - one computer that is loud is bad enough, but a few of them that are loud... ugh!
imagine being in a design house that has many of these loud G4s... at least they are shiny. you can sit back and be glad about hot pretty it is while rubbing your temples to try to alleviate the headaches:)
what happened to being in the stupio for a week, or a month?
out of all of that money, what percent of it is totally wasted on the lifestyle of being a pop star?
I don't see what the issue is, they still make a profit in that model - if they aren't making "enough" profit, then that is a sign that one of the variables in the mix is off. They are assuming that the variable that is off is the sales - the easiest one to point the blame on - but as in all business, perhaps the easier way is to be introspective - look inward and see where to cut costs.
There are a crapload of bands out there that make albums faster and cheaper than that.
If AOL would subsidize this, they would see their security problems disappear overnight.
also - I think Dick Tracy foreshadowed the cracking method used by these kids years ago with its "Mumbles" character. So by using that as an indicator, we should next look for people wearing bright colors and having odd facial features to be part of the next crack.
but the picture they show is wrong - the sedan shape will stay for a bit and the hatchback isn't coming back for some time yet unfortunately.
the key being on the center console just makes sense to me - it saves from knee laceration in an accident, and it is also a leftover in efficiency back when they were rally raced years ago (faster to get in and turn on there).
bmw and volvo are what people usually associate with the safest cars, but Saabs are the safest in real world accidents. They are incredible and put engineering first. I find it amusing when people dislike them
Red Hot Chili Peppers must have known
on
The Taste of Pain
·
· Score: 1
They've had a song "Taste the Pain" since before 1989.
the car mags all say that it SUCKS - in theory it is good, but it leaves one feeling detached and the response ends up being jerky and hard to control.
the current ones have fiber optics to connect all the electronic - for future upgrades and high data bandwidth.
I'm moving soon, so I can't have my saab - but I enjoyed it while I had it.
ergonomics are all very personal things - I personally *hate* the BMW ergonomics, but like the general look of their new cars. Saabs are quirky, but the ergonomics are perfect for me. that is the usual complaint of Saabs - but they are great cars - and safe as hell. you could drive one full speed into a brick wall and be just fine. although I'm not sure why you'd want to:)
Oh come on - I'm not really an Apple fan at all - but all companies do this (or rather things like this). There are like 8 gazillion ways to avoid taxes via corporate uses - getting paid salary $1 means no tax on his salary and then he can live off of company perks that he would otherwise have to pay for were he doing it out of pocket.
Then again, the last one I heard of doing something like this was when Enron had that dude... don't recall his name - they bought him a house, then sold it to him for cheap, and then bought it back off of him.
They are all accountings ways out of company policies of the size of bonuses, as well as legal tax loopholes.
the main issue with them is that they cause us "normal folk" to look more closely at it and then it can expose other things that they might be doing that are... less legal.
as I understand it, you use it to determine how close it is to a random walk. you want things that distance themselves from it. I don't think there is such thing as one fixed point where it can always say "this is it" - but instead as things move away from 1/2.
other sites to check out (which are pretty high on google if I recall - those are the only ones that I have in my bookmarks folder - which was what eventually reminded me of the name): http://www.bearcave.com/misl/misl_tech/wav elets/hu rst/ http://www.speedderivs.com/hurst.html http: //www.cbi.polimi.it/glossary/Hurst.html
With that, you can see how if it is more than likely random non-trending data, or if there is a pattern in there.
Re:he has some good points, but also overlooks a f
on
Fooled by Randomness
·
· Score: 1
very true.
I agree with a lot of what Taleb has to say - which is why the bible code analogy bugged me. It is frustrating when someone says a series of things that are intelligent and then goes off on a tangent like that.
the feedback loop is a problem you can't get around with stock market models. the more money you put into your model, the more you directly effect the movement of the stock and then throw off the model (I'm pretty sure I just wrote somehthing like that the other day in my journal).
the two obvious ways around this that firms use are 1) don't talk about what you do and trade "small", and 2) continuously update that trading strategy.
Saying to do XYZ today might not work out tomorrow, so find your model, trade using it, and then reevaluate your model again now that there is new data in there. and then trade again. a simplistic view - but it works.
the reason you will never see always up (or always down for that matter, although that is technically easier) in terms of bringing in money is for exactly what Taleb says the best (and also most obvious part) - it is the unknown. You might have an ideal model that will work out to give you a great series of correct runs over time - but it won't (and can't) take into account another 9/11 type event.
There are many insurance and reinsurance companies out there using neural nets to do risk assessment and are looking at data and calculating how likely a hurricane will hit Bermuda at T time on D date, or even the 5 or so firms (and schools) that are analyzing with NNs to watch for terrorist/war threats coming up.
It would be interesting to combine the output of all of those efforts into a stock model to try to account for future unknowns... but it is impossible. The neural net outputs data that has some level of error even if you give it perfectly accurate data. But if you feed it data that already has its own data (from the NN that it came from), it will then go on to produce much larger error as it goes through the new NN.
Anyway, the main point of contention was that Taleb made a really bad analogy that bothered me and made it sound like it was not possible to predict the market at all. To always get it right, nope - but to get very close over and over again - yes.
It is then a matter of using that exactly the way he does - maximize profit and minimize loss. The guys that are risking it all time and time again are either stupid or cocky.
Sorry man - but there is. Specifically dealing with Time Series data. And not that it tells you if it is random, but more importantly that it isn't random. It allows you to see if given a set of data, you can then get future events from that.
I'm at work now and don't have any of my notes here and I haven't had to use the damn thing in a few months (been working pretty much all in genetic algorithms lately) so I guess my brain just dumped it out.
and yeah, that last part I of course agree with since that is pretty much true for all that I can think of right now - that is the whole point of using neural net analysis on time series data - it is nonlinear analysis.
When I remember and/or look up the damn thing I will post it.
he has some good points, but also overlooks a few
on
Fooled by Randomness
·
· Score: 2, Interesting
The main issue I have with what he says is comparing Time Series analysis (a topic that is near and dear to my heart), to the people that read the bible and then find prophetic phrases in there. Aside from that being a bad analogy (but it sounds good if you don't know anything of time series analysis), it is also misleading in that time series analysis can be proven if it will work or not.
There is an equation (that is f'ing driving me nuts right now that I can't recall the name of it) that will tell you if your time series data is truly random or if it leads to predictability. It has been shown over and over again that the stock market, were it a perfect market would in fact be random - but since there is human intervention that drives it, it is no longer truly random and there are non-linear patterns that are within the data.
The human brain is wired to pick up on linear patterns (to the point we see them where they are in fact not really there - it is advantageous to us to see a lion in the brush and run - when in fact there isn't a lion there, than the other way around and get mauled by a lion that we never saw). That is more to what he speaks of, humans seeing what isn't there (there are a crapload of great books written about this - How We Know What Isn't So: The Fallibility of Human Reason in Everyday Life by Thomas Gilovich immediately comes to mind) - but the way he speaks, he also includes computational analysis in there and is wrong in his generalization.
Now it is going to bug me all day as to what that frickin' algorithm is called - the one that shows the strength of your time series data... I know it doesn't start with an A, or a Z... and I don't think there is a C there... other than that, I'm pretty much at a total blank.
Slashdot Mirror
It is all web based.
nice - no physical delivery costs, but the cost instead of maintaining servers and bandwidth.
there are those that will still complain that they prefer the feel of it in their hands - but online and searchable is very nice.
*note to self: read more of the articles before blabbering on like an idiot.
So it is essentially like a Netflix type system, but with books?
I'm curious what length of time they need you to hold the book in order for them to remain profitable with shipping and whatnot.
Netflix has stated what that time period is and the person needs to keep movies longer than 7 days or something like that to remain profitable on that user.
Launch.com currently allows you to vote on a scale how much you like/dislike a song - then based on that it recommends other songs to you.
But it doesn't analyze anything in the acutal music.
For that, I would recommend FFT and backprop Neural Nets being added to the existing ranking methods that they have - but in the end, your own brain is likely better at it.
I used to work at a company that wrote software for college/university settings. The admin software for the bursar, registrar, housing, etc offices.
One of the things it did was what you said - but not for individual use - it was part of the overall class and billing scheduling system.
In a similar manner - the El Farol problem too is being used towards financial models (amonger other things).
It has a basis of swarm theory behind it - but more things as well that are kind of fun as well - human behavior and psychology and the like.
I just googled all over my keyboard.
someone hand me a towel.
What's sad is it is all too true. Instead of innovating, a lot of OSS projects that are supposed to be like MS apps usually just mimick, rather than truly innovate.
Perhaps because most of the time "truly innovating" is a waste of time.
People sure do love to hate MS because they are huge and because of that push people around.
But when it comes to UI design - both they and Apple have the money to do a lot of research into actual usability.
Something that becomes obvious when you use "truly innovative" software - when someone tries to do something new just because it seems right to them.
More often then not, it looks cool, but proves useless for day to day use. (a few of MS's "features" are much like this - fading menus and such, but some people love them. Apple too has much fluff, or dare I say cruft? but for the most part, they have a very strongly researched base of design methods, hence why they are mimicked)
Obviously there are exceptions - but for the most part, MS is oft imitated because they have already invested literally millions of dollars and tons of time in research into making products that people can sit down and use.
(I'm sure someone will chime in and say that vi is far more usable for themselves and that an luser that can't see that is an idiot.
But the obvious point should be that when designing for the massees - there are certain techniques that will be seen over and over - because they work.
All that could be summed up in "why reinvent the wheel?"
mod this WAY up - I have no points right now, but this is an excellent question!
it is frustrating to always see mention of how horrible this or that on the net is, but it would be excellent to see some real numbers to gnaw on and fondle.
exactly - I was going to say there doesn't look to be any use of shared memory (not even sure how they would).
I'm curious if they are writing stuff with something like PVM for a beowulf type thing, or the more likely OpenMosix option.
Also curious as to what they are using it for. (didn't read the article but checked out that briefcase b/c it sounded and looked cool)
they have great designers - can't fault them there.
:)
:)
but I feel that they must have drunk engineers - esp. in the power supply department.
they had to recall tons of laptop power supplies because they would get too hot and cause fires. they would exchange them for free with (still warm) improved ones.
then they put out their current laptops that get hot enough to burn - but that isn't all power supply and is largely also in part of the processor and graphics card being warm little things - and then the aluminum covers conducting heat well... as aluminum does
and then this - and I would imagine that noise is a HUGE problem with many Mac types since they use it for video and sound editing.
I personally just go for cheap and fast systems that I can cluster - and I spend a little more when I build them to get the quiet stuff - one computer that is loud is bad enough, but a few of them that are loud... ugh!
imagine being in a design house that has many of these loud G4s... at least they are shiny. you can sit back and be glad about hot pretty it is while rubbing your temples to try to alleviate the headaches
what happened to being in the stupio for a week, or a month?
out of all of that money, what percent of it is totally wasted on the lifestyle of being a pop star?
I don't see what the issue is, they still make a profit in that model - if they aren't making "enough" profit, then that is a sign that one of the variables in the mix is off.
They are assuming that the variable that is off is the sales - the easiest one to point the blame on - but as in all business, perhaps the easier way is to be introspective - look inward and see where to cut costs.
There are a crapload of bands out there that make albums faster and cheaper than that.
These boys need to get laid.
If AOL would subsidize this, they would see their security problems disappear overnight.
also - I think Dick Tracy foreshadowed the cracking method used by these kids years ago with its "Mumbles" character.
So by using that as an indicator, we should next look for people wearing bright colors and having odd facial features to be part of the next crack.
still on the floor.
but the picture they show is wrong - the sedan shape will stay for a bit and the hatchback isn't coming back for some time yet unfortunately.
the key being on the center console just makes sense to me - it saves from knee laceration in an accident, and it is also a leftover in efficiency back when they were rally raced years ago (faster to get in and turn on there).
bmw and volvo are what people usually associate with the safest cars, but Saabs are the safest in real world accidents. They are incredible and put engineering first.
I find it amusing when people dislike them
They've had a song "Taste the Pain" since before 1989.
Those guys are brilliant.
the car mags all say that it SUCKS - in theory it is good, but it leaves one feeling detached and the response ends up being jerky and hard to control.
perhaps over time they will get that worked out
I loved my '00SE.
:)
the current ones have fiber optics to connect all the electronic - for future upgrades and high data bandwidth.
I'm moving soon, so I can't have my saab - but I enjoyed it while I had it.
ergonomics are all very personal things - I personally *hate* the BMW ergonomics, but like the general look of their new cars.
Saabs are quirky, but the ergonomics are perfect for me.
that is the usual complaint of Saabs - but they are great cars - and safe as hell. you could drive one full speed into a brick wall and be just fine. although I'm not sure why you'd want to
1) try pants
2) men in trenchcoats... not your friends
defacing a web page != stealing credit cards.
they shouldn't have equal sentences, but that isn't to say one of them isn't deserving of what they get...
Oh come on - I'm not really an Apple fan at all - but all companies do this (or rather things like this).
There are like 8 gazillion ways to avoid taxes via corporate uses - getting paid salary $1 means no tax on his salary and then he can live off of company perks that he would otherwise have to pay for were he doing it out of pocket.
Then again, the last one I heard of doing something like this was when Enron had that dude... don't recall his name - they bought him a house, then sold it to him for cheap, and then bought it back off of him.
They are all accountings ways out of company policies of the size of bonuses, as well as legal tax loopholes.
the main issue with them is that they cause us "normal folk" to look more closely at it and then it can expose other things that they might be doing that are... less legal.
I thought it described "goblazing windows"
I was trying to figure out if that was some sort of new windowmanager or a theme... now I see that I'm just retarded.
as I understand it, you use it to determine how close it is to a random walk. you want things that distance themselves from it.
v elets/hu rst/: //www.cbi.polimi.it/glossary/Hurst.html
I don't think there is such thing as one fixed point where it can always say "this is it" - but instead as things move away from 1/2.
other sites to check out (which are pretty high on google if I recall - those are the only ones that I have in my bookmarks folder - which was what eventually reminded me of the name):
http://www.bearcave.com/misl/misl_tech/wa
http://www.speedderivs.com/hurst.html
http
Hurst Exponents.
That is what I was thiking of.
With that, you can see how if it is more than likely random non-trending data, or if there is a pattern in there.
very true.
I agree with a lot of what Taleb has to say - which is why the bible code analogy bugged me. It is frustrating when someone says a series of things that are intelligent and then goes off on a tangent like that.
the feedback loop is a problem you can't get around with stock market models. the more money you put into your model, the more you directly effect the movement of the stock and then throw off the model (I'm pretty sure I just wrote somehthing like that the other day in my journal).
the two obvious ways around this that firms use are 1) don't talk about what you do and trade "small", and 2) continuously update that trading strategy.
Saying to do XYZ today might not work out tomorrow, so find your model, trade using it, and then reevaluate your model again now that there is new data in there. and then trade again.
a simplistic view - but it works.
the reason you will never see always up (or always down for that matter, although that is technically easier) in terms of bringing in money is for exactly what Taleb says the best (and also most obvious part) - it is the unknown.
You might have an ideal model that will work out to give you a great series of correct runs over time - but it won't (and can't) take into account another 9/11 type event.
There are many insurance and reinsurance companies out there using neural nets to do risk assessment and are looking at data and calculating how likely a hurricane will hit Bermuda at T time on D date, or even the 5 or so firms (and schools) that are analyzing with NNs to watch for terrorist/war threats coming up.
It would be interesting to combine the output of all of those efforts into a stock model to try to account for future unknowns... but it is impossible.
The neural net outputs data that has some level of error even if you give it perfectly accurate data.
But if you feed it data that already has its own data (from the NN that it came from), it will then go on to produce much larger error as it goes through the new NN.
Anyway, the main point of contention was that Taleb made a really bad analogy that bothered me and made it sound like it was not possible to predict the market at all.
To always get it right, nope - but to get very close over and over again - yes.
It is then a matter of using that exactly the way he does - maximize profit and minimize loss.
The guys that are risking it all time and time again are either stupid or cocky.
Sorry man - but there is. Specifically dealing with Time Series data. And not that it tells you if it is random, but more importantly that it isn't random. It allows you to see if given a set of data, you can then get future events from that.
I'm at work now and don't have any of my notes here and I haven't had to use the damn thing in a few months (been working pretty much all in genetic algorithms lately) so I guess my brain just dumped it out.
and yeah, that last part I of course agree with since that is pretty much true for all that I can think of right now - that is the whole point of using neural net analysis on time series data - it is nonlinear analysis.
When I remember and/or look up the damn thing I will post it.
The main issue I have with what he says is comparing Time Series analysis (a topic that is near and dear to my heart), to the people that read the bible and then find prophetic phrases in there.
Aside from that being a bad analogy (but it sounds good if you don't know anything of time series analysis), it is also misleading in that time series analysis can be proven if it will work or not.
There is an equation (that is f'ing driving me nuts right now that I can't recall the name of it) that will tell you if your time series data is truly random or if it leads to predictability.
It has been shown over and over again that the stock market, were it a perfect market would in fact be random - but since there is human intervention that drives it, it is no longer truly random and there are non-linear patterns that are within the data.
The human brain is wired to pick up on linear patterns (to the point we see them where they are in fact not really there - it is advantageous to us to see a lion in the brush and run - when in fact there isn't a lion there, than the other way around and get mauled by a lion that we never saw).
That is more to what he speaks of, humans seeing what isn't there (there are a crapload of great books written about this - How We Know What Isn't So: The Fallibility of Human Reason in Everyday Life by Thomas Gilovich immediately comes to mind) - but the way he speaks, he also includes computational analysis in there and is wrong in his generalization.
Now it is going to bug me all day as to what that frickin' algorithm is called - the one that shows the strength of your time series data... I know it doesn't start with an A, or a Z... and I don't think there is a C there... other than that, I'm pretty much at a total blank.