I liked Windows 95, but I must have failed to look at the multimedia samples. I'm not a great judge of music, but in regard to this particular song it doesn't seem to me I missed a whole lot.
It was in the "fun section". Actually it repeatedly got stuck on my Win95, which sounded like "Good times, ba-a-a-a-a" and then blue-screened. This is the main reason I remember it well; I guess it was a driver problem but never figured it out. Instead I installed Slackware/Linux a few months later...
But I find it amazing that all this was 20 years ago...
There was incredible progress in hardware since then, but surprisingly little progress in software - Google's pagerank algorithm and a bit of machine learning is really new, otherwise the windows desktop even looks almost the same today, but of course needs much more memory and processor speed.
Speaking of machine learning and its possible extended applications, I think there is a nonzero probability that you would enjoy reading Leslie Valiant's book "Probably Approximately Correct" if you haven't already read it. He believes that the concepts he and others have developed in machine learning may be applicable to evolutionary dynamics and the way humans make decisions to survive in a "theoryless" environment.
There may be nothing new in the book for you, but his ideas really helped quantify some things that I'd been thinking about for a while and were bothering me. Anyway, if you're interested, there's an Amazon link below. (Sorry about being so far off topic, but I've kind of been wanting to recommend this book to you and have had difficulty finding an appropriate opportunity. Also, I think the Amazon editorial and customer reviews are mostly worthless.)
I'm not at all certain there would be much new to you in the book, or even that the topic is among those that you currently find interesting. However, it has clarified my thinking considerably about some topics that I currently find interesting. After reading the book, the only review that I thought was on point was the one by Avi Wigderson where he wrote, "The suggested theory has all the characteristics of a great one. It is simple, general, and falsifiable, and moreover seems probably, approximately, correct!”
It seems like machine learning algorithms might be useful in searching for short term market inefficiencies. Do hedge funds make much use of them, or is that something you can't comment on.
Quants do use machine learning but one problem is the low signal to noise ratio and another one (non)stationarity. In other words, markets change their behavior, e.g. because quants figure out inefficiencies and trade them. That is as much as I want to say about that ...
>> With the Master Algorithm in hand and data as its fuel, machine learning—essentially the automation of discovery, a kind of scientific method on steroids—will become the most powerful technology humanity has ever devised. And The Master Algorithm will be its bible. (From the synopsis of Domingos' book.)
Making that sort of solid prediction of the future in the realm of what Valiant would call the theoryless is often harder than it looks. I'm somewhat certain that Valiant would agree that having the Master Algorithm isn't going to change that much.
11 comments:
... on the Windows 95 installation CD.
I liked Windows 95, but I must have failed to look at the multimedia samples. I'm not a great judge of music, but in regard to this particular song it doesn't seem to me I missed a whole lot.
It was in the "fun section".
Actually it repeatedly got stuck on my Win95, which sounded like "Good times, ba-a-a-a-a" and then blue-screened.
This is the main reason I remember it well; I guess it was a driver problem but never figured it out. Instead I installed Slackware/Linux a few months later...
But I find it amazing that all this was 20 years ago...
There was incredible progress in hardware since then, but
surprisingly little progress in software - Google's pagerank
algorithm and a bit of machine learning is really new,
otherwise the windows desktop even looks almost the same today,
but of course needs much more memory and processor speed.
>> and a bit of machine learning is really new
Speaking of machine learning and its possible extended applications, I think there is a nonzero probability that you would enjoy reading Leslie Valiant's book "Probably Approximately Correct" if you haven't already read it. He believes that the concepts he and others have developed in machine learning may be applicable to evolutionary dynamics and the way humans make decisions to survive in a "theoryless" environment.
There may be nothing new in the book for you, but his ideas really helped quantify some things that I'd been thinking about for a while and were bothering me. Anyway, if you're interested, there's an Amazon link below. (Sorry about being so far off topic, but I've kind of been wanting to recommend this book to you and have had difficulty finding an appropriate opportunity. Also, I think the Amazon editorial and customer reviews are mostly worthless.)
http://www.amazon.com/Probably-Approximately-Correct-Algorithms-Prospering/dp/0465060722/ref=sr_1_1?s=books&ie=UTF8&qid=1437403287&sr=1-1&keywords=leslie+valiant
Lee,
thank you for the link.
Btw wikipedia has an entry about PAC learning.
Wolfgang,
I'm not at all certain there would be much new to you in the book, or even that the topic is among those that you currently find interesting. However, it has clarified my thinking considerably about some topics that I currently find interesting. After reading the book, the only review that I thought was on point was the one by Avi Wigderson where he wrote, "The suggested theory has all the characteristics of a great one. It is simple, general, and falsifiable, and moreover seems probably, approximately, correct!”
If anybody is interested to have PAC learning explained by Scott A. , then I recommend to follow this link (pdf).
Wolfgang,
It seems like machine learning algorithms might be useful in searching for short term market inefficiencies. Do hedge funds make much use of them, or is that something you can't comment on.
Quants do use machine learning but one problem is the low signal to noise ratio and another one (non)stationarity. In other words, markets change their behavior, e.g. because quants figure out inefficiencies and trade them.
That is as much as I want to say about that ...
Another book about machine learning I should read.
>> With the Master Algorithm in hand and data as its fuel, machine learning—essentially the automation of discovery, a kind of scientific method on steroids—will become the most powerful technology humanity has ever devised. And The Master Algorithm will be its bible. (From the synopsis of Domingos' book.)
Making that sort of solid prediction of the future in the realm of what Valiant would call the theoryless is often harder than it looks. I'm somewhat certain that Valiant would agree that having the Master Algorithm isn't going to change that much.
Post a Comment