I decided to reveal my method of predicting ratings in very short time.
Details can be found here.
This stuff is commentware :-)
Enjoy!
Sunday, September 20, 2009
Wednesday, July 22, 2009
Fast algorithm v2
Here's another one: 0.8840 in 8 minutes. Machine spec as before.
I'm quite sure this approach would lead to much better solution, maybe even 0.8560 :-) I wish I had more time (3! months would be enough) or - at least - 12! times faster machine...
I'm quite sure this approach would lead to much better solution, maybe even 0.8560 :-) I wish I had more time (3! months would be enough) or - at least - 12! times faster machine...
Wednesday, February 11, 2009
Fast algorithm
Fast CF algorithms have a future, as I don't think that blending hundreds (or even dozens) of predictors is acceptable in real recommender system.
Unfortunately, most of published algorithms tried in this contest needs hours of computations to produce a decent result.
I've been recently working on my "fast" algorithm; the goal was to develop an algorithm with lowest RMSE and running time not exceeding 10 minutes.
Now it's time I may proudly announce the result: my algorithm can produce a 0.8927 QRMSE solution in seven (7) minutes (including reading raw data and writing result file) on my AMD 64 X2 Dual Core @5200+ machine, without using parralelism of any kind (like CUDA, OpenMP, etc.). It's not a totally new algorithm, but rather a clever modification of what was already published by one of the leading teams.
If anyone has an algorithm which offers a better solution in similar time, please let me know.
Unfortunately, most of published algorithms tried in this contest needs hours of computations to produce a decent result.
I've been recently working on my "fast" algorithm; the goal was to develop an algorithm with lowest RMSE and running time not exceeding 10 minutes.
Now it's time I may proudly announce the result: my algorithm can produce a 0.8927 QRMSE solution in seven (7) minutes (including reading raw data and writing result file) on my AMD 64 X2 Dual Core @5200+ machine, without using parralelism of any kind (like CUDA, OpenMP, etc.). It's not a totally new algorithm, but rather a clever modification of what was already published by one of the leading teams.
If anyone has an algorithm which offers a better solution in similar time, please let me know.
Wednesday, January 21, 2009
More gigabytes on board
My machine has 8GB of RAM now. I could still work with 4GB, but - you know - memory got so cheap these days... It's not worth using error-prone data & structure packaging any longer (I mean more error-prone than cf algorithms themselves). Not to mention that this challenge is not the only thing I'm currently working on.
There's bad news though: my motherboard won't accept more RAM, so the next step would be to replace the whole mobo ;-)
I still remember buying additional RAM for my first PC machine: it was in mid-1990's, I was doubling my machine's 4MB (some silly game required 8MB). BTW: Windows 95 worked really nice with 4MB, as far as I remember.
Both me and my machine(s) have changed a lot since then...
Let's find out if more RAM helps in this contest.
There's bad news though: my motherboard won't accept more RAM, so the next step would be to replace the whole mobo ;-)
I still remember buying additional RAM for my first PC machine: it was in mid-1990's, I was doubling my machine's 4MB (some silly game required 8MB). BTW: Windows 95 worked really nice with 4MB, as far as I remember.
Both me and my machine(s) have changed a lot since then...
Let's find out if more RAM helps in this contest.
Tuesday, January 13, 2009
Introduction
Hi!
My name is Wojtek Kulik. I used to hide under nickname "photek" (guess why, a limited collection handshake offered for the first correct solution; HINT: there are TWO reasons), but from now on I'll be using my real name only. I wish I could change my team's name, but apparently it's not possible (that means I've slightly polluted the leaderboard - sorry for that).
To quote Terry Pratchett - I live behind a keyboard in Warsaw (Poland), working on my startup project. As it's related to prediction, I decided to give a try in Netflix Prize challenge, hoping that success of my startup depends on the success of my approach to this challenge.
I've managed to get into TOP-40 recently. But to be honest, it's not enough for me... I've got plenty of ideas on my TODO list; unfortunately my current bottleneck is computing power.
Enough writing. Time to make a submission :-)
My name is Wojtek Kulik. I used to hide under nickname "photek" (guess why, a limited collection handshake offered for the first correct solution; HINT: there are TWO reasons), but from now on I'll be using my real name only. I wish I could change my team's name, but apparently it's not possible (that means I've slightly polluted the leaderboard - sorry for that).
To quote Terry Pratchett - I live behind a keyboard in Warsaw (Poland), working on my startup project. As it's related to prediction, I decided to give a try in Netflix Prize challenge, hoping that success of my startup depends on the success of my approach to this challenge.
I've managed to get into TOP-40 recently. But to be honest, it's not enough for me... I've got plenty of ideas on my TODO list; unfortunately my current bottleneck is computing power.
Enough writing. Time to make a submission :-)
Subscribe to:
Posts (Atom)
