Netflix Prize: Forum

Forum for discussion about the Netflix Prize and dataset.

You are not logged in.

Announcement

Congratulations to team "BellKor's Pragmatic Chaos" for being awarded the $1M Grand Prize on September 21, 2009. This Forum is now read-only.

#1 2009-09-22 21:49:36

Mathematica_Expert
Member
Registered: 2009-09-21
Posts: 4

Need Partners for Prize 2: Wolfram/Mathematica Gurus

I'm planning on competing on something other than pure speed for Round 2.  Using Mathematica will not have the speed of Java, but there are several other benefits like transparency and a unique library of interdisciplinary methods that can be utilized in non-traditional way to measure efficiency (like Data Envelopment Analysis).

Any takers?

Offline

 

#2 2009-09-23 00:42:35

CS1
Member
From: San Jose, CA
Registered: 2006-10-02
Posts: 151

Re: Need Partners for Prize 2: Wolfram/Mathematica Gurus

What's the license fee like for commercial use?  I imagine it's prohibitive.

Why Mathematica and not Python?  Or Matlab?  The last time I used Mathematica, I was in high school, eons ago, and, no offense intended, I still regard it as a toy for schoolchildren.

CS1

Offline

 

#3 2009-09-23 06:28:45

Mathematica_Expert
Member
Registered: 2009-09-21
Posts: 4

Re: Need Partners for Prize 2: Wolfram/Mathematica Gurus

CS1,
I'll try not to be offended sad     Here is a post by some professor.  Sure it's bias, but I think you will find 80% of the content objective evidence.   True, the cost is over $1000, but you can't put a price on learning and fun smile

Reply with your thoughts.



----------------------------------------------------------

Why Mathematica?
All our research courses use Mathematica® as the tool for data analysis and visualization.  Mathematica® has the following advantages and disadvantages. (We also describe how we overcome the disadvantages.)

Advantages:
*Used by many top researchers in all scientific disciplines around the world today
*Most scientific functions are already built in
*One of the industry's most powerful visualization engine is built in
*Can read and write most file formats without any extra coding
*Offers the shortest path between hypothesis and validation

Disadvantages
*Expensive.  (Resolved by iSabio's site license deal with Wolfram to provide affordable licenses to the students)
*Slow.  (Resolved by using multiple computers and compiled code)
*Too convenient.  It will spoil the student's programming ability with other computer languages.  (Resolved by teaching the students procedural implementation first, then teach the list processing equivalent


The advantages far outweigh the disadvantages.  While the slow execution of the code (inherent to all interpreted programming language) is a clear disadvantage, code execution is only 1% of the total research time.  In any research, most time is spent reading, thinking and trying out the ideas.  In the computer languages that lack the visualization tools, or file import/export conversion tools, researchers are forced to spend weeks, if not months, implementing those features only at a barely functioning level.   Mathematica® on the other hand will allow the researchers to focus on the research most of the time.  As for the slow execution of the code, iSabio has tens of computers, all equipped with the latest version of Mathematica®, ready to run our researchers' code whenever necessary.

Anyone can purchase Mathematica license for their home use from our Mathematica Purchase Page.

Mathematica vs. Maple
Maple is a competition of Mathematica.  These two software packages have similar purpose and capabilities and you can find true believer of either one.  We feel learning either one well is time much better spent than comparing these two endlessly.  We chose Mathematica for our own reasons that placed students' interest first.  For objective comparison, please refer to Stefan Steinhaus's site and his 2008 report.
Mathematica vs. Matlab

Matlab has a different purpose than Mathematica. Matlab, which stands for Matrix Lab, is designed for number crunching applications whereas Mathematica was designed for symbolic mathematics.  Many of our research projects are computational in nature, and is likely to be better served by Matlab.  However, we chose Mathematica becaues it will be more useful for our students because this is only the beginning of their career and they will still go high school and college courses, taking calculus and differential equations.  Mathematica will be a better companion on their math courses while it can still do everything Matlab does. 

Matlab's main advantage over Mathematica used to be its speed.  But, as you can see in this speed comparison (pg 29), Mathematica now exceeds Matlab in speed in many types of computations and does fall by much in other computations.
Learning to use these these packages is very much like learning to drive trucks vs. busses vs. cars.  They all have different purpose and different advatages/disadvantages.  But if you learn to drive one well, you will learn to drive others easily.   Not knowing how to drive any of them is only losing deal.

Last edited by Mathematica_Expert (2009-09-23 06:30:29)

Offline

 

#4 2009-09-23 19:55:51

edr2
Member
Registered: 2009-07-02
Posts: 58

Re: Need Partners for Prize 2: Wolfram/Mathematica Gurus

Mathematica_Expert wrote:

*Slow.  (Resolved by using multiple computers and compiled code)
.
.
The advantages far outweigh the disadvantages.  While the slow execution of the code (inherent to all interpreted programming language) is a clear disadvantage, code execution is only 1% of the total research time.  In any research, most time is spent reading, thinking and trying out the ideas.

Mathematica Expert,

I've only seen demos of Mathematica, and it looks like a really cool tool for calculation (as is MATLAB).  Slow, however can be a real problem in this problem domain. If you hunt down some of the papers where people describe their methods, your will see that some of them (but not all!) can take hours or even days (e.g. RBMs) to run in Java or C++.  If you have some serious horsepower at your disposal, this might not be a problem and  Mathematica or Matlab might be the way to go.  But you might also want to stay open to the possibility of doing at least some stuff in C/C++ or Fortran or some other traditional optimizing compiled language, as painful as it might be from a development standpoint.  The NFP was one of the few computational things I have worked on where it actually took more time to run code than it did to write it!

best regards,

EdR

Offline

 

Board footer

Powered by PunBB
© Copyright 2002–2005 Rickard Andersson