Tuesday, September 21, 2010

IBM's furious Analytics aquisitions

Is anyone keeping up with IBM and their propensity to obtain Analytics based companies?  Let's see if we can do a recap of IBM in the news in the last couple of years. 
In actuality there is an estimated $12 billion dollars of 23 Analytics based companies that IBM has acquired in the last few years.  That is quite a leap for a hardware/software/IT company.  IEOR Tools has talked about IBM's emergence as an analytics company before with new analytic centers and acquisitions.  I think its safe to assume that IBM is the de-facto analytics champion in the world right now.

So what does this mean for Operations Research and their professionals?  I believe it means the sky is the limit now.  This is a grand opportunity for the Operations Research community.  In fact I would even say that if INFORMS does not take advantage of the recent demand for analytics and decision sciences then they are missing this big picture.  Jobs should be plentiful in the foreseeable future.  There should be plenty of work to keep management happy and help drive value into organizations.  This may even be the dawn of a new day for Operations Research and Analytics.  There is so much buzz now it will leave a ringing in your ears.  Sure I might be a little optimistic but I think any news is good news right now in this economy.

I also believe that IBM is not done.  I think IBM is going to evolve even more in the Analytics realm.  Perhaps getting more involved in the software within Operations Research and statistics.  Its just a guess but who knows if SAS, Matlab, or even contributing to open source projects like R, RapidMiner  or Weka.  This is an exciting time none of the less for Operations Research. 

Thursday, September 16, 2010

Current Data Mining and Analytics Challenges

I love the Data Mining and Analytics Challenges.  There tends to be so much collaboration and open knowledge especially if the challenge has an affiliated forum.  There really is so much to learn and the challenges offer a great way to bring all of the resources and knowledge together.  Here is a list of the current challenges underway in the Data Mining community.

  1. Kaggle is hosting a three competions.   Tourism Forecasting part one challenges to predict 581 tourism-related time series. Chess Ratings - Elo vs the rest of the World is trying to determine a chess rating system that is better than the current Elo rating system. INFORMS Data Mining Contest challenges to predict intra-day stock price movements based on experts predictions, sector data, and other indicators.
  2. TunedIT is another competition hosting organization.  Currently TunedIT is hosting the e-LICO mutli-omics prediction challenge with background knowledge on Obstructive Nephropathy.  Yes, I had to look it up too.
  3. UC San Diego is hosting the 2010 UC San Diego Data Mining Contest.  This is a two task contest which tries to predict e-tailer's data on consumer and non-consumer information.  The two tasks are a binary preditor and a boolean-transformed predictor.

Wednesday, September 15, 2010

OpenOpt release 0.31

New OpenOpt Suite release is out. This is free (license: BSD) and cross-platform (Linux, Windows, Mac etc) Python language modules for numerical optimization, automatic differentiation, solving systems of linear/nonlinear/ordinary differential equations, interpolation, integration etc.

OpenOpt 0.31:

  • Lots of new NLP, NSP (nonsmooth) and GLP (global) solvers from nlopt have been connected
  • New LP solver: pclp (very premature, but permissive license (BSD) and pure Python implementation)
  • Some bugfixes (mostly wrt using sparse matrices) and code cleanup

FuncDesigner 0.21:

  • New features: Integration, Translator
  • Some speedup for functions evaluation, automatic differentiation, optimization problems with some fixed variables
  • New parameter useSparse for optimization probs and automatic differentiation (sometimes autoselect works prematurely)
  • Some bugfixes (mostly wrt using sparse matrices) and code cleanup

DerApproximator 0.21:

  • Add parameter exactShape (prevents flattering operations on result)
Welcome to our homepage: http://openopt.org

Wednesday, September 8, 2010

Computer languages and Applied Math

There is no question that computer languages have helped pushed the envelope for applied mathematics.  It is hard to imagine where we would be without airline scheduling, supply chain management, or inventory control if it were not for all of the great advances in optimization and statistical computing.  I have thought a lot about the convergence of computing and Operations Research.  In fact I brought up a discussion on the topic on OR-Exchange with the question "Is programming skills a requirement for today's OR practitioner?"  You would think with all of the advances in computing that programming would be simpler but that is not the case.

There is an interesting debate in the R-project community about the shortcomings of the R language.  Xi'an Og posted a discussion on R shortcomings re-posted from another blog.  The consensus of the R community seems to be that R is an inferior language but has a brilliant library of resources.  So where does that leave the practioner?  Does the practioner need to update their coding skills and develop something better in another computer language?  I find it really interesting that some of the first solutions to this debate is to scrap everything and start over.

I don't think this debate is ever going to change.  The computer is always going to be a valuable tool for the Operations Research practitioner.  The tools we use to complete our daily tasks need to ubiquitous but also readily available.  Let's just say that the slide rule is not going to be making any sort of comeback. 

I believe that the Open Source model has a real advantage here over the proprietary counterparts in this debate.  The community has a lot of input into Open Source software.  It is often called a meritocracy.  The best solutions continue while those that do not go away in obscurity in the Open Source model.  This is one of the reasons why I advocate Open Source software.  In the end I think R is going to be fine.  There will be advances, possible even forks of the software, but there will always be progress.  The only limitations seem to be of what we could dream.