Friday, December 18, 2009

Artificial Intelligence with Python

While stumbling across cyberspace looking for some interesting Python tools and tutorials I found this rather interesting webcast. The video is of Raymond Hettinger at PyCon 2009. Raymond describes the usefulness of Python with applying artificial intelligence and data mining. This talk is very interesting to see how useful of a tool Python can be to performing some relevant Operations Research tasks specifically with data manipulation and learning. As I have mentioned before Open Source software offers a lot in the way of Operations Research tools. I hope you enjoy it as much as I did.

Source:
http://blip.tv/file/1947373

4 comments:

Bernoulli-Blogger said...

Thanks for posting this. I have been slowly teaching my kids Python also. I also ended up watching some of the other PyCon videos. I thought the one on coverage testing was also quite good. http://blip.tv/file/1947218 While not directly related to Operations Research, seeing the challenges in code coverage could present some interesting operations problems.

Larry said...

Thanks for the feedback. Some other interesting topics that you may know about in Python

PyMathProg
CVXOPT
Numpy/Scipy
OpenOpt

Bernoulli-Blogger said...

I have played with these.
Here are my thoughts

PyMathProg:
Takes a bit of thought to comprehend how to process the results. Uses GLPK has the engine to process the math. The web site has great examples. No sensitivity analysis or shadow price available to the tool, even though I sure that GLPK can do so.

CVXOPT:
I tried it can any remember much about it. Also no No sensitivity analysis or shadow price available.

Numpy/Scipy:
Great toolkit, aimed mainly at scientific needs which has a lot of crossover for Operations. Needs other tools to get results for PL.

OpenOpt:
So far my favorite toolkit of the group for LP. Uses Numpy. Since matrix setup because you describe the problem in the same way as if you took a PL lecture. Examples show the traditional constraints subject to setup. But also no sensitivity analysis or shadow price available.

Another one that was also written in Python but does not do SA or SP is
"pulp-or" puLP: An LP modeler in Python. http://code.google.com/p/pulp-or/. Just like PyMathProg, it does have nice examples. http://code.google.com/p/pulp-or/wiki/OptimisationWithPuLP?tm=6

I am soon looking into using GNU R API to Python. GNU R can do what I am looking for with SA and SP and the API from Python would be like using the same idea as PyMathProg.

Larry said...

Awesome feedback! Thanks for the detailed reviews.