GSoC2013 Coding Challenge
Implementing Machine Learning Algorithms for Classification and Feature Selection in Mothur
My complements on submitting your applications for the idea, we're glad to let you know that all of you have submitted good applications, some of which are pretty outstanding. We have almost 3 weeks before we make the final judgement, in this time we want to give you some interesting things to do so that you can grow a bit more understanding of the problem we are trying to address. Try any of the following challenges.
- Using any open source tools i.e. R, Octave, scikit-learn, libsvm, shogun ML toolbox etc and scripting, run any one of the ML and feature selection algorithms on the data provided
- Write a prototype implementation of SVM or ENET based feature selection and try it on the data provided, it doesn't have to be a very efficient implementation, a very crude one that serves as a proof of concept will do. You can use any high level languages like python, ruby or perl to make it easier, it doesn't have to be a C/C++ implementation and it doen't have to be mothur-compliant as well. As I said, just a proof of concept, make it simple but workable.
The output should be in the following format. The rank is a relative term, not an absolute value. It denotes the relative importance between the features.
OTU Rank Otu0022 5.55 Otu0077 0.93 Otu0840 0.82 Otu0299 0.8 Otu0170 0.79 Otu0566 0.78 Otu0372 0.78 Otu0365 0.77
Finishing any of the objectives (or if you can both) will count towards bonus points for you of getting selected for this idea.
- If you have trouble accessing the input data for the challenges, personally mail mentor Kathryn Iverson and Abu Zaher asking for the data.
- When you have accomplished any of the challenges, personally mail the code snippets with instructions on howto run them to mentor Kathryn Iverson and Abu Zaher. Plus do add comments in your proposal in the melange page so that all the mentors can know what you've accomplished.