|
follow-up on KDnuggets poll: http://www.kdnuggets.com/polls/2011/algorithms-analytics-data-mining.html Also - which new methods have you learned this year? |
|
i use libsvm as my classifier tool for remote image processing ! |
|
KMeans, Kernel KMeans, Spectral Clustering, SVMs with chi2, RBF and linear kernel, Linear SGD SVMs, Approximate kernel SVMs, SVR, Neural Networks, Gaussian mixture models, CRFs, Quickshift, Mean Shift, Normalized Cuts, Multi-Instance Kernels, Multi-Instance SVMs, SVR-SVM for multiple instance learning, Non-local means and I tried the decision trees in sklearn once ;) I learned this year about lots of approximate SVM things this year and also a lot about multiple instance learning. I finally learned how alpha expansion and alpha-beta swaps are done to minimize submodular energies.
This answer is marked "community wiki".
That's a lot of stuff! How many algorithms did you code yourself? If you didn't code them, where did you obtain the software?
(Nov 19 '11 at 16:14)
Noel Welsh
1
I coded some neural networks stuff myself, some Multi-Instance stuff, kernel approximation things and I think quickshift in python. I often use my labs CUDA/Python lib to speed things up. But mostly, I use scikit-learn. It is a python machine learning library. You should use it, too (if you're not already using it). Everyone should. It is awesome :) - Also, the developers largely overlap with the top posters on metaoptimize ;) Other than that I just use code accompanying papers. Why is no one else answering the question? I think would be quite interesting.
(Nov 19 '11 at 16:29)
Andreas Mueller
in china may few will use python , more choose C++ or matlab
(Nov 20 '11 at 03:26)
lpvoid
|
