-
Notifications
You must be signed in to change notification settings - Fork 0
A list of topics for a google summer of code (gsoc) 2012
Important: Expectations for prospective students
Possible Mentor: Andreas Mueller?
Possible Candidate: Vikram Kamath
Goal: The C5.0 is an algorithm used to construct m-ary decision trees. It is a successor to the C4.5 algorithm (which in turn is an extension of the ID3 algorithm), all of which were developed by Ross Quinlan. The C5.0 source (implemented in C) has been released under the GNU General Public License (GPL). The aim is to port it and hence make it a feature of sklearn. Additionally, documentation/examples can be created (I have learned from my interaction with Ross Quinlan that the documentation of the C5.0 has not been released under the GPL and is in fact, proprietary).
References:
-
- ai.stanford.edu/~ronnyk/treesHB.pdf
Possible mentor: Olivier Grisel
Possible candidate: Vlad Niculae, ?
Goal: Online or Minibatch SGD or similar on a squared l2 reconstruction loss + low rank penalty (nuclear norm) on scipy.sparse matrix: the implicit components of the sparse input representation would be interpreted by the algorithms as missing values rather than zero values.
Application: Build a scalable recommender system example, e.g. on the movielens dataset.
TODO: find references in the literature.
Possible candidate: Vlad Niculae, Immanuel Bayer, ?
Goal: Online or Minibatch NMF using SGD + positive projections (or any other out-of-core algorithms) accepting both dense and sparse matrix as input (decomposition components can be dense array only).
Application: Build a scalable topic model e.g. on million of Wikipedia abstracts for instance using this script.
References:
Algorithms for decomposing a design matrix into a low rank + sparse components.
Possible mentor: ?
Possible candidate: Kerui Min (Minibio: "I'm a graduate student at UIUC who is currently pursuing the research work related to low-rank matrices recovery & Robust PCA.")
Applications: ?
References:
- http://perception.csl.uiuc.edu/matrix-rank/home.html
- http://www.icml-2011.org/papers/41_icmlpaper.pdf (randomized algorithm supposedly scalable to larg-ish datasets)
Possible mentor: Andreas Mueller
Possible candidate: David Marek
Goal: Implement a stochastic gradient descent algorithm to learn a multi-layer perceptron, starting from https://gist.github.com/2061456.
References:
- http://en.wikipedia.org/wiki/Multi-layer_perceptron
- http://www.springerlink.com/content/4w0bab2v3qnqhwyr/
Possible mentor: Andreas Mueller
Goal: Implement a stochastic gradient descent SVM using a low-rank kernel approximation.
References:
Possible mentor: Paolo Losi, Alex Gramfort, (others?)
Goal: Implement one of the state of art methods for Generalized Additive Models Sparse Version of it is SpAM
References:
- arxiv.org/pdf/0711.4555
- http://code.google.com/p/google-summer-of-code-2011-r/downloads/detail?name=Juemin_Yang.tar.gz
- http://en.wikipedia.org/wiki/Generalized_additive_model
- http://arxiv.org/abs/0806.4115
- http://www.stats.ox.ac.uk/~meinshau/liso.pdf
Possible mentors: Alex Gramfort, Gael Varoquaux
Goal: Implement state of art methods for optimizing sparse linear models using coordinate descent.
One objective to avoid the dependency on LibLinear for the LogisticRegression model in order to allow warm restart and Elastic-Net regularization (L1 + L2)
A second objective is to improve the Lasso coordinate descent using strong rules to automatically discard features.
References:
- http://www.jmlr.org/papers/volume11/yuan10c/yuan10c.pdf
- http://www-stat.stanford.edu/~jbien/jrssb2011strong.pdf
Possible mentors: Gael Varoquaux
- Refurbish the current GMM code to put it to the scikit's standards
- Implement a core-set strategy for GMM
http://las.ethz.ch/files/feldman11scalable-long.pdf http://videolectures.net/nips2011_faulkner_coresets/