Structured Data

This is a master project for the class Structure Data hold by Florence d'Alché Buc and Slim Essid at M2 Data Siences - Université Paris-Saclay.

We worked on the following article:

Su, H., Heinonen, M., & Rousu, J. (2010, September). Structured output prediction of anti-cancer drug activity. In IAPR International Conference on Pattern Recognition in Bioinformatics (pp. 38-49). Springer Berlin Heidelberg. (http://link.springer.com/chapter/10.1007/978-3-642-16001-1_4)

The goal was to implement the methods presented in the paper and to comment the results we obtained.

References

To do that, the authors shared with us their work and we used a lot of what they did:

Here is the GitHub repository in reference: https://github.com/hongyusu/Molecular_Classification

The folder /references contains:

• the original matlab code they shared with us
• the article in reference

Thanks a lot to them.

Files & Folders

• MST_MMCRF.ipynb: implements the Max-margin Conditionnal Random Fields (MMCRF) algorithm to resolve the problem
• SVM.ipynb: implements a Support Vector Machine (SVM) classifier to to resolve the problem
• spanning_tree.py: implements the maximum-weight spanning tree algorithm (open-source code from the 'NetworkX' package: https://networkx.readthedocs.io/en/latest/_modules/networkx/algorithms/tree/mst.html#maximum_spanning_tree)
• /data_clean: contains the dataset used in the above notebooks
• /report: contains our conclusions (based on our own work). It is actually the report that we gave to our Professors
• /references: contains the original MATLAB code from Juho Rousu (Department of Computer Science, Aalto University).

Special note:

In case you wish to use any of the present material, please ask the authors to refer to the book chapter (Rousu et al, 2007) introducing MMCRF. In order to cite, please refer to: https://scholar.google.fr/scholar?hl=fr&q=Structured+Output+Prediction+of+Anti-cancer+Drug+Activity&btnG=&lr=.